Sediment Problems: Strategies for Monitoring, Prediction, and Control

TITLES RECENTLY PUBLISHED BY The International Association of Hydrological Sciences (IAHS)

In the Series of Proceedings and Reports

prepared by an IHP-III Working Group Publ.no.r~ • ' " ~

Proceedings of the symposia held dur ing the IUGG Assembly, Vancouver , August 1987:

La rge Scale Effects of Seasonal Snow Cover Publ.no.166 (1987), price $42

Fores t Hydrology a n d Wate r shed Management Publ.no. 167 (1987), price $55

The Influence of Cl imate Change and Climatic Variabili ty on the Hydrologie Regime and Wate r Resources Publ.no.168 (1987), price $55

I r r igat ion and Wa te r Allocation PuM.no.l69(1987), price $32

The Physical Basis of Ice Sheet Modelling Publ.no,170(1987), price $40

Hydrology 2000. Report of the IAHS Hydrology 2000 Working Group Publ.no.171 (1987), price $22

Side Effects of W a t e r Resources Managemen t . Report ed by an M P - m Wor' " ~ ).l72(1988),price$40

Groundwate r Monitor ing and Management . Proceedings of the Dresden Symposium, March 1987 PuM.no.173 (1990), price $55

Sediment Budgets . Proceedings of the Porto Alegre Symposium, December 1988 Publ.no.l74 (1988), price $60

Consequences of Spatial Variability in Aquifer Propert ies and Data Limitations for Groundwa te r Modelling Pract ice. Report prepared by a Working Group of the International Commission on Groundwater Publ.no. 175 (1988), price $45

Kar s t Hydrogeology and K a r s t Envi ronment Protection. Proceedings of the IAH/IAHS Guilin Symposium, October 1988 Publ.no,176(1988), price $55

Estimation of Area! Evapot ranspi ra t ion . Proceedings of a workshop held during the IUGG Assembly, Vancouver, August 1987 Publ.no. 177 (1989), price $45

Remote Da ta Transmission. Proceedings of a workshop held during the IUGG Assembly, Vancouver, August 1987 PubUo.178 (1989). price $30

Proceedings of symposia held dur ing the Thi rd IAHS Scientific Assembly, Bal t imore , M a r y l a n d , M a y 1989:

Atmospheric Deposition Publ.no. 179 (1989), price $45

Systems Analysis for Wa te r Resources Management : Closing the G a p Between Theory and Pract ice Publ.no.180(1989), price $45

New Directions for Surface W a t e r Modelling PuM.no. 181 (1989), price $50

Regional Charac ter iza t ion of W a t e r Quality Publno.182 (1989), price $45

Snow Cover and Glacier Variat ions PubUo. 183 (1989). price $30

Sediment a n d the Environment Publ.no.184 (1989), price $40

Groundwa te r Contaminat ion Publ.no.l85(1989), price $40

Remote Sensing a n d Large-Scale Global Processes Publ.no.l86(1989), price $40

FRIENDS in Hydrology. Proceedings of the Bolkesjo Symposium, April 1989 Publ.no.187 (1989). price $50

Groundwate r Management : Quant i ty a n d Qual i ty . Proceedings of the Benidorm Symposium, October 1989 Publ.no. 188(1989), price $60

Erosion, T ranspo r t and Deposition Processes. Proceedings of the Jerusalem Workshop, March-April 1987 Publ.no.l89(I990), price $40

Hydrology of Mounta inous Areas . Proceedings of the âtrbské Pleso Workshop, Czechoslovakia, June 1988 Publ.no.190 (1990), price $45

Regionalization in Hydrology. Proceedings of the Ljubljana Symposium, April 1990 Publ.no. 191 (1990), price $45

Research Needs a n d Applications to Reduce E r o s i o n and Sedimentation in Tropical Steeplands. Proceedings o f die Suva, Fiji, Symposium, June 1990 Publ.no.l92(1990), price $50

Hydrology in Mountainous Regions. I - Hydrologica l Measurements ; the W a t e r Cycle. Proceedings of two Lausanne Symposia, August 1990 Publ.no.193 (1990), out of print

Hydrology in Mounta inous Regions. II - Artificial Reservoirs ; W a t e r and Slopes. Proceedings of two Lausanne Symposia, August 1990 Publ.no.194 (1990), price $50

ModelCARE 90: Cal ibrat ion a n d Reliability in G r o u n d w a t e r Modelling. Proceedings of the symposium held in T h e Hague, September 1990 Publ.no.I95(1990), price $55

Groundwa te r Contaminat ion Risk Assessment: A G u i d e to Unders tanding and Manag ing Uncer ta int ies . Prepared by a Working Group of the International Commission on Groundwater Publ.no.196 (1990), price $40

The Hydrological Basis for W a t e r Resources M a n a g e m e n t . Proceedings of the Beijing Symposium, October 1990 Publ.no. 197 (1990), price $60

Hydrological Processes a n d W a t e r M a n a g e m e n t in U r b a n Areas . Invited lectures and selected papers of the Urban Water '88 Symposium at Duisberg, Germany, April 1988 Publ.no. 198 (1990), price $50

Soil W a t e r Balance in the Sudano-Sahelian Z o n e . Proceedings of the Niamey Workshop, February 1991 Publ.no.I99(1991), price $60

L a n d Subsidence. Proceedings of the Fourth Land Subsidence Symposium held in Houston, May 1991 Publ.no.200(1991), price $65

Proceedings of t h e symposia held du r ing t h e I U G G Assembly, Vienna, August 1991 :

Hydrology for the W a t e r M a n a g e m e n t o f L a r g e River Basins Pubi.no.201 (1991), price $55

Hydrological Basis of Ecologically Sound Managemen t of Soil and G r o u n d w a t e r PubI.no.202(I991), price $55

Sediment and S t ream W a t e r Quali ty in a Changing Envi ronment : Trends and Explana t ion Publ.no.203 (1991), price $55

Hydrological Interact ions Between A t m o s p h e r e , Soil a n d Vegetation Publ.no.204(r99I), price $60

Snow, Hydrology and Forests in High Alp ine Areas Publ.no.205(199l), price $50

Hydrology of Na tu ra l and M a n m a d e L a k e s Publ.no.206 (1991), price $50

Glaciers-Ocean-Atmosphere In teract ions , Proceedings of the St Petersburg Symposium, September 1990 Publ.no.208 (1991), price $60

Erosion, Debris f lows a n d Envi ronment in M o u n t a i n Regions. Proceedings of the Chengdu Symposium, Ju ly 1992 Publ.no.209 (1992), price $75

Erosion and Sediment T ranspo r t Moni tor ing P r o g r a m m e s in River Basins . Proceedingiof the Oslo Symposium, Augus t 1992 Publ.no.210 (1992), price $75

Application of Geographic Informat ion Systems in Hydrology a n d W a t e r Resources M a n a g e m e n t . Proceedings of the HydroGIS 93 Conference, Vienna, April 1993 Publ.no.211 (1993), price $80

Please send orders and enquiries to: I A H S Press , Inst i tute of Hydrology Office of t h e T r e a s u r e r I A H S , 2 0 0 0 Flor ida WalUngford, Oxfordshire OX10 8BB, UK Avenue N W , Washington, D C 2 0 0 0 9 , USA Telephone: +44 491 838800; telex: 849365 hydrol g Telephone: + i 202 4626900; telex: 7108229300; Fax: +44 491 832256 Fax: +1 202 3280566

P l e a s e s e n d c red i t c a r d o r d e r s ( V I S A , A C C E S S , M A S T E R C A R D , E U R O C A R D ) a n d I A H S m e m b e r s h i p o r d e r s to t h e Wai l ingford a d d r e s s on ly . A c a t a l o g u e o f pub l ica t ions m a y b e ob ta ined free o f c h a r g e f rom e i the r o f t h e a b o v e a d d r e s s e s .

1AHS1 Sediment Problems:

Strategies for Monitoring,

Prediction, and Control

Edited by

RICHARD F. HADLEY Department of Geography, University of Denver, Denver, Colorado 80208, USA

TAKAHISA MIZUYAMA Department of Forestry, Kyoto University, Kyoto, Japan

Proceedings of an international symposium held at Yokohama, Japan, 19-21 July 1993. The Symposium was organized by the International Commission on Continental Erosion of the International Association of Hydrological Sciences.

IAHS Publication No. 217

Published by the International Association of Hydrological Sciences 1993. IAHS Press, Institute of Hydrology, Wallingford, Oxfordshire OX10 8BB, UK.

IAHS Publication No. 217. ISBN 0-947571-78-7

IAHS is indebted to Department of Geography, University of Denver, and the Department of Forestry, Kyoto University, for the support and services provided that enabled the Editors to function effectively and efficiently. Without this support the proceedings of Yokohama Symposium 114 would not have been pre-published.

The designations employed and the presentation of material throughout the publication do not imply the expression of any opinion whatsoever on the part of IAHS concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries.

The use of trade, firm, or corporate names in the publication is for the information and convenience of the reader. Such use does not constitute an official endorsement or approval by IAHS of any product or service to the exclusion of others that may be suitable.

Special thanks are due to everyone who helped with the production of this Proceedings volume, and in particular to Penny Kisby who coordinated its production, and to Mrs Gail Nickels, Department of Geography, University of Denver, who produced the camera-ready copy from edited manuscripts and word-processor diskettes supplied by some authors.

The final camera-ready copy for the papers was finished, printed out, and assembled at IAHS Press, Institute of Hydrology, Wallingford, Oxfordshire, UK, by Penny Kisby.

Printed in Great Britain by Galliard (Printers) Ltd, Great Yarmouth.

Preface

The International Commission on Continental Erosion (ICCE) has organized, sponsored or co-sponsored two symposia since 1981 that were concerned primarily with the topics of erosion and sediment transport measurements, and the prediction of erosion and sediment yield. The International Symposium on Erosion and Sediment Transport Measurement (IAHS Publication no. 133) was held in Florence, Italy, in June 1981. In July 1982 a symposium was held in Exeter, UK (IAHS Publication no. 137) that dealt with the explanation and prediction of erosion and sediment yield. Both of those symposia focused on problems of measurement, monitoring, and prediction of processes of erosion and sediment yield. The themes of the Yokohama Symposium Proceedings published in this volume are also concerned with sediment problems related to monitoring, prediction, and erosion control.

The papers that comprise the Yokohama Symposium may be divided into six general groups: (1) erosion and sediment yield; (2) landslides and pyroclastic flows: characteristics and controls; (3) deposition processes in reservoirs; (4) modelling and monitoring of sedimentation and erosion processes; (5) soil erosion, sediment losses, and drainage basin characteristics; and (6) monitoring processes of erosion and sediment transport. None of these general groups contains a sufficient number of papers to cover adequately the sediment problems that face us in the wide variety of climates and environments that we live in. There are, however, papers that represent problems in arid, semiarid, and humid regions. These papers present a broad perspective on contemporary studies of erosion and sediment yield.

In the group of papers on erosion and sediment yield there are many interesting approaches to a variety of problems. In the semiarid southwestern United States, sediment yields are estimated using different types of sediment sampling equipment. These estimates are compared with estimates made using the Revised Universal Soil Loss Equation (RUSLE). In China, rates of soil erosion are related to physical-chemical properties of soil, climate, topography, vegetation cover, and human activity. In Tanzania, seasonal variation of sediment yield on interrill areas, in a semiarid region, is being investigated.

In investigations of landslides, pyroclastic flows, and debris flows, methods of modelling and monitoring are discussed. Hydrological observations that relate shallow landslides to rock type and regolith thickness are presented. Pyroclastic flows reach farther downstream depending on the volume of ash that is erupted from a volcanic vent. Debris flows occur even in periods of low rainfall; sediment concentration depends on the volume of runoff. Models are presented for the numerical solution of landslides and pyroclastic flows, and areas that are susceptible to landslides are designated using Landsat Thematic Mapper data.

Several papers consider the use of reservoirs in water management, and the

VI Preface

influence of reservoirs on depositional processes. A one-dimensional numerical model is presented for a reservoir in the Carpathian region of southern Poland which demonstrates the influence of water management on the distribution of sediment in the reservoir and on the backwater profile. The influence of recent climate, sediment particle-size, and reservoir shape and depth are also discussed.

Several innovative methods for monitoring and modelling are presented in this volume. An on-site sediment prediction model is developed for forest roads and timber harvest areas that will assist managers in the rehabilitation of areas disturbed by timber production. Among the recent techniques of monitoring erosional features is the use of an airborne laser altimeter to provide input to natural resources models.

We are hopeful that the papers briefly introduced here and the research findings of all the authors will stimulate discussion of all aspects of sediment problems and their potential solution.

Co-convenors:

Richard F. Hadley Department of Geography, University of Denver

Denver, Colorado 80208, USA

Takahisa Mizuyama Department of Forestry

Kyoto University, Kyoto, Japan

Contents

Vil

Preface by Richard F. Hadley & Takahisa Mizuyama

1 Erosion and Sediment Yield

Sediment yield in a semiarid basin: sampling equipment impacts /. R. Simanton, W. R. Osterkamp & K. G Renard 3

Incorporating social and environmental factors into a regional soil erosion system analysis Gu Hengyue, Qian Xiarong, Ian Douglas & He Min 11

Characteristics and control of soil erosion in Hubei Province, China Yao Huaxia 23

Seasonal variation of sediment yield on a gentle slope in a semi-arid region, Tanzania S. Onodera, J. Wakui, H. Morishita & E. Matsumoto 29

Correlation analysis of rainfall: classification method of rainfall in view of sediment yield and transport Tom Shimada & Kuniaki Miyamoto 39

2 Landslides and Pyroclastic Flows: Characteristics and Controls

Underlying rock type controls of hydrological processes and shallow landslide 47 occurrence Yuichi Onda

Characteristics of pyroclastic flows and debris flows accompanying the Mt Unzen-Fugendake eruption Hiroshi Ikeya & Yoshiharu Ishikawa 57

Model of pyroclastic flow and its numerical solution S. Yamashita & K. Miyamoto 67

Study for prediction of occurrence of hillside landslides S. Hiramatsu, T. Mizuyama, S. Ogawa & Y. Ishikawa 75

Inference of landslide susceptible areas by Landsat Thematic Mapper data L. Samarakoon, S. Ogawa, N. Ebisu, R. Lapitan & Z. Kohki 83

3 Deposition Processes in Reservoirs

An overview of reservoir sedimentation in some African river basins M. M. A. Shahin 93

Investigation on sediment deposition in a designed Carpathian reservoir K. Banasik, J. Skibinski & D. Gârski 101

Influence of recent climate on sedimentation in Burrinjuck Reservoir R. Srikanthan & R. J. Wasson 109

A simulation model for sedimentation process in gorge-type reservoirs T. Okabe, S. Amou & M. Ishigaki 119

Reservoir sedimentation for different size particles Ryoei ho 127

V l l l Contents

4 Modelling and Monitoring of Sedimentation and Erosion Processes

Development of an onsite sediment prediction model for forest roads and timber harvest areas Peter R. Robichaud, Randy B. Foltz & Charles H. Luce 135

Development of debris flow Takahisa Mizuyama, Sumiji Kobashi, & Ou Guogiang 141

Measurement of sediment transport components in a drainage basin and comparison with sediment delivery computed by a soil erosion model B. Hasholt & M. Styczen 147

Monitoring streambank and gully erosion by airborne laser Jerry C. Ritchie, Joseph B. Murphey, Earl H. Grissinger & Jurgen D. Garbrecht 161

Erosion and runoff monitoring and modelling in a semiarid region of Brazil V. S. Srinivasan & C. O. Galvao 167

5 Soil Erosion, Sediment Losses and Drainage Basin Characteristics

Forest clearcutting and site preparation on a saline soil in East Texas: impact on sediment losses Alexander Sayok, Mingteh Chang, Kenneth G. Watterston 111

An investigation of the influence of edaphic, topographic and land-use controls on soil erosion on agricultural land in the Borrowdale and Chinamora areas, Zimbabwe, based on caesium-137 measurements T. A. Quine, D. E. Walling & O. T. Mandiringana 185

Experimental study on the Theological properties and hydrological mechanism of the occurrence of a volcanic mudflow Y. Taniguchi 197

Environmental and hydrological implications of the development of multipurpose reservoirs in some catchments of Kenya: meeting Kenya's water demands by the year 2010 George S. Ongwenyi, Shadrack M. Kithiia, F. O. Denga & P. O. Abwao 207

An overview of the soil erosion and sedimentation problems in Kenya George S. Ongwenyi, Shadrack M. Kithiia & Fred O. Denga 217

6 Monitoring Processes of Erosion and Sediment Transport

A method for estimating soil erosion caused by surface runoff using sloping lysimeters K. Banzai & Y. Hayase 227

Motion, debris size and scale of debris flows in a valley on Mount Yakedake, Japan H. Suwa, K. Okunishi & M. Sakai 239

Examining the transition from sediment transport in water to mass movement C. W. Rose, A. L. Presbitero & R. K. Misra 249

A comparative study on suspended-sediment discharge initiated by snow- or glacier-melting Kazuhisa Chikita 257

Modelling soil erosion in arid zone drainage basins K. D. Sharma, R. P. Dhir & J. S. R. Murthy 269

Influence of heterogeneous sediment transport on the function of sediment control of a check dam H. Malta 277

1 Erosion and Sediment Yield

Sediment Problems; Strategies for Monitoring, Prediction and Control (Proceedings of the Yokohama Symposium, July 1993). IAHS Publ. no. 217, 1993. 3

Sediment yield in a semiarid basin: sampling equipment impacts

J. R. SIMANTON United States Department of Agriculture, Agricultural Research Service, Southwest Watershed Research Center, 2000 E. Allen Road, Tucson, Arizona 85719, USA

W. R. OSTERKAMP United States Department of Interior, Geological Survey, Denver Federal Center, Mail Stop 413, Lakewood, Colorado 80225, USA

K. G. RENARD United States Department of Agriculture, Agricultural Research Service, Southwest Watershed Research Center, 2000 E. Allen Road, Tucson, Arizona, 85719, USA

Abstract Sediment yields from two small semiarid subbasins within the Walnut Gulch Watershed in southeastern Arizona, USA, are estimated using three types of sediment sampling equipment. Because efficiency of sampling equipment affects these estimates, measured sediment yields from each subbasin were compared by sampling method and then to estimates using the Revised Universal Soil Loss Equation (RUSLE). Sediment yield versus storm runoff volume were related for each sampling method and results showed that as sampling equipment became more efficient, regression line slope increased indicating an increase in subbasin sediment yield. The magnitude of difference between RUSLE estimates and measured sediment yields changed with sampling method.

INTRODUCTION

Sediment-yield estimates and erosion prediction models depend on experimental data collected from field and small drainage basin areas. Fluvial sediment samplers range from those that measure suspended loads to those that measure time-integrated total loads. Sediment-yield measurements are not only affected by the portion of flow being measured but also the sampling action and technique. Results using one sampler may not be comparable to those using another, even though each sampler collects from the same part of the water column. Because of vertical and horizontal spatial variability of sediment concentration within the stream flow, samplers approach the spatial concentration problem with different techniques. Sampler efficiency can have a major impact on the accuracy of the sediment-yield estimates and the subsequent validity of erosion model estimates.

This study determined whether sediment yields changed with sampling equipment and how changes modified erosion model estimates. The study was conducted in subbasins of the 150 km2 Walnut Gulch Experimental Watershed in southeastern Arizona, USA (31°43'N, 110°41'W) (Fig. 1). The watershed is representative of semiarid brush and grass rangeland of the southwestern USA and is transitional between the Chihuahuan and Sonoran Deserts. Annual precipitation averages about 300

4 / . R. Simanton et al.

Fig. 1 Location map of the Walnut Gulch Experimental Watershed, Arizona, United States of America.

mm: 70% occurs as thunderstorms from July to mid-September. Runoff during this season accounts for 99 % of the annual total.

Runoff and sediment-yield data from adjacent subbasins, Lucky Hills 3 (LH3)(3.68 ha) and Lucky Hills 4 (LH4)(4.53 ha), were used in this study. The subbasins have similar vegetation and topography, but different channel characteristics and parent material of the soils. The main channel of LH3 is deeply incised, relatively straight, and actively eroding; LH4 has a meandering channel with gently sloping banks. This channel difference could be a function of soil type in which the channel has formed. The LH3 channel was formed in a fine textured, relatively rock-free soil whereas the LH4 channel formed in a soil similar to that of both subbasins. Sediment yields from LH3 are about 3 times greater than from LH4, the difference being a greater amount of channel erosion in LH3 (Osborn & Simanton, 1989). Livestock have been excluded from both subbasins since 1962 but little change in vegetation type and density and surface ground cover has been measured. Shrubs less than 1 m high and with a canopy area of about 35% dominate the subbasins. Vegetation consists of creosote bush (Larrea tridentata), white-thorn (Acacia constricta), tarbush (Flourensia cernua), snakeweed (Gutierrezia Sarothrae), and burroweed (Aplopappus tenuisectus).

Pedogenesis has resulted in medium-depth (1.2 m), well-drained, calcareous loams. Calcretes are common at depths of 0.5 to 2 m. The uppermost 10 cm of the soil profile contains up to 60% gravel but underlying parts of the profile usually contain less than 40% gravel (Gelderman, 1970). Surface rock fragment cover (erosion pavement) ranges from negligible on shallow slopes to over 70% on the very steep slopes (Simanton et al, in press).

Sediment yield in a semiarid basin 5

INSTRUMENTATION

Event runoff and sediment-yield data were collected from 1973 through 1989. During the period 1973 to 1977 broad-crested V-notch weirs were used for runoff measurement. Suspended sediment passing over the weir was measured with an automatic pump sampler (Allen et al, 1976) equipped with a depth-integrating sampling tube. Bed load was measured after each runoff event as trapped sediment in the stilling basins behind the weirs. The weir stilling basins were large enough that all bed load was trapped. The weirs were replaced in 1977 at LH3 and in 1978 at LH4 with supercritical flow flumes (Smith et al, 1981) equipped with total-load automatic traversing slot samplers (Renard et al, 1986).

Three methods were used to measure sediment yields: (a) depth integrated pump sampler (DIP), (b) DIP plus the bedload trapped sediment behind the weir (DIPT), and

Super Critical Flume and

Traversing Slot Instrumentation

Slot Drive Chain

a:

Fixed Slots at Uniform Spacing

Cross Section Downstream

Row Depth Recorder

Sample Storage ~ / Housing

Bj Sample Bottle

Sample Storage Housing

Row Depth Recorder

4 % Floor Slope

Rxed Slots

Longitudinal Section

Fig. 2 Schematic of the traversing-slot sediment sampler and associated instrumentation.

6 J. R. Simanton et al.

(c) total-load traversing slot sampler (TS). The DIP method takes periodic sediment samples, throughout the hydrograph, that consist of fine suspended sediment from the water column at the intake tube. Thus, concentrations may not be representative of the entire flow width. It is assumed that the sample is vertically integrated because the bottom-pivoting sampling tube in the flow profile samples from equally spaced holes in the tube. The DIPT and DIP methods are similar except that the DIPT method includes coarse sediment, trapped behind the weir, that was measured and removed after each runoff event. The trapped sediment weight was added to the DIP sediment to give a total load for the event. The TS method uses a supercritical flume with a traversing slot sediment sampling device that diverts a water-sediment mixture at the flume discharge into fixed slots (Fig. 2). The water-sediment mixture then flows into sample bottles within the sample storage housing. The traversing slot was assumed to provide a horizontally and vertically-integrated, total-load sediment sample. Temporal changes in sediment concentration were quantified by periodic sampling throughout each hydrograph.

DATA ANALYSIS AND RESULTS

Direct comparison of sediment yields by the DIP and DIPT methods were made, but neither method can be compared directly with the TS method because its equipment replaced the V-notch weirs. Comparisons of DIP and DIPT to TS were made by relating storm runoff volume to sediment yield. Total sediment yields, in kilograms per hectare (kg/ha), for selected storms were regressed with storm runoff volume (depth over the subbasin area), in millimeters (mm), to determine differences in sediment yield among the sampling methods. In the regression analysis, the regression was forced through the origin and produced the equations given in Table 1. The coefficients of the equations (sediment yield (kg/ha) per unit of runoff volume (mm)) for the two subbasins indicate that sediment yields increased when the sampling method changed from DIP to DIPT. Changing from the DIPT to the TS method did not affect sediment-yield measurements at LH3 (Fig. 3(a)) but the sediment yields at LH4 showed an increase (Fig. 3(b)). However, by eliminating one point (runoff = 14.5 mm and sediment = 626 kg/ha) in the DIPT regression analysis for LH4, the regression

Table 1 Regression equation and coefficient of determination (R2) for the storm sediment yield (kg/ha) v*. storm runoff volume (mm) relations of the Lucky Hills subbasins.

Lucky Hills 3 Method Equation

DIP Y= 83.1 X DIPT Y= 187.8 X

TS Y= 193.1 X

R2

0.91 0.95

0.88

(LH3) N

20 20

30

Method

DIP DIPT

DIPT(w/o TS

Lucky Hills Equation

Y= 29.0 X Y= 59.5 X

) Y= 79.5 X Y= 80.1 X

4 (LH3) R2 N

0.88 20 0.73 20 0.73 19 0.86 13

Y= Sediment yield (kg/ha), X= Runoff Volume (mm) DIP = Depth integrated pump sampler DIPT = Depth integrated pump sampler plus trapped sediment DIPT(w/o) = DIPT without one data point (x= 14.5, y= 626) TS = Traversing slot sampler

Sediment yield in a semiarid basin 7

(a)

3000

— 2000

1000-

Walnut Gulch, . Lucky Hills 3

Figure 3a

"

•

_ 0

S

p><* s

fêZ^^*' * _ jp, J

USA

o / / - '

S' 0 * ^ * ^

1

TSO /

/ / ' 'DIPT+

4-

^ DIP *

(b)

10 Runoff (mm)

15 20

500

| xs to

"l Walnut Gulch, USA Lucky Hills 4 / T s 0 Figure 3b /

/ o /

o / / + /

/ of /

v.- ^-^»

, DIPT +

'

+

^ DIP *

10

Runoff (mm) 15 20

Fig. 3 (a) Runoff volume vs. sediment yield of three sampling methods on the Lucky Hills 3 subbasin. (b) Runoff volume vs. sediment yield of three sampling methods on the Lucky Hills 4 subbasin.

coefficient became 79.5 (Table 1) which would plot very similar to the TS regression line. Student's t-tests of mean sediment yield showed that measured sediment yields were significantly (P < 0.05) greater when measured with the DIPT and TS methods than with the DIP for both subbasins; sediment yields were not significantly (P < 0.05) different between the DIPT and TS methods for either subbasins. Runoff volumes during the periods of different sampling methods were not significantly (P < 0.05) different for either subbasin. Annual sediment yields (kg/ha/yr) from both subbasins were compared to estimates of soil loss using the Revised Universal Soil Loss Equation (RTJSLE) (Renard et al., 1991). For this study it was assumed that soil loss equals sediment yield and that each subbasin had a sediment delivery ratio of 1.0. Comparisons of the RUSLE estimates to measured sediment yields showed differences

J. R. Simanton et al.

Table 2 Measured annual vs. RUSLE estimated soil losses (kg/ha/yr) and the difference (measured-estimated) from Lucky Hills 3 subbasin. (Measured data from Osborn & Simanton, 1989.)

Year

1973 1974 1975 1976

1973 1974 1975 1976

1977 1978 1979 1980

Method

DIP DIP DIP DIP

DIPT DIPT DIPT DIPT

TS TS TS TS

Lucky Hills 3 (LH3)

Measured

1330 880 1260 1020

2780 4860 8590 2420

6810 2000 470 560

I Estimated Di (kg/ha/yr)

680 850 1960 310

Mean difference

680 850 1960 310

Mean difference

870 480 270 290

Mean difference

.fference

650 30

-700 710 170

2100 4010 6630 2110 3710

5940 1520 200 270 1980

DIP DIPT TS

Depth integrated pump sampler Depth integrated pump sampler plus trapped sediment Traversing slot sampler

Table 3 Measured annual vs. RUSLE estimated soil losses (kg/ha/yr) and the difference (measured-estimated) from Lucky Hills 4 subbasin. (Measured data from Osborn & Simanton, 1989.)

Lucky H i l l s 4 (LH4)

Year Method Measured Estimated Difference (kg/ha/yr)

1973 DIP 460 630 -170 1974 DIP 390 770 -380 1975 DIP 1980 1840 140 1976 DIP 270 290 -20 1977 DIP 370 820 -450

Mean difference -180 1973 DIPT 780 630 150 1974 DIPT 1680 770 910 1975 DIPT 3180 1840 1340 1976 DIPT 700 290 410 1977 DIPT 2980 820 2160

Mean difference 990 1978 TS 180 440 -260 1979 TS 0 240 -240 1980 TS 220 270 -50

Mean difference -180 DIP = Depth integrated pump sampler DIPT = Depth integrated pump sampler plus trapped sediment TS = Traversing slot sampler

Sediment yield in a semiarid basin 9

that varied considerably in magnitude (Tables 2 and 3). For both subbasins, RUSLE estimates were similar to sediment yields measured by the DIP method (because of low degrees of freedom, Student's t-test were not attempted). Greater differences occurred when RUSLE estimates were compared with the total-load measurements of sediment yields. Years of greater than normal water and sediment discharge measurements showed larger differences with RUSLE estimates than did years with lower than normal measured water and sediment discharges. A possible explanation is that much of the coarse sediment does not reach the basin outlet during low-volume runoff events, and that the assumption of unity for the sediment-delivery ratio is inappropriate. A delivery ratio greater than unity may be appropriate for these subbasins because of the relatively large amount of gully erosion observed in LH3.

In all field studies the design should ensure, as practically and economically as possible, that complete measurement of data is made; ie. suspended and bed load both need to be sampled to get accurate information about the sediment yield of the Lucky Hills subbasins. Limitations of sampling techniques should be recognized and deficiencies be included in interpretations of results.

REFERENCES

Allen, P. B., Welch, N. H., Rhoades, E. D., Edens, C. D. & Miller, G. E. (1976) The Modified Chickasha Sediment Sampler. USDA, ARS ARS-S-107.

Gelderman, F. W. (1970) Soil Survey, Walnut Gulch Experimental Watershed, Arizona. Special Report, USDA-SCS.

Osborn, H. B. & Simanton, J. R. (1989) Gullies and Sediment Yield. Rangelands, 11:51-56. Renard, K. G., Simanton, J. R. & Fancher, C. E. (1986) Small watershed automatic water quality sampler. Proc.

4th Fed. Interagency Sedimentation Conference, Las Vegas, Nevada, March 1986, Vol. 1, pp. 1.51-1.58. Renard, K. G., Foster, G. R., Weesies, G. A. & Porter, J. P. (1991) RUSLE Revised universal soil loss equation.

Journal of Soil and Water Conservation 46:30-33. Simanton, J. R., Renard, K. G., Christiansen, C. M. & Lane, L. J. Spatial distribution of surface rock fragments

along catenas in semiarid Arizona and Nevada, USA. Catena Supplement No. 25, in press. Smith, R. E., Chery, D. L., Jr, Renard, K. G. & Gwinn, W. R. (1981) Supercritical flow flumes for measuring

sediment-laden flow. USDA Agriculture Technical Bulletin No. 1655.

Sediment Problems: Strategies for Monitoring, Prediction and Control (Proceedings of the Yokohama Symposium, July 1993). IAHS Publ. no. 217, 1993. 11

Incorporating social and environmental factors into a regional soil erosion system analysis

GU HENGYUE, QIAN XIAORONG Chongqing University, Chongqing, Sichuan, China

IAN DOUGLAS & HE MIN Department of Geography, University of Manchester, Manchester M13 9PL, UK

Abstract In the upper Changjiang and middle Huanghe the rate of soil erosion depends on climate and hydrometeorological conditions, the physical-chemical properties of the soil, topography and geomorphology, vegetation cover and human activity. In large basins, such factors as the position of the main area of heavy rain, the manner in which rainstorms move across the basin, localised intense rainfalls and the number of storms on a given day, all affect soil detachment and transport and the sediment delivered to the main river. In Sichuan, the dumping of over 33 Mt/yr of industrial waste into the Changjiang adds to pollutant levels, sediment loads and flooding.

INTRODUCTION

Soil erosion is a biophysical outcome of the integration of forces of soil detachment and transport with forces of resistance and deposition. The process is therefore the interaction of two sub-systems: an erosional dynamic system (rainfall, gravity, wind and frost action) and a land surface condition system (geology, geomorphology, soil, vegetation, cropping systems and management). This paper uses data from areas of severe soil erosion in China to analyse the factors involved in the assessment of regional soil erosion systems. Since 1984, varied construction activities have increased areas of severe soil erosion in Chongqing city by about 67 km2/yr, with sediment yields rising by 1.5 Mt/yr.

FACTORS AFFECTING SOIL EROSION

Soil erosion is a complex dynamic process involving three sub-systems: land environment, erosion dynamics and human activity. These subsystems reflect the geology, geomorphology, climate and agricultural land use in a given region. For a region, the process of soil erosion may be said to be:

Ggif)

Se(t) = Cm(t) ( 1 )

Ap(t)

12 Gu Hengyue et al.

Se{ti) = [Gg(ti)f)Cm(ti)f)Ap(ti)/ti] (2)

Soil erosion amount in the time period tl — t2 Se(T):

h

Se(T) = { Gg(f)f\Cm(t)r\Ap(t)dt (3) fi

where: Se(t): net result of soil erosion processes in a region; Gg{t): geologic and geomorphic processes; Cm{t): climate and hydrometeorological processes; Ap(t): agricultural production; and Se(ti): the erosion at time tt.

Results from the upper Changjiang and middle Huanghe (Gu & Douglas, 1989; Zhao et al, 1989; Gu & Ai, 1987) have shown that the rate of soil erosion depends on the physical-chemical properties of the soil, climatic conditions and weather, topography and geomorphology, vegetation cover and human activity, as equations (1), (2) and (3) imply.

Table 1 The relationship of soil type and erosion in Wufengdi watershed, inner Mongolia.

Soil slope angle

wind blown sand soil type: loess sand and silt

9

9 9

Ks value

mms-1

3.0-3.5

2.1-2.8 0.6-1.2

erosive rain

mm

190.6

190.6 190.6

runoff depth

mm

15.0

41.7 64.8

runoff coe. %

7.86

21.90 34.00

sediment yield

tha-1

4.31

21.19 42.72

V: vegetation

Table 2 The relation of erosion and rock property and soil quality in Chongqing, China.

place area rock soil geomorphology slope land erosion ha degree use t/ha

Tianjia 0.015 J2S light middle hilly 5 W-SP 2.4 soil

Qianjin 0.019 J3S middle middle hilly 3 W-SP 7.2 soil

Qianjin 0.027 J3S M-L middle hilly 8 W-SP 7.4 soil

W:wheat SP: sweet potato M-L: middle and light

Social and environmental factors and regional soil erosion analysis 13

Geologic and soil factors

Surficial geologic and soil conditions directly influence soil erodibility. The properties and degree of weathering of earth surface materials affect permeability, cohesion and resistance to erosion. Data from north and south China (Zhao et al., 1989; Wohlke et al., 1988; Gu & Ai, 1987) show the variability of erosion with soil type. In inner Mongolia, under uniform slope, vegetation and rainfall conditions, soil permeabilities 3 to 5 times greater in one soil than another cause runoff 4.3 times greater and sediment yield 10 times greater (Table 1). In Chongqing, in the upper Changjiang basin, different soils under uniform slope and cultivation, have threefold differences in sediment yield (Table 2).

Climatic and hydrologie factors

The Universal Soil-Loss Equation recognises the role of rainfall amount, form, intensity and erosivity in producing runoff and splash erosion. However in large basins, such factors as the position of the main rainfall area, the way in which rainstorms move across the basin, localised intense rainfalls and the number of storms on a given day, can all affect soil detachment and transport and the sediment delivered to the main river.

In the 156 142 km2 Jialinjiang basin, upper Changjiang, two storm events in which rain fell over the same general area and produced the same runoff (Table 3(a)) caused sediment yields differing by 1.9 times because rainfall intensities were 2.6 times greater in one storm. When the area of rainfall is not identical, sediment yields differ by as much as 5.3 times (Table 3(b)). In large basins, the area of maximum rainfall intensity has a major effect on sediment delivery. In this basin, differences in rainfall intensities and numbers of storms, despite giving the same monthly runoff, cause wide variations in sediment yield (Table 3(c)).

Topographic and geomorphic factors

Topography and geomorphology directly affect rainfall, runoff, rate of water erosion and intensity of wind erosion. Slope angle is the dominant parameter reflecting topography and geomorphology. Under otherwise uniform conditions and varying slopes below 40 degrees, soil erosion increases as slope increases (Tables 4 and 5).

Erosion plot data show that at a constant slope angle, length of slope affects soil erosion. But the effect of length of slope is more complex than that of slope angle because the relationship between length of slope and slope angle is disturbed and restricted by local environmental conditions and rainfall intensity. In wet regions or during heavy rainfalls, the longer the slope, the more water accumulates and higher is the erosivity and transporting capacity of the flow. In arid or low rainfall regions, the longer the slope, the greater the infiltration of water, and the smaller the surface runoff and soil erosion.

14 Gu Hengyue et al.

Table 3(a) Comparing rainfall with runoff and sediment under different rainfall intensity.

station area runoff runoff rainfall sediment event intensity

km2 time Gm3 mm/day Mt

Beipei 156142 2/8-11/8/84 6.78 26.8 6830 Beipei 156142 1/9-10/9/78 6.88 10.1 3630

Table 3(b) Comparing rainfall, runoff and sediment under different position of main rainfall area.

station runoff rainfall runoff position sediment event main rain time mm B m3 area Mt

Beipei 2/8-11/8/84 80.5 6.78 high erosion 68.3 potential

Beipei 19/9-28/9/78 79.6 6.68 low erosion 12.9 potential

Table 3(c) Comparing rainfall, runoff and sediment under different rainfall, rainfall intensity and rain day.

Time month month month average >= 30mm >= 50mm Max. runoff sediment rain rain days days storm

intensity Bm3 Mt mm mm/day mm

My, 1981 July, 1966

Aug., 1983 July 1978

Sep., 1968 Aug., 1984

11.6

12.1

9.77

10.1

30.4

30.3

29.3

178.0

18.6

80.9

25.9

236.0

102.5

218.7

126.4

124.3

139.1

151.9

9.3

10.9

11.5

12.4

10.7

13.8

1

4

0

2

3

4

0

2

0

1

0

2

48.0

92.6

23.8

71.7

49.0

61.6

Data from the Tanjiaba station on the Jialinjiang, with a basin area of 9538 km2

Vegetation cover

The relationship between vegetation and soil erosion is readily apparent. In south China water erosion predominates (Table 6), but in north China wind erosion is more frequent (Table 7). Despite this difference in process, the same rule applies: as

Social and environmental factors and regional soil erosion analysis 15

Table 4 Variation of runoff and sediment yield with slope in the upper Changjiang basin.

slope

angle

5 10 15 20 25

Yulin Yufeng P mm

642.5 642.5 642.5 642.5 642.5

R erosion m3 ha-1 t ha-1

1930.5 37.1 1682.3 51.2 2569.4 87.5 2153.1 116.3 1892.7 159.0

Xulin Shannin R ( m3ha-l

631.6 813.3

1031.6 1133.9 1760.0

srosion tha-1

7.2 68.7 92.7 109.5 148.3

Jiangjln erosion

tha-1

13.1 19.2 39.3 57.9 77.7

observated data. P: rainfall R: runoff

Table 5 Comparison of slope and sediment yield in the Huangpuchuan basin, inner Mongolia.

Wufengdigou Wubujingou slope sediment slope sediment angle yield angle yield

t ha-1 t ha-1 _ _ _ _ _ _

12 11.087 17 59.699 15 22.766 19 101.278 30-50 100-300

Table 6 Runoff and erosion under different vegetation cover in Jianyang county of Sichuan Province.

place cover runoff erosion surface lowering (%) m3ha-l Tha-1 (mm)

waste 20) 4ÎI 125 0.166 slope

bush 90 132 0.45 0.033

Table 7 The relation of grass cover and soil erosion in Wufengdi watershed, inner Mongolia.

grass cover (%) < 1 24 28 49 67 72

erosion (t ha-1) 9.246 5.142 3.650 1.750 0.732 0.466

Gu Hengyue et al.

Land surface env i ronment

Large scale human activi ty

Erosion dynamic

T—1 Geology | |

[ Soil |

1 Vegetation |—'

1 Logglng.grazing J—j

Reclamation cultivation J —

1 Irrational land use h-

j Construction.mining |—

1 1 Grass and tree planting 1—

j Conservation ploughing j

1 Soil improvement [—

j Hydraulic engineering |—

| Water flow |—

J Wind action j —

1 1 Gravity |—

} Freezing and thawing \~

Soil erosion

Accelerated erosion

Natural erosion

Soil conservation - Erosion

control

. Erosion Process

Fig. 1 Soil erosion and control system.

vegetation increases, erosion of the land surface decreases. Not only do natural forests and grasses have the function of decreasing erosion on the land surface, but man-made vegetation, like crops in fields, performs the same function. In inner Mongolia and Helongjiang, the C value (the index of reflecting vegetation cover in the USLE) varies not only with the crop cover, but also with the type and growing period of individual crops (Table 8).

Agricultural land use

Human activity affects soil erosion mainly by agricultural and construction activities. The effects may be either positive or negative. The positive effects are expressed by

Table 8 C value in Keshan county of Helongjiang Province, China.

Crop grown periods seeding seedling ripe harvest average

Corn Sorghum Millet Bean average

0.49 0.50 0.41 0.88 0.57

0.31 0.54 0.50 0.43 0.45

0.16 0.67 0.48 0.09 0.35

0.28 0.25 0.25 0.16 0.24

0.31 0.49 0.41 0.39 0.40

Social and environmental factors and regional soil erosion analysis 17

200

150

100

1 2 3 4 5 6 7 8 9 10 11 12

Fig. 2 Soil erosion over a year in Yanting, Jialinjiang basin.

-— Rainfal!

•+• Runoff

•*• Erosion 1

H*~Erosion2

such actions as cultivation on steep slopes, logging of forests, excessive grazing, and mining which aggravate soil erosion. The negative effects, which reduce erosion, are field improvements, planting of grasses and trees, establishment of engineering structures to trap sediment, and water and soil conservation measures.

In areas of severe water loss and soil erosion, periods of bare soil or reduced crop cover during a year would affect soil erosion. Soil erosion in the Yangting

Table 9 The effect of urban and rural activities on soil erosion in 11 counties of Sichuan Province.

items number new erosion area (km2)

annual erosion amount (Mt)

Mining Construction made stone road building logging other total

3947 15022 66423 23944(km) 35406 94259

374.62 363.48 377.03 734.72

565.82 438.72

2854.39

16.5 5.53 5.98 26.78 5.59

14.58 74.97

18 Gu Hengyue et al.

experimental watershed in the Jialinjiang basin, shows two peaks (Fig. 2), the first peak in May accounting for 58% of the annual total, the second peak in July to September corresponding to seasonal rainfall and runoff peaks. In May, rainfall only accounts for 13% of the total wet season rainfall, runoff is only 3% of the total seasonal runoff, but sediment yield is 58% of the total for the year. The reason is that after harvesting, crops are sown in late spring, crop cover is very low, most fields are bare, soil is dry and loose, then soil is scoured easily by flow.

Urban and rural construction activities

Human activities, such the building of houses, road construction, quarrying, dam construction, cuttings and embankments, tree felling and waste disposal, change land surface features and vegetation cover. Waste material from these engineering works often causes problems such as debris flow and severe soil erosion. In Sichuan, over 33 Mt/yr of industrial waste is dumped into rivers causing not only pollution, but aggravating flooding. In 11 Sichuan counties, human activity from 1984 to 1986, including mining, quarrying, logging and road construction, increased the area of severe erosion by 2854.39 km2, accounting for 1.64% of the total area of these counties, or a total sediment yield of 74.97 Mt/yr (Table 9).

COMPREHENSIVE ASSESSMENT OF SOIL EROSION

Data from Chongqing are used to illustrate a correlation method of analysing soil erosion. (a) Using seven variables (jx, ..., y7) comprising the parameters geomorphological

type, slope, ruggedness, altitude, surface materials condition and land use as and relating the dependent variable erosion intensity for a statistical sample of 38 sites (z'j, ..., i38) in the hills northwest of Chongqing City, the following calculation was made for each site:

M

Xj = E V * d = 1,2,3-.38;; = 1,2,..7) (4) *=i

where Sk represents the percentage of the total site area occupied by each class (k grade) (e.g. each of seven slope categories) of each variable at each site; and Pk

represents an assigned numerical value allocated to each class of each variable at each site (e.g. 0 to 5° slope has an assigned value of 1, a 5° to 10° slope a value of 2 ...). The assigned values would be decided by experts based on field experience and observations.

(b) Multiplying Sk area percentage of every grade in i sample by the assessments value of the corresponding grade for every grade, and summing all the conceptualized values in i sample gives a characteristic value of assessment factors and sets up a matrix for correlation analysis:

Social and environmental factors and regional soil erosion analysis 19

X=

cn xl2...xhn

X2\ X22---X2m

Xnl Xn2- • -Xnm

(5)

(c) The following steps eliminate differences of parameter values.

X'ij = ^ ^ (i = 1,2,....,«; j = 1,2,...,m) Dk

where

Xk = iyxik

n

Y,(Xik-Xky2

Dk = ; = 1

(6)

(7)

(8)

n-\ (d) Coefficient of correlation matrix

n

rij = (£ X'ki * X'kj)ln (ij = 1,2,... ,m)

giving a correlation matrix

(9)

R =

r i l r\Vr\m

r2l r22...rlm

J" Y Y

ml ml"' mm

(10)

Table 10 Correlation matrix of environmental factors influencing erosion in Chongqing.

1. Geomorph.. type 2. Slope 3. Surface materials 4. Height 5. Ruggedness 6. Land Use 7. Erosion Intensity

2 0.781

3 -0.046 -0.082

4 0.088 -0-160 -0.004

5 0.717 0.612 -0.024 0.221

6 0.618 -0.789 -0.070 -0.270 0.221

7 0.520 0.701 -0.300 -0.287 0.457 0.796

20 Gu Hengyue et al.

(e) The data in Table 10 show that the largest coefficient of correlation between erosion intensity and the six factors in the matrix is that with land use, r16 (0.796), indicating the importance of human activities for soil erosion intensity in Chongqing city. Soil erosion intensity is also closely associated with slope, r-72 (0.701); with good correlations with geomorphological type, r71 (0.520); and

Geomorphology

D.502"\ .

- 0 . 2 0 7 / ^

Height

Slope

0.701

Erosion intensity

0.457

Terrain ruggedness

Ground surface characteristics

^ ^ ^ 0 . 3 0 0

^ \ 0 . 7 9 €

Land use

Fig. 3 Relationship between erosion intensity and environmental factors.

ruggedness, r15 (0.457). Thus people's activities clearly have a greater effect than natural factors on erosion processes on the present land surface in Chongqing (Fig. 3). Overall, the factors promoting increased erosion include grazing, forest clearance,

logging, mining, industrial and residential construction, and waste disposal, while those counteracting erosion include improvements to field boundaries, retention of water on site, detention ponds and other conservation measures. In areas subject to severe water losses and soil erosion, variations in crop cover through the year create specific opportunities for erosion which need to be modified by retention of organic matter, minimizing fallow periods and avoiding excessive cultivation of steep slopes. Detailed analysis of the way in which human activities intervene and different scales in the basin sediment problem is absolutely essential if an effective management plan is to be established.

REFERENCES

Gu Hengyue & Douglas, I. (1989) Spatial and temporal dynamics of land degradation and fluvial erosion in the middle and upper Yangtze river Basin, China. In: Land Degradation and Rehabilitation, 1, 217-235.

Gu Hengyue & Ai Nanshan (1987) Sediment sources and fluvial processes of the Jialin river, China. In: International Geomorphology, Pari I (ed. by V. Gardiner), 791-800. Wiley, Chichester, UK.

Wohlke, W., Gu Hengyue & Ai Nanshan (1988) Soil erosion and fluvial processes in the basin of the Jialinjiang. Geojournain, 103-115.

Social and environmental factors and regional soil erosion analysis 21

Zhao Yu, Jin Zhengpin, Shi Peijun & Hao Yongchong (1989) Study of Soil Erosion in Inner Mongolia. China Science Publishing House.

Zhao Chunyong, Tian Baoscheng, Zhao Xiaolu & Yi Dawei (1988) Subdivision of Chongqing Municipality into Districts of Agriculture Geomorphology. Chongqing Publishing House.

Sediment Problems: Strategies for Monitoring, Prediction and Control (Proceedings of the Yokohama _ _ Symposium, July 1993). IAHS Publ. no. 217, 1993. 23

Characteristics and control of soil erosion in Hubei Province, China

YAO HUAXIA Institute of Geodesy and Geophysics, Chinese Academy of Sciences, 54 Xu Dong Road, Wuhan 430077, China

Abstract Surface soil erosion in Hubei Province, China is statistically calculated and characteristics such as climate-, plant-, and soil-conditions, erosion severity and distribution, and damages and causes are analyzed. An annual soil thickness of 1.66 cm, an amount of 319 billion kg, is eroded from mountains, hills, and farmlands causing unfavorable effects on the environment and eco-social systems. Also, measurements and planning for erosion control are discussed. It may be summarized that monitoring research and erosion control needs to be expanded.

INTRODUCTION

Hubei Province is located in central China, with an area of 185 900 km2 and a population of 51.4 million (Fig. 1). The climate is typical of the sub-tropical monsoonal zone. Annual rainfall is 1166 mm and annual runoff is 528 mm. Nearly 74 percent of the land surface consists of loamy soils, and only 20.3 percent is covered by forests. Moreover, the landscape is mainly mountains and hills, 70.6 percent of the total area. Mountains higher than 500 m above sea level comprise 56 percent of the area.

Several factors — steep hillslopes, poor vegetation cover, flood flows, torrential streamflow, and dense population — contribute to severe soil erosion, surface soil is eroded by various forces such as rainfall, wind, and human activities, and transported along river valleys onto the plains and lakes, causing substantial damages to the agricultural economy and environment. Therefore, the characteristics and controls of soil erosion are analyzed in this paper, based on hydrological data gathered from field monitoring networks, and on remote-sensing data.

SEVERITY OF EROSION

Erosion occurs in an area of 61 300 km2, or about one-third of the total land area. Annual sediment yield is 319 billion kilograms; of which 180 billion kilograms come from the highly eroded regions of 19.600 km2 in area. Compared with an ares of 50 120 km2 in the early 1950's it has increased by 11,180 km2, or 22.3 percent, in spite of erosion control practices on 27 842 km2. The soils are primarily eroded and transported by surface runoff or by runoff-related landslides and debris flows.

24 Yao Huaxia

Fig. 1 The location of Hubei Province in China.

Total erosion in the area may be divided into five categories according to erosion severity, as shown in Table 1. According to four criteria, for example, surface soil type, underlying rock type, erosion force, and resistance to erosion to surface material. Fluvial erosion of the land surface in Hubei Province may be spatially mapped as follows (see Fig. 2): (A) Soft, weathered rocks and yellow-colored loamy soils are exposed in the

northwest mountains. Because vegetation cover has been severely depleted or destroyed, and cultivation of steep mountain slopes, erosion is generally more than 3.0 x 106 kg/km2. It is the most severely eroded region of Hubei Province.

(B) Northeast and southeast mountains are made up of granitic rocks and yellow or red loam soils, with a weathering depth of 20 to 30 m. Caused by biological, physical, and chemical weathering, the soil and rock matrix contain large amounts of original minerals, but organic material is less than one percent. Therefore, the top layer is loose, coarse, and lacking clay particles. It is also a severely eroded region.

(C) In the northern hills the soil is predominately Quaternary clay. In the upper layer, clay-size particles (<0.1 mm in diameter) comprise more than 50 percent of the material, and organic material comprises less than one percent. Therefore,

Table 1 Severity division.

Category (xl(Fkg/km2- a)

Light Moderate Severe Extremely severe Most severe

Erosion index (km2)

0.5-2.5 2.5-5.0 5.0-8.0 8.0-13.5 >13.5

Area (%)

7 965 33 738 9 202 7 461 2 903

Area percentage (%)

13.0 55.1 15.0 12.2 4.7

Soil erosion in Hubei Province, China 25

Fig. 2 Spatial division of soil erosion.

the soils have low hydraulic conductivity and a high runoff coefficient, being easily eroded.

(D) The Three-Gorge valleys are underlain by a violet granite bedrock matrix, and yellow or violet loamy soils exposed at the surface. The region is noted for its severe erosion. It is located at the upper reaches of Gezhouba dam and the Three-Gorge reservoir. The granite bedrock is very permeable and nutrients are drained to depth. The topsoil has little resistance to erosion, and the steep valley is sparsely vegetated. On some sites as many as 8000 t of soil per km2 may be removed in one year.

(E) The southwest mountains are covered by loam and limestone bedrock, with a good vegetation cover. The limestone is readily eroded and topsoil is consequently very shallow. Where vegetation cannot be sustained, soil erosion occurs and bedrock is exposed. Furthermore, there are many eroded gullies, and gravity collapse (piping) is common in the region.

EROSION DAMAGE

Soil erosion destroys both land and water resources. It also changes in soil properties, nutrient content, and plant communities. Major damages are described in the following discussion.

Nutrient loss

On average, a thickness of 1.66 cm of topsoil is removed each year. Of this soil, there are 1.7 billion kg of nitrogen, 3.2 billion kg of phosphorus, and 40.9 billion kg of potassium. Overland flow transports fine soil and nutrients from upland areas, deposits coarse sands and gravels, and as a result farmlands degrade progressively from year to year. On the other hand, soils coming from mountainous areas may cover farmlands and completely inundate crops.

26 Yao Huaxia

River or lake deposition

Erosion has increased sediment concentrations in the flows of more than 20 rivers by 20 to 40 percent since 1973. The Yangtze River and its tributaries in Hubei Province receive 2.57 billion kg of sediment annually, creating a substantial threat to river-dike safety and social development. Efficiency and substantial threat to river-dike safety and social development. Efficiency and storage capacity of reservoirs and lakes are also challenged. One of the major hydro-power plants in Hubei Province, Danjiangkou reservoir, has a capacity of 16 billion m3. The annual sediment load of 0.12 billion m3

raised the bed level by 5.5 m in the period 1967-1976. As for Honghu Lake, the largest reservoir in Hubei Province, its area decreased rapidly from 687 km2 in the I960's to the present 353 km2, and the flood regulation capacity was reduced by 4 billion m3.

Environmental deterioration

All types of vegetation cover — trees, grasses, and agricultural crops do not grow well in areas of severe erosion. Ecosystems in these regions become fragile and mountain torrents, debris flows, and landslides present severe natural hazards.

CAUSES OF EROSION

The main erosion process in the region, overland flow, is controlled by natural and artificial factors. Key factors are discussed in the following paragraphs.

Storm flow

In the mountains and hilly topography, annual rainfall usually exceeds 1000 mm, and 50 to 70 percent of the rainfall occurs in the four-month period May to August when the rainfall pattern of sustained rain is accompanied by wind. These heavy rainstorms are a primary force in erosion.

Steep topography

Although eastern, northern, and western territories are high mountain area, the south-central area is a vast plain. Between the steep mountains and the vast plain there are distributed many mounds and grasslands. Steep slopes in these areas are easily eroded.

Sparse vegetation

There are 20 677 km2 of barren hills in the region that have not been forested yet, and the average cover-ration is as low as 20.3 percent. Furthermore, a large population

Soil erosion in Hubei Province, China 27

makes the forested stand per person to be only 800 m2, a value of 8.7 percent. On the other hand, these forests are not uniformly distributed, and northern regions have more than half the forest lands. This poor vegetation cover condition promotes soil erosion.

Poor land use

Many farmers still use a primitive pattern of cultivation. They reclaim land and grow crops on steep terrains. They cultivate soils perpendicular to the hillslope contours in place of along contours. Hydraulic, transportational, industrial, mining, and urban development also have caused severe soil erosion.

PLAN FOR EROSION CONTROL

Water and soil conservation have been studied and tried for 40 years, and much experience and progress has been made. Measurements include: (a) creating or repairing 1.16 million pools; (b) construction of 6200 reservoirs on 1100 rivers; (c) afforestation of trees on 20 861 km2 of land; (d) terracing steep fields on 3567 km2 into flat fields; (e) putting 7667 km2 of farmlands into order; (f) reforming 5687 km2 of low-yield farms; and (g) conservation on small watersheds totalling 2284 km2. After all these activities, 56 000 km2 of land have been controlled. Twenty-six billion m3 of runoff can be stored and managed, and a 27 842 km2 area has been primarily controlled.

Now local governments and scientists are starting on a plan for soil and water conservation (1991-2000), which includes three aspects: (a) to afforest with protective trees and timber trees on barren hills in an area of

14 674 km2; (b) to bring slope cultivation under control, which is the main source of erosion.

About 1667 km2 of steep farmland will be terraced into flat fields; and (c) to construct small flood-control dams in the region of severe erosion. After the

above plan is completed, a new area of 23 000 km2 will be protected. Finally, it might be concluded that both protection and control should be given

much attention and intensive measurements be made for severely eroded fields. Ground cover might be used as the alternative for regions of low erosion.

Sediment Proble?ns: Strategies for Monitoring, Prediction and Control (Proceedings of the Yokohama Symposium, July 1993). IAHS Publ. no. 217, 1993. 29

Seasonal variation of sediment yield on a gentle slope in semi-arid region, Tanzania

S. ONODERA STA Fellowship, Forestry and Forest Products Research Institute, Tsukuba Norin Danchi, Ibaraki, 305, Japan

J. WAKUI JICA, Japan

H. MORISHITA Graduate School of Technology, Tokyo Institute of Technology, Tokyo, Japan

E. MATSUMOTO Institute of Geoscience, the University of Tsukuba, Tsukuba, Japan

Abstract This research was conducted on a gentle slope in inland area of Tanzania for two years, in order to clarify the seasonal variation of sediment yield on inter-rill area in semi-arid zone. The sediment yield and runoff indicated the clear relationship, except for the value at the beginning of the rainy season. Sediment yield varied seasonally with the amount of erodible soil and vegetation coverage. The comparison of sediment yield at bare land with grassland which grows only during the rainy season suggested the following mechanisms. The decrease in the sediment concentration at the beginning of the rainy season was controlled by the existence of a lot of erodible soil on the ground, and the decrease in it in the late rainy season was controlled by the developing vegetation. The annual sediment yield was estimated from the relationships of sediment yield factors. Soil erosion rate at the complete bare land was 3.1 mm/yr, on the land covered with an acacia tree was 1.0 mm/yr and on the grassland was 1.2 mm/yr.

INTRODUCTION

In recent years, it has been confirmed that sediment yield was most remarkable in semi-arid, seasonal mediterranean, or tropical monsoon conditions (Walling & Webb, 1983). Especially, in semi-arid regions, the soil erosion caused by overland flow is an effective geomorphic agent. The predominant denudation processes in semi-arid zones are rainsplash, sheet erosion and "rain-flow transportation" proposed by Moss et al. (1979). Rapp et al. (1972), based on air photo interpretation and field checking on Dodoma in Tanzania, made clear that more sediment yield was caused by rainsplash and sheet erosion than by rill erosion. The "rain-flow transportation" is the process that occurs when raindrops impact on the soil surface covered by thin layers of water flowing over it. Moss et al. (1979) indicated that the process is especially effective on gentle slopes.

Evans (1980) reviewed about sheet erosion on the inter-rill that the factors controlling the temporal and special variation of erosion are the rainfall, vegetation,

30 S. Onodera et al.

soils, and slopes. However, the seasonal variation of sediment yield and its control factors on a gentle slope have not yet been made clear.

The objective of the present research is to confirm the seasonal variation in the sediment yield on the inter-rill areas, to make clear the factors of sediment yield on bare land and land covered with vegetation, and to estimate and forecast the annual soil erosion rates in this area.

EXPERIMENTAL SITE AND METHODS

The experimental site is located on a slope of Makutapora basin, 25 km north of Dodoma, capital of Tanzania. This slope is gentle, about 4 degrees, and its elevation is 1100 m above the sea level (Fig. 1). The basement of this area mainly consists of granitic rocks and a small block of metasediments of Precambrian. The basement is covered with a thin soil layer of about 0.5 m thick. The soil layer consists of reddish silt with granule and the saturated hydraulic conductivity is 10"3 cm/s. The specific gravity of the surface soil around this slope is 2.28 to 2.35 (g/cm3) and the porosity is 31.9 to 38.5%. The climate in this area is a tropical savanna. Most of the annual rainfall, about 550 mm, falls during the rainy season, from December to April.

The measurements of the runoff and sediment yield were conducted at four field plots installed on an inter-rill area. Four field plots are 3.2 m x 3.2 m (10 m2) in size, and each land surface is managed as follows. The plot PI is kept bare throughout the year, the plot P2 is covered with an acacia tree of 2.5 m high, and the remaining two, P3 and P4, are left grasses to grow during the rainy season, where P3 was measured in 1990, and P4 in 1991. The measurements were carried out from December to February in 1990, and from December to February in 1991.

The erodible soil amount was estimated from the weight of soil that was able to be collected by sweeping out by a soft brush two small plots of 400 cm2 (20 x 20 cm). A number of areas were randomly selected on the inter-rill area around this site. The erodible soil amount were measured each 15 days from December to February in 1991. The erodible soil depth (mm) is calculated by dividing the soil amount by a specific gravity and volumetric soil content (%). Vegetation coverage (%) was surveyed from the photo analysis each 2 to 10 days from December to February in 1991.

RESULTS OF FIELD OBSERVATION

Figure 2 shows the variations of the runoff and sediment yield at the field plot PI from December to February in 1990. From this figure, the characteristics of the runoff and sediment yield on the interrill area are as follows: (1) the runoff relates to the rainfall; (2) in the beginning of the rainy season, sediment yield is high; (3) during the rainy season, sediment yield decreases gradually; and (4) in a heavy rain conditions, the sediment yield is high.

Factors of seasonal variation

In order to confirm the factors that control the variation of sediment yield, the

Seasonal variation of sediment yield in semi-arid Tanzania 31

S. Onodera et al.

Dec. Feb.

20 •

40 •

u. 60 • z < a:

F o \ — l i

O s 1 LE

cri

n _ j LJ

>-h-y Ld

n hi (/)

60 eoo

600

400

200

0 4000

3000

2000

tooo

0

PI

PI

ULI

Apr. 1990

Ï 11 I11' l'I" l'fïlf 1 ' M 1

11 L „i U i .

Fig. 2 Variations in the runoff and sediment yield on the bare land, field plot PI, from December, 1989 to April, 1990.

following statistic analysis was carried out. In the present paper, the authors considered the vegetation coverage, existence of the erodible soil and runoff amount as the control factors of variation of soil erosion on the area which have a similar morphology.

The sediment yield is related to the runoff by the following correlation:

S = 12.5(2 - 99.9 r = 0.88 at PI (1)

S = 17 AQ - 100.9 r = 0.88 at P3 (2)

where S is a sediment yield (g/10m2), and Q is a runoff amount (l/10m2). Other studies (Wischmeier, 1975 and Elwell, 1981) had indicated that sediment

yield decrease exponentially with the increasing of a vegetation coverage. The present paper considers not only a vegetation coverage but also the erodible soil amount as one of the seasonal variation factors except for the runoff amount. Figure 3 shows the relationship between time since the first rain event in the rainy season and the sediment concentration at the bare area (PI) in a whole year and the bare area (P3) only along the dry season, where grasses gradually grew during the rainy season. This figure indicates the following variation: (1) The sediment concentration decreases rapidly at the beginning of the rainy season. (2) The sediment concentration decreases slowly in the late rainy season. (3) The sediment concentration at PI decreases more rapidly than at P3 at the

beginning of rainy season.

Seasonal variation of sediment yield in semi-arid Tanzania 33

30 r

Dec. Feb. 1990

Fig. 3 Variations in the sediment concentration at PI and P3 from December, 1989 to February, 1990.

(4) The sediment concentration at PI decreases more slowly than at P3 in the late rainy season.

(5) The sediment concentration C (g/1) is related to time (day) counted since the first rain event in the rainy season by the following correlation curve:

C = 32.2T - 0.32 r = -0.48 at PI

C = 154AT - 0.73 r = -0.66 at P3

(3)

(4)

where T is the number of days since the first rain event in the rainy season. The above (3) term suggests that the decreasing of sediment concentration in the

beginning of rainy season is controlled by the erodible soil amount on the ground, and the (4) suggests that the decreasing in the late rainy season is controlled by the developing vegetation. The authors consider that the dispersion of the correlation is caused by the variation of energy of rain drops that means rainfall intensity and the variation of antecedent soil moisture condition.

Figure 4 shows changes of the erodible soil amount, vegetation coverage and soil erosion amount at PI and P4 from December, 1990 to February, 1991. The relationship of correlation between the sediment concentration and time in 1991 is shown by the following equation at PI.

25. i r - 0.41 r = -0.58 at PI in 1991 (5)

The relation is approximately similar to equation (3) for 1990. Figure 4 indicates that there is more erodible soil in the beginning of rainy season (1.2 mm) than in the late rainy season (0.8 mm). This decreasing tendency agrees with the results of soil erosion during the rainy season. These results suggest that the decreasing of sediment concentration in the beginning of rainy season is controlled by the existence amount of erodible soil.

Vegetation cover at the plot P4 covered with grasses is less than 20% up to January 7 in the beginning of rainy season, but it increases after the heavy rain of 70 mm on

34 S. Onodera et al.

30

25

20

O 15 \ en

" 10

ci 00 5

0

2

1.5

1= E 1

S 0.5

x PI

—PI op4

.... p4

X

x E.S.

° veg. o

o

o x o

o

o o

o

X

o

X

o

X

100

eo

60

40

Jan. Feb. 1991

> o o

•20 -5

Fig. 4 on the

Variations in the sediment concentration at P I and P4, the erodible soil amount bare land, and the vegetation coverage at P4.

January 7, and it reaches 80% at the beginning of February in 1991. This gradually increasing tendency is similar to the reverse of the decreasing tendency of sediment yield at P4.

From the above results, the authors infer that the factors controlling sediment yield on the inter-rill area in the similar morphologic condition are the overland flow velocity and its amount, the amount of erodible soil yielded by weathering and deposited by wind on the slope especially during the dry season, and the growth and development of vegetation.

Forecasting of sediment yield

For forecasting of soil erosion rate on the gentle slope in a semi-arid zone, the present paper led the following equation of relationship among the soil erosion (S), runoff ( 0 and time (7) since the first rain event of the rainy season:

S = 32.2QT - 0.32 at PI

S = 197.3QT - 0.71 at P2

S = 154AQT - 0.73 at P3

(6)

(V)

(8)

In the present research, the authors measured the sediment yield of the rainy season in

en

O _ i UJ

>-h -Z Ld

Q LiJ Ut

O UJ

< DC ZD O _ l <

Fig. 5 Relationship between the measured sediment yield and calculated sediment yield.

1990 only from December, 1989 to February, 1990, though measured the runoff during the rainy season in 1990 (December, 1989 to April, 1990). For estimation of soil erosion rate in a whole year, the equations (6) to (8) and the measured runoff were used. Figure 5 shows the relationship between sediment yield calculated by equation (6) and the sediment yield measured at PI from December, 1989 to February 1990. The calculated sediment yield amount approximately agrees with the measured one in this figure.

The soil erosion rate (mm/yr) is estimated from the following equation and the sediment yield amount (g/yr) that has been calculated by the above step.

SR = l0St/G/V/A (9)

where SR is the annual soil erosion rate (mm/yr), St is the annual soil erosion amount (g), G is the specific gravity (g/cm3), V is the volumetric solid content (%) and A is the area of plot (cm2). From these parameters, the annual soil erosion rates were estimated as follows. Soil erosion rate at the complete bare land was 3.1 mm/yr, on the area covered with an acacia tree was 1.0 mm/yr and on the grassland was 1.2 mm/yr. These values are higher than the value of 0.7 mm/yr that Rapp et al. (1972) estimated from the sedimentation in the reservoir in the same region or the value on the steep slope in humid region (0.1 to 1.0 mm/yr) that Saunders & Young (1983) quoted in reviewing several studies. The above results supposed that the soil erosion may grow more seriously than in 1970 in this area, because of human activities such as the deforestation and grazing (Morishita, 1991). Especially, the extension of bare area caused by human activities makes the increasing of soil erosion accelerate, because the rainsplash, sheet erosion and "rain-flow transportation" act more effectively on the bare land than on the land covered with vegetation, and overland flow generate much on the bare area (Onodera, 1991).

Seasonal variation of sediment yield in semi-arid Tanzania

4000 r

35

3500

3000

2500

2000

1500

(000

500

P1

-J L_

0 500 1000 1500 2000 250Q 3000 3500 4000

36 51. Onodera et al.

The present paper can show quantitatively the seasonal variation. The value of 17% of the annual soil erosion rate yielded at PI from December 5 to December 16, 1989, in the beginning of rainy season when 11% of the annual rainfall fell. While 50% of the annual soil erosion rate yielded from March to April, 1990 in the late rainy season when 55% of the annual rainfall fell.

CONCLUSIONS

The results of the present study are summarized as follows: (1) The runoff and sediment yield indicate the clear relationship. (2) Sediment concentration divided sediment yield by runoff decrease with time

counted since the first rain event in the rainy season. (3) In the beginning period of rainy season, the sediment concentration was very high,

then rapidly decreased. This tendency is similar to the decreasing tendency of the erodible soil amount.

(4) In the late rainy season, the sediment concentration gradually decreased at P3. This tendency is similar to the reverse of the increasing tendency of vegetation coverage.

(5) The above results infer that the decreasing of the sediment concentration in the beginning of rainy season is controlled by the existence of the erodible soil amount on the ground, and the decreasing in late rainy season is controlled by the developing vegetation.

(6) The annual sediment yield were estimated from these results and clarified the relationships between sediment yield and its control factors. Soil erosion rate at the complete bare land was 3.1 mm/yr, on the land covered with an acacia tree was 1.0 mm/yr and on the grassland was 1.2 mm/yr.

Acknowledgements This research is a part of the Japan-Tanzania Joint Research led by Prof. S. Shindo, Chiba University and was supported by the members of this project, Prof. Ikeda, Prof. Sato, Dr Kondo, Dr Miyauchi and Mr Kongola from Ministry of Water, Tanzania. Special thanks are given to Mr Nahozya and Mr. Katanga for helping us to conduct the field researches.

REFERENCES

Elwell, H. A. (1981) A soil loss estimation technique for southern Africa. In: Soil Conservation: Problems and Prospects, 281-292. Wiley.

Evans, R. (1980) Mechanics of water erosion and their spatial and temporal controls: an empirical viewpoint. In: Soil Erosion (ed. by Kirkby, M. J.), 109-128. Wiley.

Morishita, H. (1991) Forecasting of soil erosion. In: Study on the recharge mechanism and development in the inland area of Tanzania (ed. by Shindo, S.), Progress Report of Japan — Tanzania Joint Research, 35-37 and plate 4-6.

Moss, A. J., Walker, P. H. &Hutka, J. (1979) Raindrop-stimulated transportation in shallow water flows. Sediment. Geology. 22, 165-184.

Onodera, S. (1991) Measurement of infiltration rates for estimation of overland flow amount on the semi-arid region in Tanzania. International Hydrology and Water Resources Symposium, 3, 815-816.

Seasonal variation of sediment yield in semi-arid Tanzania 37

Rapp, A., Murray-rust, D. H., Christlasson, C. & Berry, L. (1972) Soil erosion and sediment transportation in four catchments near Dodoma, Tanzania. Geografiska Annaler 54, A, 255-318.

Saunders, I. & Young, A. (1983) Rates of surface processes of slopes, slope retreat and denudation. Earth Surf. Processes andLandforms 8, 473-501.

Walling, D. E. & Webb, B. W. (1983) Patterns of sediment yield. In: Background to Palaeohydrology (ed. by Gregory, K. J.), 69-100. Wiley.

Wischmeier, W. H. (1975) Estimating the soil loss equations cover and management factor for undisturbed areas. In: Present and Prospective Technology for Predicting Sediment Yields and Sources. Agricultural Research Service Publication ARS-s-40, 118-124.

Sediment Problems: Strategies for Monitoring, Prediction and Control (Proceedings of the Yokohama Symposium, July 1993). IAHS Publ. no. 217, 1993. 39

Correlation analysis of rainfall: classification method of rainfall in view of sediment yield and transport

TORU SHIMADA Kokusai-Kogyo Corp, 3-6-1 Asahigaoka, Hino-city, Tokyo, 191, Japan

KUNIAKI MIYAMOTO Sabo Technical Center, 3-4 Ichigaya-sadohara-cho, Sinjuku-ku, Tokyo, 162, Japan

Abstract It is important to understand the characteristics of rainfall in order to make sediment disaster prevention plans. We use the method of correlation analysis to know the temporal changes in the spatial distribution of precipitation, and find that the method is efficient to know the characteristics of rainfall.

INTRODUCTION

The processes of erosion and deposition in rivers and streams are determined by the hysteretic relationship between discharge and yield of sediments depending on rainfall and its runoff, and these processes influence sediment disaster patterns. Therefore, when making sediment disaster prevention plans, it is important to understand the characteristics of rainfall.

So far, when making the plans, the plan precipitation has been defined as the precipitation causing sediment yield, because the sediment, which is passing at a point in river stream and domains sediment disaster, is yielded at anywhere in upstream watershed and transports by rainfall. Recent improvements in numerical calculating method of river beds variation make it possible to take into account discharge processes of sediments. Therefore, it becomes possible now to regard the plan precipitation as the precipitation which cause not only yielding but transporting sediment. The sediment transport at any point in a river depends on flood discharge at the point determined by temporal changes in the spatial distribution of precipitation in the watershed. Therefore, rainfall properties must be treated as temporal changes in spatial distribution of precipitation.

This paper will present a method of approaching a precipitation time series from the perspectives mentioned above.

COMPUTATION OF REPRESENTATIVE SPATIAL AND TIME SCALES FOR PRECTPITATION

Individual time series precipitation data having specific spatial and time scales will be treated as 'noise signal' that fluctuates in terms of both time and space. When precipitation is treated in this manner, its spatial and time scales are characterized by the reciprocity in the correlation between temporally and spatially lagged data. In correlation analysis, which usually deals with time series data, the correlation between

40 Torn Shimada & Kuniaki Miyamoto

data can be evaluated using the cross-correlation coefficient as shown in equation (1):

(1) Rxy(T) = C^WJCyioyCyip)

lim r r/2 X(t) • y (t + r) at (2)

In equation (2), x(r) and y(r) are time series data, while r is the temporal lag. When the time series x(r) and y(r) are identical, the correlation coefficient is called auto-correlation coefficient, as shown in equation (3):

RXX(T) = CJfilCJp) (3)

Auto-correlation coefficient under random movement decreases as r increases, which usually can be expressed as one of the equations (4), (5):

Ra(j) = exp{-| r \lra} (4)

Rb(r) = exp{-(7/rZ?)} (5)

Here, ra and rb express representative time scales of changes in the time series, with each satisfying its respective equation:

ra = ÎRa(r] )dr (6)

u 2

TO = — - ÎRb(T)dT (7)

What is referred to here as a representative time scale does not indicate the time between beginning and end of precipitation but rather the duration of rainfall of similar intensity. Figure 1 shows an example of RXX(T) derived from actual precipitation data. The change in RXX(T)IS approximated in equation (4). Then, a suitable measure for indexing a representative hyetograph time scale is shown in equation (6).

The autocorrelation coefficient or the time scale ra exhibits temporal characteristics of hyetograph as 'noise signal' at certain observation points but it does not show spatial and time scales of precipitation. In order to know the spatial scale of rainfall and associated temporal changes, it is necessary to compare precipitation time series from

o> ^ 0-5

0.0

10 20 30 4 0 x (hour)

Fig. 1 An example of auto-correlogram.

Correlation analysis of rainfall 41

1.0

0.5

0.0 S3aS*âèîSÈ

Ï J

*sfe^ J&

""^^^fe^iS -20 -10 10

r(hour)

20

1.0

0.5

0.0

f./i

Wfi

\ \

Ik • ^ "^""N ~s

as^àâs -20 -10 10

z (hour)

20

Fig. 2 An example of Cross-correlogram.

several observation points. Figure 2 shows the correlation coefficient calculated between precipitation time series obtained at the center and the other observation points which are selected at the nodes of north-south and east-west grid. For observation points along the north-south axis, the maximum correlation coefficient is at roughly r = 0, which means there is at most only a small time lag in the precipitation pattern. On the other hand, the correlation coefficient along the east-west axis peaks at the point r is not equal to zero, showing a time lag in the precipitation pattern. Furthermore, if we examine the relationship between time lag r and the correlation coefficient of the peak in east-west axis, we can see that the coefficient of cross-correlation becomes smaller as the time lag becomes larger. These two results shows that the rainfall moves in east-west direction.

Incidentally, the usual way to quantify such correlations of random signals is coherence and phase as shown in equations (8), (9). In equation (10), TW expresses phase as a function of time:

coh2(o>) = V")2

£ * » V«) (8)

6xy (w) = tan (9)

T U _ M«) (10)

42 Toru Shimada & Kuniaki Miyamoto

S (co) is the cross spectrum of X(T) and V(T) and is defined by equation (11). Additionally, Q^d) and K^w), as shown in equation (12), define the complex composition of cross spectrum ^ (w) :

OS

s^(w) = 2^ f {cxy (T)' exp(-w)}d' (11)

E„(W) = K^-iQJp) (12)

where Î is the imaginary unit. Coherence is to exhibit cross-correlation coefficient of each frequency component

of random signal, assuming that the process varying in time is the set of frequency component. If the precipitation time series caused by the same rainfall are recorded at some points, coherency for them shows a strong correlation. Therefore, the similarity of rainfalls can be determined by the value of Coherency. Phase 6 (a) refers to the phase difference of a couple of time series for each frequency, the phase difference for the prevailing frequency derived from the coherence of the precipitation time series between two points for identical rainfall can be regarded as the time needed for rainfall activity to move between two observation points.

Coherence and Phase, derived from the precipitation which is used in Fig. 2, are depicted in Figs 3 and 4. The solid lines in the figures represent the calculated value for observation points between the center point and the point approximately 50 km away from the center, while the broken lines represent the same at a distance of about 120 km.' This example shows that time- series of precipitation in short distance are more similar than one in long distance. This tendency is recognized obviously in the component of small frequency.

0.3 0.4 0.5 FrequencyOiour"1)

Fig. 3 An example of coherence.

0.5

Frequency(hour~')

Fig. 4 An example of r a>.

Correlation analysis of rainfall 43

In order to depict the temporal and spatial characteristics of precipitation which is used in Fig. 2, Coherence and Phase for coherent frequencies between the center and the other observation points, are plotted against the distance from center observation point in Figs 5 and 6. Figure 5 shows that Coherence becomes small as distance increase. And Figure 6 shows that relation between distance of observation point and Phase is approximately liner, then the gradient indicates velocity of the rainfall moving.

coherence

0.5

- 1 5 0 -100 - 5 0 50 100 150 Distance of observation points (km)

Fig. 5 The relation between coherence and distance.

5.0 T

4.0

3.0

2.0

1.0

-150 -100 -50 -1.0

e -2.0

- 3 . 0

- 4 . 0

- 5 . 0 1

(hour)

50 100 150

Distance of observation

points (km)

Fig. 6 The relation between T to and distance.

CONCLUSIONS

The conclusions of this research are as follows: (1) A suitable measure for indexing a representative hyetograph time scale is ra in

equation (6). (2) Cross-correlogram (Fig. 2) is efficient to know the characteristics of precipitation. (3) The degree of identity of a rainfall which may be observed at several observation

points and the movement of it can be evaluated using Coherence and Phase of the precipitation time series at the observation points. The methods, which are presented above and verified by actual data in this paper,

are very useful in order to classify the rainfalls. After classify the rainfalls into some categories which are characterized by range and intensity of the rainfalls, and life time and movement of them, the long term precipitation pattern will be able to make as the plan precipitation which may cause not only yielding but transporting of sediment.

2 Landslides and Pyroclastic Flows: Characteristics and Controls

Sediment Problems: Strategies for Monitoring, Prediction and Control (Proceedings of the Yokohama Symposium, July 1993). IAHS Publ. no. 217, 1993. 47

Underlying rock type controls of hydrological processes and shallow landslide occurrence

YUICHI ONDA Laboratory of Soil and Water Conservation, Department Forestry, Nagoya University, Nagoya 464-01, Japan

Abstract Hydrological observations were conducted in the areas where the shallow landslide density varied with rock types, and determined that the difference in landslide occurrence density can be explained by the different hydrological environment. In Obara region, central Japan, where few landslides were observed in the Granodiorite area and many in granite area, occurrence of landslides was found not to be controlled by regolith shear strength, instead by regolith zone thickness, which determines storm water storage capacity. In Niigata region, central Japan, few landslides were observed in Paleozoic sedimentary rock area and many in granite area. Hydrological observation showed that in Paleozoic basin, storm water can be discharged by Hortonian overland flow, whereas in Granite basin storm water can infiltrate in the regolith, thereby the regolith would saturate, resulting in landslides.

INTRODUCTION

Hydrological processes such as runoff generation pattern is recognized to vary with underlying rock types (Hewlett & Nutter, 1970; Freeze, 1972; Onda, 1992), and the frequency of shallow landslides also known to vary with underlying rock types (e.g. Hayashi, 1985). Nonetheless, few studies have provided the conclusive field evidence for the differences. Since hydrological environment seems to be the most important control on landslide occurrence (e.g. Selby, 1979), hydrological researches of two areas, where the landslide density varies with rock types, were performed.

GRANITE AND GRANODIORITE

Study area

In the investigated area there were 200-500 m high granite hills that were underlain by two types of Cretaceous granitic rocks in sharp contact with each other (Nakai, 1974): (a) a medium grained granodiorite covered in the northern part, and (b) a coarse grained granite in the southern part. The granite has higher resistance to chemical weathering than the granodiorite (Iida et al., 1986). Landslides due to a heavy rainfall (210 mm/3 hours) occurred in 1972 in Obara area, Aichi, Japan, where the landslide occurrence density varied with rock types; few landslides were observed in the granodiorite area and many in granite area (Fig. 1).

48 Yuichi Onda

Fig. 1 Scars resulting from 12 July 1972 landslide (Photo date 19 July 1972) and location of the experimental basin.

Field observation has been performed since 1986 to explain this contrast, with two experimental basins (0.41 — 0.86 ha) being established in each area since 1986. Discharge was taken at the outlet of each basin by thin-plate V-notch weirs, located at C-l, C-2, and F-l basins, and a partial flume, located at F-2 basin. Several manual tensiometers, vertically buried with different depths, were placed at the crest of in F-l and C-l basins. The experimental basins are located within a distance of 3 km; this approximates identical climatic conditions for the experimental basins. A raingauge was placed at C-2 basin.

Results and discussion

A regolith's shear strength, which can be measured by a direct shear test, is believed

Rock type, hydrological processes and shallow landslides 49

to be the catalyst in slope failures, and was determined using a 12-cm diameter shear box and saturated samples. The sampling was only performed in the F-2 and C-l basins, because sampling was easier than the other basins, which were underlain by the same bedrock. Normal stresses were applied at 100, 300, 500, 900 kPa. Shear speed was 1 mm/min. The regolith was collected at the potential slope failure depth (Iida & Okunishi, 1983) of 1 m. The granodiorite and the granite regolith's saturated shear strength were respectively C = 37.2 kPa/0 ' = 31.8° and C = 26.5 kPa/0 ' = 48.2°. Since the granite's shear strength was greater than the granodiorite's, the difference in slope failure distribution cannot be attributed to the regolith's shear strength in saturated conditions.

Regolith shear strength normally decreases with respect to its water content (Selby, 1979), therefore hydrological observations must be performed to investigate regolith properties. Figure 2 shows hydrograph of the F-1 and C-l basins during the 1986 rainy season (16 June-16 July), which had a total rainfall of 462 mm. Compared with the C-l basin, the F-1 basin had lower peak runoff and relatively higher and more steady base flow. The runoff ratio during the rainy season was 12.6% for the F-1 basin, and greater than 39.8% for the C-l basin. The runoff ratio for the F-2 basin was greater than 18.8% and in the C-2 basin was 39.7%, indicating that runoff characteristics depend on the underlying rock types.

The water budget equation (Freeze & Cherry, 1979) shows the remainder of the precipitation must be stored in the basin, thus the regolith acts as reservoir during the rainy season and afterwards supplies the base flow. It has been reported that the regolith structure may control the runoff characteristics (Hewlett & Nutter, 1970),

l/sAm;

200

<rw - v f i 1 m'

F-1 Granodiorite

' J ' T ' l : -

n/h

10 < u_

H 20

30 ce

' 16 17 18 192021 222324-252627282930 1 2 3 I 5 6 7 8 9 10 1112 13 U I 5 16

JUNE 1986 JULY

Fig. 2 Hydrograph of the F-1 and C-l basins during 1986 rainy season.

50 Yuichi Onda

therefore regolith zone thickness was investigated using a cone penetrometer. The regolith zone is defined as the depth that Nl0 < 50 (Iida & Okunishi, 1983). The N10

value is defined as a number of blows for a 10-cm penetration b'y a cone penetrometer with a cone diameter of 3 cm, a weight of 5 kg, and a falling height of 50 cm.

Figure 3 shows the subsurface structure and suction values in the F-l and C-l basins. The regolith thickness at the crest of the F-l basin (Fig. 3a) was 5.9 m, however, the average thickness was 4 m, therefore 800 mm of storm water could be stored in the macro pores (Onda, 1989). In contrast, the C-l basin's regolith zone thickness in the granite area was only 1 m, having a calculated water storage capacity of 210 mm (Onda, 1989). The regolith thickness also varied with underlying rock type in the C-2 and F-2 basins.

Figures 3b and 3c show the suction values during the 1986 rainy season obtained by manual tensiometers. After a rainfall the F-l basin's suction value (Fig. 3b) in the regolith's shallow region abruptly decreased, however with increased depth the suction value only gradually decreased, i.e., the suction value at 300 cm decreased in a specific interval, followed by smaller values at 400 cm, 500 cm, and 590 cm. This implies a gradual lowering of the wetting front, whereas in the C-l basin (Fig. 3c), the suction value decreased to 20 cm H20 after each rainfall, indicating that the entire regolith zone is close to being saturated. However, after the rainfall, the suction value recovered, then decreased over time, demonstrating that the subsurface water movement and the regolith water content are primarily controlled by regolith thickness.

These results indicate that landslides were not controlled by regolith shear strength, instead by regolith zone thickness, which determines storm water storage capacity.

PALEOZOIC SEDIMENTARY ROCKS AND GRANITE

Study area

Landslides due to a heavy rainfall (350 mm/day) occurred in 1968 in Kaetsu region, Niigata, Japan; few landslides were observed in Paleozoic sedimentary rock area and many landslides and debris flow were found in granite area (Fig. 4). The mountains have approximately 900 m elevation and the granite covers the northern part and the Paleozoic sedimentary rocks (shale, sandstone etc.) cover the south. The elevation between the Paleozoic and the granite is about the same, caused by the similar mode of uplift (Watanabe & Une, 1985). Therefore the sediment budget in both regions is considered to be about the same.

A large (1.95-2.04 km2) and a small (0.03-0.04 km2) experimental basins have been established in each area since 1991. These basins are located within the distance of 3 km, and the Paleozoic and the granite basins are named PL and PS, and GL GS, respectively. Discharge and electrical conductivity (EC) was measured automatically in these basins.

Results and discussion

The observation showed that on the large Paleozoic basin, flows generated mainly from large springs, whereas flows generated as seepage from small seeps but a great deal

Rock type, hydrological processes and shallow landslides

fa)

C - 1 Granite Tensiometer y

F - 1 Granodiorite Tens iometV*

(b)

M R e g o l i t h z o n e

F - 1 Granodiorite

Ô300 X

(cm

2: a 200 — i—

o = 100 en

y " ' T ' < " W" »\

it i \ t

t t-

- \ /

\ .

r | • I'iniu

-' * 50cm

100cm 200cm

X

* • ' ' - * * ' / fy •' '

T 5 10 u.

2: 20 5

a:

E - 200

O

-- - • • " " ' *

'•MJt'"

"W i

300cm • -iOOcm

500cm 590cm

. '•fc'V1! * • ? * • " • " • * • • • " " • • . M0

i ^ ^ w ' ^ - ^ 2 » s" ' - . ,

• 3°° ^ ^ N r - t — , " " ^ " ^ ' - - •%. . • ' . ._ ._ .< ^-^Vv-yf

100 -

u 16 21 JUNE 1986 JULY

JUNE 1986 JULY

Fig. 3 Subsurface structure (a) and suction values in the F-l basin in the granodiorite region (b) and the C-l basin in the granite region (c) during rainy season in 1986. No observations were made from 22 to 26 June.

52 Yuichi Onda

Fig. 4 Location of study site and landslides resulting from 28 August 1968.

of them can be found in the granite region. The thickness of the regolith, obtained by the soundings, and infiltration rate, measured by tube infiltrometer (tube diameter of 8 cm) are listed in Table 1. This shows the regolith thickness was about the same but the smaller infiltration capacity was measured in the Paleozoic basin.

Figure 5 provides hydrograph for PL and GL basins during 1-14 August 1991. The PL basin's hydrograph (Fig. 5a) shows that runoff did not respond to precipitation in smaller intensity of rainfall (7 August), but responded quickly with precipitation > 30 mm/h (8-9 August), with the peak dissipating rapidly. The value of electric conductivity was generally constant during rainless periods but decreased abruptly with rainfall. The GL basin's hydrograph (Fig. 5b), on the contrary, responded quickly to each rainfall, having high discharge peak with delayed recession. The EC response to precipitation varied considerably from event to event.

Table 1. Regolith thickness, and infiltration capacity in the small watersheds.

Regolith Sample Infiltration capacity (mm/h): thickness (cm) number Average Minimum Maximum

Paleozoic (PS) 60 5 310 96 618

Granite (GS) 60 4 833 425 1254

Rock type, hydrological processes and shallow landslides

(#S/cm)(mm/h) MOr 0

10 I I 12 13 U 30 Uj

August 1991 Fig. 5 Hydrograpb. of the PL and GL basins in August 1991. No EC data were obtained in PL basin from 8 August.

To quantify the contrasting EC response to precipitation, the plot of 1-hour rainfall against deviation of EC from average value is made (Fig. 6). The plot of the Paleozoic basin (PL) shows EC-value gradually decreases with rainfall intensity, whereas little correlation can be found between rainfall intensity and EC value in the granite basin (GL). This illustrates that marked dilution effect was observed in PL basin, but was not found in GL basin. The small infiltration rate (Table 1) and the dilution effect during rainfall peak (Fig. 6) suggest that Hortonian overland flow may occur during the rainfall peak in the Paleozoic basin.

In the light of above results, Hortonian overland flow would contribute much of the peak runoff at its peak, occurring at the time of the heavy rainfall ( > 30 mm/hour). In contrast to this, in the granite basins, flows were judged as generated as translatory flow (Pearce et al., 1986) in terms of seepage outflow. The average regolith thickness in both area was the same (Table 1), suggesting that the difference in landslide

54 Yuichi Onda

E

a

+10

O)

v -10 UJ

c .o to S

Q

20

o

o D

s

cP

- D O

0

D e

©

e

D

0

• Granh • Paleo.

*°

D

i I

10 20 30

1-hour rainfall (mm)

40

Fig. 6 The relationship between 1-hour rainfall against the deviation of EC from average value in PL and GL basins.

occurrence was explained as the difference in the runoff mechanisms; in the Paleozoic area, storm water can be discharged by Hortonian overland flow, whereas in the granite area storm water can infiltrate in the regolith, thereby the regolith would saturate, resulting in landslides.

CONCLUSION

The above two cases suggested that the difference in landslide occurrence density can be explained by the different hydrological environment, which is controlled by underlying lithologies.

REFERENCES

Freeze, R. A. (1972) Role of subsurface flow in generating surface runoff 2. Upstream source areas. Water Resour. Res. 8(10), 1272-1283.

Freeze, R. A. & Cherry, J. A. (1979) Groundwater, 193-236. Prentice-Hall, Englewood Cliffs. Hayashi, S. (1985) Some relationships between the landslide-area ratio and hydrologie data. / . Japan. For. Soc.

67(6), 209-217. Hewlett, J. D. & Nutter, W. L. (1970) The varying source area of streamflow from upland basins. In: Proceedings

of the Symposium on Interdisciplinary Aspects of Watershed Management, 65-83. American Society of Civil Engineers, New York.

Iida, T. & Okunishi, K. (1983) Development of hillslopes due to landslides. Z. Geomorphol., Suppl, 46, 67-77. Iida, T., Yoshioka, R., Matsukura, Y. & Hatta, T. (1986) Development of the weathered zone on granitic rocks due

to dissolution. Trans. Japan. Geomorphol. Union 7(2), 79-89. (in Japanese with English abstract) Nakai, Y. (1974) Compositional variations of the Inagawa granitic rocks in the Asuke area, Aichi prefecture, Central

Japan. / . Japan. Assoc. Min. Petr. Econ. Geol. 69(3), 215-224. Onda, Y. (1989) Influence of water storage capacity in regolith zone on runoff characteristics and slope failure on

granitic hills in Aichi, Japan. Trans. Japan. Geomorphol. Union 10(1), 13-26. (in Japanese with English abstract)

Rock type, hydrological processes and shallow landslides 55

Onda, Y. (1992) Influence of water storage capacity in the regolith zone on hydrological characteristics, slope processes, and slope form, Z. Geomorphol 36(2), 165-178.

Pearce, A. J., Stewart, M. K. & Sklash, M. G. (1986) Storm runoff generation in humid headwater catchments 1. Where does the water come from? Water Resour. Res. 22(8), 1263-1272.

Selby, M. J. (1979) Hillslope Materials and Processes, 117-151. Oxford Univ. Press, Oxford. Watanabe, M. & Une, H. (1985) Active faulting and mountain building in the Eastern marginal are of Niigata Plain,

Central Japan. Geogr. Rev. Japan 58(8), 536-547.

Sediment Problems: Strategies for Monitoring, Prediction and Control (Proceedings of the Yokohama Symposium, July 1993). IAHS Publ. no. 217, 1993. 57

Characteristics of pyroclastic flows and debris flows accompanying the Mt Unzen-Fugendake eruption

HEROSHI EKEYA Sediment Control Department, Ministry of Construction, 2-1-3, Kasumigaseki, Chiyoda-ku 100, Japan

YOSHEHARU ISHIKAWA Public Works Research Institute, Ministry of Construction, 1, Asahi, Tsukuba-shi, Ibaraki-ken 305, Japan

Abstract Mt Unzen-Fugendake erupted on November 17, 1990. An investigation of the pyroclastic flows and debris flows occurring in the Mizunashi River during the period from May 1991 to August 1992 indicates the pyroclastic flows and debris flows had the following characteristics. The larger the amount of material flowing in a pyroclastic flow material ejected from the volcano, the further downstream the front edge of the pyroclastic flow will reach. The main body materials of a pyroclastic flow moves downward, to some degree, along a valley configuration, but the hot ash cloud moves straight down, without being much influenced by the topography. The temperatures of the hot ash clouds accompanying the pyroclastic flows were over 450°C and their wind velocity was more than 50 m/s. Debris flows are likely to occur even when rainfall is low. The larger the total volume of water runoff, the larger is the sediment concentration of the debris flow.

INTRODUCTION

Mt Unzen-Fugen (1359 m) (Fig. 1), located near the center of the Shimabara Peninsula, awoke in 1990 from its 198 years of inactivity and began to erupt. Following this eruption, debris flows and pyroclastic flows have been occurring continuously from May 1991 until the present time (August 1992) at the Mizunashi River, which lies at the eastern foot of Mt Unzen-Fugen. These debris flows and pyroclastic flows have caused much damage to people and buildings. This study describes characteristics of the pyroclastic flows and the debris flows produced by the Mt Unzen-Fugen eruptions.

COMMENCEMENT OF ERUPTION AND OCCURRENCE OF PYROCLASTIC FLOWS

Starting around July 1990, the frequency of earthquakes and tremors began to increase in the Shimabara Peninsula and in the area to the west. The eruptions started on November 17, 1990, at the Jigokuato crater and Kujukushima crater. After that initial activity, eruptions abated, then started again on February 12, 1991, at Byobuiwa crater,

58 Hiroshi Ikeya & Yoshiharu Ishikawa

Fig. 1 Map of the peripheral area of Mt Unzen Fugen (as of August 27, 1992).

near the peak of Mt Fugen, and a large amount of volcanic ash was erupted. Accumulated volcanic ash reached a maximum thickness of about 1.5 m near the peak of the mountain. A pyroclastic flow (the first dome) was extruded out of the Jigokuato crater on May 20. This pyroclastic flow grew in height every day, and on the 24th some of the lava began to flow down to the source of the Mizunashi River. This was the beginning of the pyroclastic flows, and since that time very small pyroclastic flows have been occurring frequently.

PYROCLASTIC FLOWS OF JUNE 3, JUNE 8, AND OCTOBER 15

A pyroclastic flow larger than all the earlier flows occurred at about 3:50 a.m. on June 3, and it swept over some parts of Kitakamikoba-machi and Minamikamikoba-machi, causing a disaster with 43 fatalities, 10 injuries, with 49 residential buildings burned down or collapsed completely (Figs 2 and 3, Table 1).

Later, a second dome emerged at the same location where the first dome had collapsed, and continued to grow. Most of this second dome collapsed, and a pyroclastic flow larger than all the previous flows occurred at around 7:50 p.m. on June 8. The main body of the new pyroclastic flow ran down along the Mizunashi River, filling up the river course with volcanic debris, and nearly reached Route 57,

Pyroclastic flows and debris flows of the Mt Unzen-Fugendake eruption 59

Fig. 2 Area influenced by the pyroclastic flows on June 3, 8 and September 15

which lies 5.5 km downstream from the crater (Fig. 2). Due to the breaking of a lava lump accompanying the growth of the third dome, very small pyroclastic flows started to occur on the northeastern slope (Oshiga Valley, which is the left branch of the

Fig. 3 Overview of route and deposition area of the pyroclastic flow of June 3 (Mt Unzen-Fugen in the front and Mt Mayu to the right).

60 Hiroshi Ikeya & Yoshiharu Ishikawa

Table 1 Casualties and damage to houses caused by the pyroclastic flows and debris flows.

Type

Pyro­clastic flow'

D ate of occurrence

1991

1992

May 2 9th

June 3rd

June 8th

Sept. 15th

Aug. 8th

River name

Mizunashi Riv.

Mizunashi Riv.

Mizunashi Riv.

Mizunashi Riv.

Mizunashi Riv.

Puroclastic Flow . Total

Debris flow

1991

1992

May 15th

June 30th

Aug. 8th

Aug. 12th

Aug. 15th

Mizunashi Riv.

Mizunashi Riv.

Yue Riv.

Miz.unashi Riv.

Mizunashi Riv.

Mizunashi Riv.

Debris Flow Total

Total

Casualties (persons)

Fatalities

0 43 0 0 0 43 0 0 0 0 0 0 0 43

The injured

l 10 0 0 0 11 0 0 1 0 0 0 1 12

Damage to building (ridges)

Residential buildings

0 49 72 53 6

180 0 64 34 35 84 56

273 453

Non­residential

0 130 135 165 12 442 1

87 17 7 4

11 127 569

Note: The number of fatalities includes those listed as missing (as of August 18, 1992).

Mizunashi River). At 9:21 p.m. on September 6, a small pyroclastic flow occurred in Oshiga Valley. At this time the front part of the hot ash cloud moved down along the Oshiga Valley and reached a point about 3.1 km from the crater. At 6:45 p.m., September 15, the largest pyroclastic flow up to that time occurred in Oshiga Valley. This pyroclastic flow struck against Taruki Platform, turned to the right, and flowed downhill. When it met the main stream of the Mizunashi River the main body (bottom part) changed its direction, veering left, and flowed down along the Mizunashi River main stream, reaching a point near Shiratani Bridge, located about 5.8 km downstream from the crater (Fig. 2).

CHARACTERISTICS OF THE PYROCLASTIC FLOWS

Using aerial photographs taken at different times, aerial cross section photogrammetry was conducted on the pyroclastic flow deposition area in the drainage basin of the Mizunashi River. Cross sections were set at intervals of 100 m along the river. Based on these cross sections, the amount of soil deposited by each pyroclastic position area in the drainage basin of the Mizunashi River. Cross sections were set at intervals of 100 m along the river. Based on these cross sections, the amount of soil deposited by each pyroclastic flow was determined for each photogrammetry interval (Fig. 4). Considering the amount of deposited soil produced by the very small pyroclastic flows that occurred in these three periods, it is estimated that the volumes of deposited sediment (volume of flow sediment) caused by the pyroclastic flows occurring on May 26, June 3, and June 8 should be approximately 300 000 m3, 2 500 000 m3, and 3 500 000 m3, respectively.

Pyroclastic flows and debris flows of the Mt Vnzen-Fugendake eruption 61

Crater (lava dome)

Second waterfall

(1.5km) .Volcanic debris of the pyroclas ; flow of May 26 •

Ash cloud of May 26 Volcanic debris of the pyroclastic flow of June 3 (4.5km)

Deposited «diment

June 4 - June 16

May 8 - June 4

M*y 16-M*y28

o,H;„„, „„ _ , . „ Ash cloud of the pyroclastic (5.5km) _ Severn control dam n o w o f w 3 y0\cm^ debris of the pyroclastic

flow of June 8 | Junction with ' ~ Y ' ; T 7 JC 5 ; l k m ) , •

jAkamaiw Valley / Ash cloud of the pyroclastic T — ./low of June 8 _____ Route 251

, Route 57 . The river mouth

5.5 6.0

—j.2S*-j 12' [-20* -j 1

0.6U I l.7kB ' 3 |___

( < a b

Row distance (reached) (km) r.s.. h

Original ri

5.5 7.0 7-5 7.75

s than 2° -j

The upper stream | (about 39OO0O0m'ij " (about 3100000m*)

T bed inclination I .5kra Thedownstream

1 (about 500000m') ~

JfiSj

Legend

Deposited sediment

Tr-TTI

~] j j June 4 -June 16

- [>__ May 8 - June 4

_fi May 16- May 28

lllll^.irr-rTTlTTTlH-rkTXTtT^ ,

Flow distance (reached) (km)

Fig. 4 Flow down and deposition of volcanic debris of pyroclastic flows at the Mizunashi River main channel.

Also, the amounts of soil deposited by the pyroclastic flows of August 26, September 6, and September 15 are estimated to be approximately 800 000 m3, 1 000 000 m3, and 4 000 000 m3, respectively.

Figure 5 shows the relationship between the amount of deposited flow down sediment produced by pyroclastic flows occurring in the Mizunashi River basin since

Mizunashi River mainstream

o -_ 400

300

! !

11 Fig. their

200

100-

Oshiga Valley

A A

Front edge of the deposited volcanic debris of the pyroclastic flow main bodies (indudingithin soi! volcanic debris pans) Front edge of ash cloud (volcanic ash) caused by pyroclaslic

1 2 3 4 Distance reached (km)

5 Amount of deposited volcanic debris produced by major pyroclastic flows and reaching distance.

62 Hiroshi Ikeya & Yoshiharu Ishikawa

the last third of June, and the travel distance of the pyroclastic flow. As seen in Fig. 5, a tendency can be observed, whereby the pyroclastic flows carrying more material (more material ejected by the volcano) are seen to reach a greater distance from the crater. Using this relationship, the distance reached by the front edge of a pyroclastic flow can be estimated to some extent from the amount of material in the pyroclastic flow (amount ejected from the crater).

DAMAGE CAUSED BY THE PYROCLASTIC FLOWS HOT ASH CLOUDS

The areas damaged by the hot ash clouds of the pyroclastic flows of June 3 and June 8 in the drainage basin of the Mizunashi River covered about 3.0 km2 and 4.1 km2

respectively. This was more than twice the area of the deposits of the main bodies which were about 1.0 km2 and 1.9 km2 for June 3 and June 8 respectively.

In the areas affected by the hot ash clouds, wooden buildings were burned down and vehicle tires melted. Judging from the temperatures at which wood catches fire and at which tires melt, it is estimated that the temperature of the hot ash cloud near the exit from the valley was over 450°C. Considering the way in which trees and concrete utility poles were felled and vehicles up-ended in the path of the hot ash cloud, it is estimated that wind pressures equivalent to a maximum instantaneous wind velocity of 40-50 m per second occurred near the valley exit during the pyroclastic flow of June 3. Since there was a forest of coniferous trees and broad-leaved trees spreading along the Mizunashi River, many trees were blown down by the wind pressure from the hot ash cloud. From Fig. 2, it can be said that although the main body sediments of a pyroclastic flow move downwards, following to some degree along the valley configuration, there is a tendency for the hot ash cloud to move straight down, without being influenced by the terrain of small valleys. This suggests that, even if the flow direction of the main body debris is changed by a training dike, the flow direction of the hot ash cloud does not necessarily follow it. This is an important point to be taken into a consideration when planning and designing facilities to counter pyroclastic flows.

DEBRIS FLOW AND INFLUENCE OF RAINFALL AND VOLCANIC ASHFALL

On May 15th, 1991, debris flows were caused by rainfall and as much as 70 000 m3

of debris flowed down from the upper stream of the Mizunashi River main stream. After the 15th, debris flows again occurred on the 19th, 20th, 21st, and 26th. However, no injuries or damage to buildings occurred, except for damage to two bridges, one hut and three utility poles.

On June 30, 1991, a large debris flow was triggered by a heavy rainfall (64 mm/hr) at the Mizunashi River. As of June 30, the river channel of the Mizunashi River was filled up to a point near route 57, with material produced by the pyroclastic flow of June 8. Therefore, after the two debris flows met, one from the Akamatsu Valley and the other from the left branch of the Mizunashi River, the confluent debris flow ran down onto the alluvial fan, concentrating on the relatively lower left bank of the Mizunashi River channel (Fig. 1). Based on aerial photography interpretation, the new area covered by the sediment from the debris flow was about 350 000 m2 and the volume of deposited soil was about 380 000 m3. At the same time, it is estimated that

Pyroclastic flows and debris flows of the Mt Unzen-Fugendake eruption 63

some of the debris flow changed into a hyper-concentrated flow and flowed downstream to the Mizunashi River main stream river channel, further down from Route 57, resulting in the deposition of as much as 80 000 m3 in the river channel near Route 251.

Figure 6 shows the relationship between changes in the amount of rainfall per day, maximum rainfall per hour measured at Mt Unzen meteorological station, and occurrence of debris flows after February 12, when a lot of ash fell. On May 15, the rainfall per day and the maximum rainfall per hour was the greatest since February 12. Therefore, it is considered that this heavy rain(fall) directly caused the debris flow in the Mizunashi River. There seems to be a tendency for debris flows to repeat, after the first one occurs after an ashfall, even with low rainfall. On May 26 and 29, a small pyroclastic flow occurred, and a large amount of pyroclastic flow material was accumulated, which created a condition whereby a debris flow could occur even with small amount of rainfall. However, after June 3 and 8, when relatively large pyroclastic flows occurred, there was no debris flow occurrence.

It is considered that this was because of the following: The temperature of the

Ï I

I t s.

Ï

Eruption from Byôbuiwa crater

.(muchajhtallKltt))

Occurrence of pyroclastic flow

Fig. 6 Changes in rainfall per day and the maximum rainfall per hour and occurrence of debris flow (rainfall observation station: Mt Unzen meteorological station).

64 Hiroshi Ikeya & Yoshiharu Ishikawa

deposits from the pyroclastic flow was still high, therefore, rain was evaporated and precluded surface water runoff, therefore, deposits accumulated at the upper and middle reaches of the Mizunashi River main stream did not become a debris flow source. Since the 64 mm maximum rainfall per hour rate, recorded on June 30, occurs only once in every four years, it can be said that the ashfall produced by the eruption became a major factor which supplied the conditions where a debris flow could easily occur. Rainfall per hour was about 10 mm, and continuous rainfall of 20-100 mm had occurred at the time when debris flows (including those similar to hyper-concentrated flows) occurred on May 15-26. These rainfalls were less than those usually experienced when a torrential debris flow occurs. Therefore, it is evident that there was an influence from the great amount of ashfall on the upper stream basin of the Mizunashi River.

At 1:28 a.m. on March 1, 1992, there was a debris flow from the Mizunashi River, caused by heavy rainfall from a weather front activity. The Oshiga Valley, along the left branch of the Mizunashi River, supplied a lot of material of the debris flow. Gullies have formed directly below the lava dome along the Oshiga Valley, because the pyroclastic flows had hardly occurred in the Oshiga Valley since January, 1992. Conversely, significant gullies have not formed, and there was little flooding in the main stream of the Mizunashi River. In the Akamatsu Valley, no significant erosion was noted in the pyroclastic deposits. The debris flow of March 1 did not cause fatalities or serious damage to houses, but the Shimabara Railway, Route 251, were blocked by flood water and deposits, which disrupted transportation. On March 15, 1992, there was a debris flow similar in size to the one occurring on March 1.

2

Occurrence of

debris flow

Yes

70

60

50

40

30

20

10

No

Legend (maximum one-hour rainfall of more than 5mm)

Feb. 12 to May 31,1991

June 1 to 29,1991

June 30 to Dec. 31,1991

Jan. 1 to Aug. 20,1992

O Jun. 30

f Aug. 8

• Aug. 12-13

•7 Mar. 1

120 0 20 40 60 80 100 Continuous rainfall preceding maximum one-hour rainfall (mm/hr' (with less than three-hour interruption)

Fig. 7 Relationship between occurrence of debris flows and debris and precipitation (measured by the Unzen meteorological station).

Pyroclastic flows and debris flows of the Mt Unzen-Fugendake eruption 65

On August 8, 12-13, and 15, 1992, relatively large debris flows were triggered by heavy rainfall along the Mizunashi River. At that time the river channel of the Mizunashi River was opened to the middle reaches with the erosion of the right river bank after March 1992. Therefore, in August the debris flows flowed down along the original channel of the Mizunashi River, resulting in deposits totaling about 470,000 m3

in the river channel. As a result, the river bed was raised up to the top of the bank protection levee, and the hyper-concentrated flow flooded onto the fan and caused great damage to buildings.

Figure 7 shows the maximum one-hour rainfall and the continuous rainfall (amount falling with interruptions of three hours or less) measured by the Unzen Meteorological Station since February 12, 1991 when a large quantity of ash fell during a volcanic eruption and the relationship between these rainfall data and the debris (and hyper-concentrated) flow occurring during the same period. Rainfall on June 30, 1991 and August 8, 1992 significantly exceeded that on other days, and the volume of debris flow transported and the area flooded were also larger. Maximum one-hour rainfall was approximately 10 mm and continuous rainfall about 20-100 mm at the time when debris flow (including that similar to hyper-concentrated flow) occurred in May of 1991. The rainfall that triggered debris flow in 1992 was within the above range.

The volume of sediment (V) which deposited in the downstream area (down Route 57) of the Mizunashi River by debris flows since May, 1991 was measured by surveys, and the total volume of water runoff (Q) of each rainfall was calculated as: the catchment area of the Mizunashi River (up Route 57) times continuous rainfall, times runoff rate (0.9). Figure 8 shows the relationship between the average sediment concentration [100V/(V + 0 ] and the total volume of water runoff for each debris flow. Figure 8 indicates more of the debris flows occurring on the Mizunashi River were low sediment concentration flows, containing only several percent sediment. Figure 8 also indicates that the higher volumes of runoff tend to have the higher concentrations of sediment.

100V

J u n e 3 0 , 1 9 9 1

o

192

1 I 1700

103m' ) -inoff Rate)

Fig. 8 Average sediment concentration of debris flows in the Mizunashi river.

T3

<a t/)

25

20

15

10

5 Aug .

~ O 15,1992

1

Aug.12-13,1992 O

1

Aug.8,19

O

500 1000

Total volume of Water Runoff Q (X ( C o n t i n u o u s Rainfall y Tatrhnipnt Arna x Rur

66 Hiroshi Ikeya & Yoshiharu Ishikawa

CONCLUSIONS

The larger the amount of sediment flowing down in a pyroclastic flow (amount ejected from the crater), the further downstream the front edge of the pyroclastic flow reaches. The main body sediments of pyroclastic flows move downwards, to some degree, along a valley configuration, but the hot ash cloud moves straight down without being much influenced by the topography. The temperature of the hot ash clouds of pyroclastic flows can be over 450°C, and their wind velocity can exceed 50 m/s. These features caused the loss of lives and the heavy damage to houses and forests in the Mizunashi River basin.

When volcanic ash from an eruption accumulates on a mountain slope, and it absorbs water from rainfall, its permeability is lowered, and the surface runoff rate increases. Since deposited volcanic ash itself becomes liquefied by absorbing water, and will flow down as a debris flow, debris flows or hyper-concentrated flows are likely to occur, even with low rainfall. When the temperature of the pyroclastic flow deposit is still high, rain is evaporated and there is no surface water runoff. Therefore, when deposit temperatures are high, debris flow does not occur with low rainfall. After the temperature of the pyroclastic flow deposits cool, debris flow is likely to occur, even with low rainfall. The debris flows which occurred at the Mizunashi River were low sediment concentration flows, containing several percent sediment. In addition, it was found that the greater volume of water runoff, the higher the concentration of sediment in the debris flow.

Acknowledgments We would like to express our deep appreciation to the Erosion Control Division, Erosion Control Department, Ministry of Construction, the Shimabara District Promotion Bureau, Nagasaki Prefecture, for extending their valuable time and assistance to us for our investigation during continuous pyroclastic flows and debris flows.

REFERENCES

Ikeya, H. & Ishikawa, Y. (1991a) Disaster caused by pyroclastic flows and debris flows occurring at Mt Unzen in 1991. J. Japan Society of Erosion Control Engineering 44(2), 46-56.

Ikeya, H. & Ishikawa, Y. (1991b) Disaster caused by pyroclastic flow and debris flow occurring at Mt Unzen in 1991 (the second report). J. Japan Society of Erosion Control Engineering 44(5), 35-46.

Ikeya, H. & Ishikawa, Y. (1992a) Disasters caused by debris flows at Unzen-Fugen Volcano in March, 1992, and Countermeasures. J. Japan Society of Erosion Control Engineering 45(1), 56-60.

Ikeya, H. & Ishikawa, Y. (1992b) Disasters caused by debris flows at Unzen-Fugen Volcano in August, 1992. J. Japan Society of Erosion Control Engineering 45(3).

Sediment Problems: Strategies for Monitoring, Prediction and Control (Proceedings of the Yokohama Symposium, July 1993). IAHS Publ. no. 217, 1993. 67

Model of pyroclastic flow and its numerical simulation

S. YAMASHITA Sumiko Consultants Co., Ltd., 16-9, Kabuki-Cho 2, Shinjuku-Ku, Tokyo 160, Japan

K. MIYAMOTO Sabo Technical Center, 3-4, Ichgaya-Sadohara-Cho, Shinjuku-Ku, Tokyo 162, Japan

Abstract A gravity flow of granular bodies is considered as a model of main body of pyroclastic flow. Characteristics of the granular flow can be described by Kanatani's constitutive equations (Kanatani, 1984), which are obtained in consideration of the energy loss caused by the only inter-particle friction by the collision between the particles. Using this model, the 1991 pyroclastic flow at Mt Unzendake were reproduced by the numerical simulation. The result of calculations roughly agree with the actual phenomena. This study also shows inter-particle friction decreases as a result of the formation of a pressure gradient by gas ascent velocity.

INTRODUCTION

Pyroclastic flow is one of many types of damage-causing volcanic activities such as lava flow, volcanic mudflow and volcanic ash fall. Being a very rapid phenomenon, it has the potential to cause great damage. Then, in the past, several pyroclastic flows have occurred, and caused disaster around the volcanos.

Information about pyroclastic flows is indispensable when preparing hazard maps, establishing warning and evacuation systems, and planning facilities for use as counter-measures.

In this study, we examined pyroclastic flow fluid models and equations describing its composition, treating pyroclastic flow as a particle flow. We designed a pyroclastic flow simulation model to simulate the flows at Mt Unzendake in 1991, and compared the computed results to the actual statistics. We also examined the relationship between steam generated from contact between flow bed and hot pyroclastic flow, and fluidity of pyroclastic flow.

MODEL AND CONSTITUTIVE EQUATIONS OF PYROCLASTIC FLOW

There are relatively few instances worldwide of pyroclastic flow being directly observed and recorded (e.g. Afamaki, 1973; Mizuyama & Yamada, 1990; and Mizuyama et al., 1990). Unfortunately, the data available covers only a relatively narrow range of the diverse spectrum of pyroclastic flows, and most of these are similar types of flows.

68 S. Yamashita & K. Miyamoto

Pyroclastic flows move along slopes. The force maintaining the flow is believed to be gravitational force, as shown in Fig. 1. Pyroclastic flows consist of a gravitational flow layer of coarse particles in the lower part of the flow (the 'lower layer') and an upper mixed solid/gas layer of fine particles and gas (the 'upper layer'). In this study, we will examine the flow mechanism in the lower layer only, treating the flow as a particle flow.

Kanatani (1984) has proposed a model which takes particle flow to be the flow of completely elastic particles, and the mechanism of energy consumption to be the conceiving friction between particles. Applying the composition equation derived by Kanatani to a two-dimensional shear flow and approximating it slightly gives:

P = 1 200 l-(dct)

1/3 :?>£>2 du (1)

„4/3

200JW i-(c/c.) 1/3 TenaD2 du\

1À du Tz

(2)

Here, P and T represent flow pressure and shearing stress respectively, du/dz is the velocity gradient (z-axis is perpendicular to the flow bed), Te is a constant expressing the state of flow, a is particle density, D is particle diameter, fi is coefficients of friction and of restitution between particles c* is the particle concentration at the time of deposition and c is particle concentration. From equations (1) and (2), the concentration in a state of local equilibrium particle concentration can be obtained by the following equation:

c = 10 h (3)

Here, ie is the energy gradient expressed as TIP. We will use these constitutive equations of Kanatani's as the flow model of the

pyroclastic flow.

Distribution of Velocity

Distribution of Density

Fig. 1 The model of pyroclastic flow.

Model of pyroclastic flow 69

GAS ASCENT AND COEFFICIENT FRICTION BETWEEN PARTICLES

Under normal conditions, dry debris particles flowing from a slope steeper than the angle of repose will usually come to a halt when the gradient decreases. For example, data from Mt Komagatake in 1921 and Mt Fugendake, Unzen in 1991 (Ikeya & Ishikawa, 1991) indicate that pyroclastic flows were deposited at gradients much smaller than 30°, the normal angle of repose for debris flow.

Thus, it can be concluded that apparent JX will always be equal to the smaller than IJL between particles. This suggests that pressure from particle collision and friction between particles has decreased, and that another pressure to support the weight of the particles has developed. The first pressure is presumed to be balanced with the weight of the particles. The new pressure gradient is created by gas filling the pores. Here we will consider the generation of steam via contact between hot pyroclastic products and pore water in the flow bed deposit layer, and gas ascent and the formation of the pressure gradient caused by the generation of steam.

As in Fig. 2, drag F working on a single particle can be expressed as:

F, = ±PCD ^D2

4 (4)

gas density. Using N as the number of particles per unit volume, the pressure gradient is the with ub the velocity of gas ascent from the flow bed. Here, CD is a drag coefficient and p is sum total of drags at that location. If N is represented by the particle concentration c, the pressure gradient can be expressed as follows:

dp dz ^L = NF,

2 2 pCDcub 3D

(5)

Considering the balance of forces in the control volume shown in Fig. 2, the apparent gradient increases via the formation of the pressure gradient. Furthermore, taking into

Fig. 2 The force working on a single particle in pyroclastic flow.

70 S. Yamashita & K. Miyamoto

account the steady state, using \xa as the apparent coefficient of friction between particles, from equation (3) the relationship between \x.a and \x can be expressed as:

Hi COS0

cos0- _i_ap co dz

-co (6)

Using the relationship njix. = tan0/tan0a. Here, dP/dz ^ 0. Thus 6a is proportionally small when the velocity of gas scent and pressure gradient are large, and that 6a is proportionally small when the gradient is large. The decrease in 6 can thus be quantitatively evaluated from ub.

Figure 3 shows the relationship between dimensionless air ascent velocity and gradient determined from an experiment (Yamada et al., 1990) where the experimental channel was set to a gradient lower than the angle of repose in order to examine the relationship between gradient and air ascent velocity from the flow bed. The values obtained in equation (6) (the continuous line in Fig. 3) are consistent with the experimental values. Here CD is assumed to be governed by the shape and concentration of particles; if CD is set to conform to the test results, then CD = 1.7. We believe that the apparent decrease of 6 can be explained by the above mentioned theory from these results.

10000

~»|-a 1000

100

-

ê n

—!_ •—-~<L

o—

d=0.0I9cn - £ - & '

^ - ^ A

n~-̂

. 1

A

•

O

S i -

8-

i

H/d=25

H/d=50

H/d=100

Ca leu I at ion

v t f o \

A

•

\ 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7

tan 8

Fig. 3 The relationship between dimensionless air ascent velocity and gradient.

FUNDAMENTAL EQUATIONS FOR THE NUMERICAL SIMULATION

Pyroclastic flows are described in terms of governing equations for compressible fluids, including the laws of conservation of mass, momentum, and energy, and the state equation. We described these equations by dividing the law of conservation of mass into the law of conservation of flow volume (continuous equation) and the law of conservation of particle volume. Accordingly, we approximated the law of conservation of momentum as an incompressible fluid for our computations. We used a staggered scheme for basic equation differences and used windward differences for space differences in the inertia terms of momentum equations.

Model of pyroclastic flow 71

The following basic pyroclastic flow equations are obtained by taking the z-axis as a vertical axis and the x-y plane as a plane orthogonal to this axis, ignoring the bidirectional components of the flow and averaging through integration in z directions:

dh ^ dM ^ dN _ Q

dt dx by

d(ch)^JzJ(cM)J(cN)_0

dt * dt dx dy

(?)

(8)

dM d&nM) J^nF) udH F J 2 2

-di+v-dx-+v-nr=~8hi*+jyu>»+v>» (9)

dN^(umN)^d(vmN)

dt dx dy udH F„ J 2 2

„4/3

F = 3 2 / H T 1 ~ (c/c J 1/3

Telxa D 2

h

(10)

(H)

Here, M (=umh) and N (=vm/z) are discharge fluxes in the x and y directions respectively; um and vm are x and v-direction components of average velocity, h is flow depth zb flow depth level, H the upper elevation of the flow layer (=zb + h), and /3 .(=4.3) is the coefficient of momentum correction. Also, pt is the density of the flow as a whole, given by Pt = ca.

COMPUTATIONS REPRODUCING THE PYROCLASTIC FLOWS AT MT UNZENDAKE

Reproduction calculations were carried out for the June 3 and June 8 flows at Mt Unzendake in 1991. The flow on June 3 inflicted heavy damage, accounting for 43

0

• \ * •

Dis t r i bu t i on of June 3 flows

; ; Reproduction ca lcu la t i on ! - - - - : (V=2.5 m i l l i on m \ ^=0 .30)

Jigokuato-Crater //

Mt. Fugen ^ ^ t r i ^ i ^ v^n^—-* \V " •

1 2 Km

s"~

^ ^

4y

A Ver

ira ^ J A «-;

1 , : ^ =

Fig. 4 The comparison of calculated extent of pyroclastic flows and actual outcome (June 3 flows).

72 S. Yamashita & K. Miyamoto

-O-

D i s l r i but ion of June 8 f l ows

R e p r o d u c t i o n c a l c u l a t i o n

(V=3.5 m i l l i o n m 1 , H'-O. 1

J igokuato-Crater

A ° Mt.Fugen

2 km

Fig. 5 The comparison of calculated extent of pyroclastic flows and actual outcome (June 8 flows).

dead and missing. Deposited sediments were estimated at 2.5 million m3. And the larger flow on June 8, estimated at 3.5 million m3, reached approximately 5.5 km from the crater, flowing along the Mizunashi River and blocking the channel. There were no casualties, although 73 houses were burned or destroyed (Ikeya & Ishikawa, 1991).

Normally chronological order for the volume of pyroclastic products is required as a boundary condition for calculation, but since this was difficult to estimate, an assumption of a steady supply of products lasting five minutes was made instead. The constants used in the basic equation were D = 0.3 m, a = 2 500 kg m"3 and Te = 1.0. After some exploratory calculation, (x values of 0.28 for June 3 and 0.18 for June 18 were decided upon, in order that the computed reaching distance would agree with the actual reaching distance. Gas ascent from the flow bed is not taken into consideration in reproduction calculations.

Results of reproduction calculations

Figure 4 and Fig. 5 show the calculated extent of pyroclastic flows compared to the actual outcome. In both cases the calculated reach distance and extent of flow were basically consistent with the actual outcome. However, looking at the distribution of sediment deposits in the longitudinal direction if Fig. 6, the actual volume of deposits on the downstream side is relatively small compared to the calculated figure.

The value of n is smaller for the June 8 flow, which had a larger volume of pyroclastic products and reached a point with a gentler gradient. In terms of the characteristics of materials, the value of p. should be roughly the same. The difference in the two y. values adopted seems to be attributable to the fact that a pressure gradient developed, for reasons including the generation of steam on the flow bed by flowing hot pyroclastic products, fi apparently changed because the pressure gradient differed in terms of space form one flow to another. Presumably, the reason why the distribution of deposited sediments was not consistent with the actual distribution is that the formation of the pressure gradient and its spatial distribution, were not taken into account in the reproduction computation.

Model of pyroclastic flow 73

(XIOV) m soo

[ACTUAL]

(X103m3) 600

! i 0 } S

Distance (km)

Q June 4 -

• May 28 -

• May 16 -

- June 16

- June 4

- May 28

TmrH-nrrTi rm

[CALCULATION]

D June 8 0 = 0 . 2 7 )

June 3 0 = 0 . 5 6 )

3.0 4.0 Distance (km)

Fig. 6 The comparison of actual deposits volume and calculated figures.

6.0

GAS ASCENT VELOCITY

Gas ascent is cause dy steam generated through contact between hot pyroclastic products and pore water in the flow bed. Gas ascent velocity ub can be given by the following equation as the quantity of generated steam per unit area and unit time:

uh = Kpwa{l-cx) RT mp0

(12)

Here, m (=18.0153 x 10"3 kg mol"1) is the molecular weight of water, pw (kg m"3) is the density of water, R ( = 8.31451 J mol"1 K"1) is the gas constant, P0 (Pa) is the pressure of generated steam, T (K) is the temperature of generated steam and K (S"1) is the proportion evaporated pore water per unit time. Also, d (m) is the layer thickness by which heat exchange with hot pyroclastic products takes place in the flow bed deposit layer containing pore water. Here, assuming that RT/mP0) is constant, ub of gas depends on the conditions of the flow bed.

Here, let us infer gas ascent velocity and a, taking the example of pyroclastic flows at Mt Fugendake, Unzen for which reproduction calculations were carried out. It is assumed that n is 0.6, the general value for debris, and therefore from the results of reproduction calculations that the apparent coefficient of friction between particles is 0.2. Also assumed is the value tan0 = 8.1. p = 0.598 x 10"3 kg m"3 is taken as the

74 S. Yamashita & K. Miyamoto

gas density, assuming 100°C, 1 atm steam. When the gas ascent velocity is determined from equations (3), (6) and (12) using these values, the value obtained for ub is 80 m s"1. Assuming that c» = 0.6 and pw = 1000 kg m"3, j8a = 0.12 m s'1.

Thus, assuming that pore water exists on the flow bed, it can be concluded that the deposit layer is rapidly eroded and mixed and that a high gas ascent velocity develops from the evaporation of pore water, significantly reducing friction between particles.

CONCLUSION

The calculations reproducing the pyroclastic flows at Mt Unzendake, using a fluid model of particle flow, were able to express fairly well the extent of deposition of the lower layer. This seems to indicate that, in general, the movement of the lower layer of a pyroclastic flow can be expressed by a particle flow composition equation, taking into account energy dissipation by fx only, and the coefficient of friction between particles.

This work has explained, and examined ways of evaluating the phenomenon whereby fx apparently decreases as a result of the formation of a pressure gradient by gas ascent velocity, with corresponding increase in the fluidity of the pyroclastic flow.

However, the technique of assigning drag coefficients for evaluating pressure gradients and the thickness of the heat exchange layer poses a continuing problem, where much research is needed.

FUTURE WORK

We are planning to prepare numerical simulation models which take into consideration the formation of pressure gradients. It is hoped that the accuracy of pyroclastic flow simulation can be improved by comparing results obtained with actual statistics from Mt Unzendake and other volcanoes.

REFERENCES

Aramaki, S. (1973) Small-scale pyroclastic flows generated during the February-March, 1973, eruption of Asama Volcano. Bulletin of the Volcanological Society of Japan 37'(3), 74-94.

Ikeya, H. & Ishikawa, Y. (1991) Disasters caused by pyroclastic flows and debris flow at Unzen Volcano in 1991. Journal of the Japan Society of Erosion Control Engineering 44(2), 46-56

Kanatani, K. (1984) A theory for the flow of granular materials (2nd Report, Developed flow). Transactions of the Japan Society of Mechanical Engineers 45(B), 515-520.

Mizuyama, T., Yamada, T., Yajima, S. & Shimoda, Y. (1990) Motion of the pyroclastic flows occurred at Mount. Semeru Volcano in 1989. Journal of the Japan Society of Erosion Control Engineering 43(3), 13-19.

Yamada, T., Mizuyama, T. & Yajima S. (1990) Experiments of pyroclastic flow. Proceedings of the 1990 Japan National Conference of the Japan Society of Erosion Control Engineering, 320-323.

Sediment Problems: Strategies for Monitoring, Prediction and Control (Proceedings of the Yokohama _^. Symposium, July 1993). IAHS Publ. no. 217; 1993. 75

Study for prediction of occurrence of hillside landslides

S. HIRAMATSU Pacific Consultants Co. Ltd, 7-5, Sekido 1-chôme, Tama city, Tokyo 206, Japan

T. MIZUYAMA Faculty of Agriculture, Kyoto University, Kitashirakawa, Sakyo-ku, Kyoto 606, Japan S. OGAWA Faculty of Agriculture, Ehime University, 3-5-7, Tarumi, Matsuyama 790, Japan

Y. ISHIKAWA Public Works Research Institute, Ministry of Construction Government of Japan, Asahi-1, Tsukuba city, Ibaraki 305, Japan

Abstract In this study, a hillside landslide model is presented that takes into consideration: (a) The mountain slope distribution in a three-dimensional space: and (b) The transient flow of water in saturated-unsaturated soils which

profoundly affect the potential occurrence of hillside landslides and the time of occurrence, was proposed.

This simulation model was applied to a basin where many hillside landslides occurred. The model predicted locations of landslides and was accepted as valid. Evaluation of parameters affecting the stability of the mountain slope were made through numerical experiments. The following matters were confirmed by this study: (a) The model simulated accurately changes of groundwater level with

time for hillsides of different shapes. (b) The value of the hydraulic conductivity is most important in

predicting the timing of hillside landslides.

NOTATION

C specific moisture capacity c cohesion of soil Gs specific gravity of soil particles h groundwater depth Ix and Iy hydraulic gradients in x-axis and y-axis directions k hydraulic conductivity ks saturated hydraulic conductivity n porosity qx and q fluxes in x-axis and y-axis directions qz amount of water supplied from unsaturated domain t time X,Yand Z distances along X-axis, F-axis and Z-axis directions /3 angle of inclination of unit slope

76 S. Hiramatsu et al.

y w unit-volume weight of water 6 water content in percent of total weight X effective porosity 4> internal friction angle of soil (p moisture tension Az vertical soil layer division width.

INTRODUCTION

Hillside landslides frequently occur in many parts of Japan in the period of heavy rains every year, thus forming a major source of sediment production in the basins. Hillside landslides are triggered by heavy rain and are explained as the physically transient process of the surface soil layer on the mountain slope. In this study, a model of hillside landslides was proposed through saturated-unsaturated infiltration analysis, taking note of the storm water hydrological process in the interior of the surface soil layer on the mountain slope, and we studied the effectiveness of the model by applying it to a small basin which has experienced frequent hillside landslides.

GENERAL DESCRIPTION OF STUDY AREA

The study area is a small basin (catchment area: 0.52 km2) in the upper stream of the Tenryu River basin in Japan as depicted in Fig. 1. In this basin, many hillside landslides occurred in the heavy rain caused by Typhoon No . 10 in 1982, affecting 1.33 % of the area. The amount of continuous rainfall was 403.0 mm and the maximum hourly rainfall intensity was 45.0 mm.

PREPARATION OF HILLSIDE LANDSLIDE M O D E L

Construction of the model

To predict the location of the hillside landslides and its scale and determine the degree of danger of basin collapse, it is necessary to evaluate the storm water hydrologie process on a hillslope in a realistic manner. In this study, hillside landslides were regarded as a phenomenon whereby a slope surface layer moves as it becomes destabilized from decrease of soil strength due to increased weight brought about by the infiltration of storm water into the interior of the hillside slope or the formation of a groundwater zone and the rise of saturation.

In this model, a hyetograph was modified using the concept of the theory of the one-dimensional vertical unsaturated infiltration flow. Storm water supplied to the bedrock surface was traced on the basis of this one-dimensional vertical unsaturated infiltration analysis as a saturated throughflow by Darcy 's law (Ota et al., 1983). Fo r the construction of the model, the basin was mesh-divided as depicted in Fig. 2, providing each mesh with a vertical soil column using the thickness of the surface soil layer as its height, and this column was used as the basic element of the analysis (Hiramatsu et al., 1990). Storm water supplied to the basic element infiltrates the

Prediction of occurrence of hillside landslides 11

32° N

-40° N

36° N

Tokyo

Study area

130° E 134° ENa8°ya

Fig. 1 Location of the study area.

unsaturated soil layers, reaches the bedrock, forms a saturated zone and moves between the elements as a saturated throughflow.

Infiltration analysis

Equation (1) led by Richards (Richards, 1931) by extending Darcy's law to the unsaturated domain was applied in analyzing the one-dimensional vertical unsaturated infiltration process of storm water in the unsaturated zone from the slope surface layer to the surface of the bedrock:

C - ^ + A at az

ot<P _ i

az (1)

The saturated throughflow process was traced, using the amount of water supplied to the surface of the bedrock by equation (1). The equation of continuity in this saturated throughflow process is expressed by:

x m ah ^ mx ^ <*%

at ax ay (2)

( A ) ( B )

C e l l

Za

C e l l

Zb Z c , £ -

a * - '

^T

q ^ -

- i -

Fig. 2 Hydrologie process on a hillslope.

N

^ N q x l

( C )

<Jy2

Qx2

w = (13 °

IS

S qyl

78 S. Hiramatsu et al.

Also, the equation of motion is given by:

according to Darcy's law.

h-k- L (3)

Slope stability analysis

In a hillside landslide, generally the collapse length is greater than the collapse depth and the slip plane is often flat. So, estimation on slope stability was analyzed through the equation of stability analysis on an infinite-length slope:

Fs = {c + ff0 - h • yw) • cos2(3 • tan<£}/(<r0 • cos/3 • sinjS) (4)

i{6(Zi).yw + (l-n)-Gs-yw}-Az (5)

APPLICATION OF HILLSIDE LANDSLIDE MODEL

Conditions of analysis

The input conditions necessary for carrying out hillside landslide simulation were set for each of the unit slopes obtained by dividing the basin, using 25 m meshes, on the basis of a 1/5000 topographic map. Thickness of the surface soil layer was given at intervals of 0.1 m, using the range of 1.0 ~ 1.5 m for every unit slope and based on the collapse depth of a slope where a landslide had occurred in the past. For soil data,

Tine (hr)

Parallel slope

0 20 40 60 80 100 120 140 160 180 200

Tine (hr)

Fig. 3 Change with time of groundwater level for the difference of topographic conditions.

Prediction of occurrence of hillside landslides 19

the results of a soil test by soil specimens taken from the basin was used. Our numerical simulation was carried out by inputting rainfall in Typhoon No. 10 of 1982 and using the time of rainfall start as the time of computation start.

Response characteristics of groundwater level for the difference of slope shapes

Figure 3 shows changes with time of groundwater level for the difference of topographic conditions for one typical slope selected from each of concave, parallel and convex slopes in the basin. In all slope shapes, groundwater levels began to rise suddenly 25 hours after the start of the rain. Then, the degree of rise began to decline 40 hours after the time of start of computation, i.e. the time of occurrence of peak rainfall. This decline of the degree of rise of the groundwater level was particularly noticeable for the convex slope. This probably occurs because the supply of storm water from the ground surface to the soil layers stops after the end of rainfall and the amount of one-dimensional vertical unsaturated infiltration decreases; furthermore, the amount stored in the interior of the slope flows toward the slopes in the vicinity as saturated throughflows for the convex slope.

Occurrence of slopes with hillside landslides

Figure 4 shows the distribution of slopes with hillside landslides obtained by numerical computation. It indicates that slopes with safety ratios of less than 1.0 are generally consistent with slopes where landslides actually occurred and replicates fairly well the trend for landslides to occur in sites such as the upper reaches of the left tributary, the lower reaches of the right tributary and the lowest reaches of the basin itself. But in reality, landslides occurred on a few slopes with safety ratios of 1.0 or higher although generally on slopes with safety ratios of under 1.0. This may be attributable to the dispersion of input conditions such as the amount of rainfall, the thickness of the surface soil layer and soil strength and the accuracy of topography reproduction by meshes.

EVALUATION OF IMPACTS OF LANDSLIDE-CAUSING FACTORS

When using a numerical simulation model to determine whether hillside landslides will occur, the results are greatly affected by topographic conditions, such as the thickness of the surface soil layer and the slope gradient, by saturated hydraulic conductivity which governs the flow of storm water in the interior of slope soil layers and by soil strength constants in case that the amount of rainfall is constant. Here, we took note of surface soil layer thickness and saturated hydraulic conductivity — i.e. all the above factors, which can be determined relatively easily by topographic measurement, field exploration, etc. - and evaluated the extent to which these factors affect the stability of slopes, using the hillside landslide model mentioned in the preceding chapter.

Changes in slope stability with changing surface soil layer thickness of 120, 130 and 140 cm were examined. It was found that the predicted likelihood of a landslide

80 S. Hiramatsu et al.

25 30

V : Sites of hillside landslides occurrence obtained by numerical calculation

Q : Sites of hillside landslides occurrence during heavy rainfall in 1982

Fig. 4 Distribution of slopes with hillside landslides.

increased with the thickness of surface soil layer. Also, comparing the change of groundwater levels and slope safety ratios by surface soil layer thicknesses, it was found that, during and after rainfall, the probability of slope collapse was high in proportion to the thickness of the surface soil layer.

It is supposed that saturated hydraulic conductivity greatly affects the change of groundwater depth because storm water which has infiltrated the surface soil layer and thereafter reached the surface of the bedrock can be calculated by Darcy's law (equation (3)) as a saturated throughflow. Change of slope stability pursuant to the change of saturated hydraulic conductivity was investigated using different values of saturated hydraulic conductivity: ks = 0.002, 0.005 and 0.010 cm/s. It was found that the time of occurrence of hillside landslides was advanced with the increase of saturated hydraulic conductivity and the time taken to reach the final distribution of hillside landslides decreased accordingly. We also confirmed that differences in saturated hydraulic conductivity on the behaviors of the groundwater level began to have an effect from when the saturated throughflow exceeded water supplied from the unsaturated domain, and that the time of this appearance was advanced with the increase of saturated hydraulic conductivity. We deduced from these results that hillside landslides on concave slopes by the influx of a saturated throughflow are more likely with high saturated hydraulic conductivity, while on convex slopes, hillside landslides are less likely by increased runoff. Thus, it was concluded that saturated hydraulic

Prediction of occurrence of hillside landslides 81

conductivity is an important factor in predicting the occurrence of hillside landslides and especially the time of their occurrence.

FUTURE PROBLEMS

We have presented a saturated-unsaturated infiltration analysis for hillside landslides as a phenomenon of sediment production triggered especially by heavy rain, and have proposed a hillside landslide occurrence model which takes into consideration storm water infiltration and flow into the hillside surface soil layer. This model proved effective in predicting hillside landslides when applied to the local basin. However, there are still no established methods for setting surface soil layer thickness (especially values for slopes without landslides), saturated hydraulic conductivity and soil strength (dispersion from place to place), factors believed to be important in predicting hillside landslides. It is necessary therefore to establish methods for setting these parameters hereafter, to improve analysis accuracy and, at the same time, to develop a universal model by applying it to other basins as well.

Acknowledgements We would like to express our gratitude to Dr Hiroyuki Yoshimatsu of Ministry of Construction Government of Japan for his advice in conducting the numerical simulation in this paper.

REFERENCES

Hiramatsu, S., Mizuyama, T. & Ishikawa, Y. (1990) Study of a method for predicting hillside landslides by analysis of transient frow of water in saturated and unsaturated soils. J. Japan. Saabo. Soc. 43(1), 5-15 (in Japanese).

Ota, T., Fukushima, Y. & Suzuki, M. (1983) Research on runoff from hillsides by one-dimensional transient saturated-unsaturated flow. J. Japan. For. Soc. 65(4), 125-134 (in Japanese).

Richards, L. A. (1931) Capillary conduction of liquids through porous media. Physics 1, 318-333.

Sediment Problems: Strategies/or Monitoring, Prediction and Control (Proceedings of the Yokohama Symposium, July 1993). IAHS Publ. no. 217; 1993.

Inference of landslide susceptible areas by Landsat Thematic Mapper data

L. SAMARAKOON Research and Development Center, Nippon Koei Co. Ltd, 2304 Takasaki, Kukizaki, Inashiki, Ibaragi 300-12, japan

S. OGAWA, N. EBISU, R. LAPITAN College of Agriculture, Ehime University, Tarumi 3-5-7, Matsuyama 790, Japan

Z. KOHKI College of Agriculture, Ryukyu University, Nishiharamachi, Okinawa 903-01, Japan

Abstract Geological structure and the exceptional water retention characteristic of soil, said to be the major factors that causes landslide in the upper basin of the Yoshino river in Shikoku Island, Japan. An indirect approach to estimate landslide susceptible areas in this river basin was investigated using Landsat Thematic Mapper (TM) data. Inference of landslide potential areas was made by investigating moisture status in forest canopy deduced by Landsat TM data. Choosing suitable TM data sets, attempt was made to estimate the forest cover that represents relatively high moisture concentration in two different seasons; just after a long spell of rainfall and following a considerable dry period. The scene acquired in September belonged to the wet season of 1986, and November scene represented a dry season of the same year. Forest cover that has been subjected to seasonal changes during the two dates was excluded from the study by comparing Normalized Difference Vegetation Index (NDVI) of the two dates. Subsequently, samples were randomly selected over the forest cover estimated as unchanged, and corresponding digital counts of TM band 3,4, and 5 of both dates were extracted. Factor analysis was carried out using the two dates band ratios 4/3 and 4/5 of the selected samples. Physical meaning of the resultant factors were interpreted according to the spectral properties of the bands that showed high correlation with respective factors. Subsequently, factor score scatter diagram of the samples was plotted and the samples that represented relatively high moisture in both the dates were extracted. These samples were utilized in separating relatively high moisture areas within the river basin using Landsat data. The extrapolation of training results was based upon a maximum likelihood classifier. The above estimated high moisture areas were compared with the existing landslides map of the area, and it was found that most of the landslides were concentrated over the estimated high moisture areas. This showed that the areas with relatively high moisture in both dates are more susceptible for landslides than other areas. Therefore, it could be said that the Landsat data acquired in rainy and dry season can be used for preliminary investigation of landslides susceptible areas as well as to monitor landslides in an area.

84 L. Samarakoon et al.

INTRODUCTION

An ideal land and water management plan should account and take care of the surface parameters and their changes in time scale. Landslide is one of the phenomenon that account for land degradation, which could cause mass destruction to life and property. Remotely sensed data has been used in land and water resources management since early sixties. Through wide range of scientific and application research works, it has been demonstrated that satellite technology can effectively be used in observing spatial parameters like surface water, land use, snow cover, surface temperature, etc. Estimation and monitoring of landslides is becoming an active field of remote sensing technology, but the appropriateness in practical application is yet to be assessed.

Landslide is a phenomenon that causes mass movement of earth due to factors such as geology, ground water, land cover. In fractured or volcanic altered zones, presence of excess ground water could influence landslides by transforming soil into soft clay. It has been shown that ground water in these geologically altered areas influences the generation of landslide (Takagi & Murai, 1991). Direct measurement of ground water in a large area is practically not viable, therefore prediction through indirect approaches have to be developed. Fractured zones in the bedrock may show high water content in the soil, and this may be detected by luxuriant plant growth in that area. This could

Fig. 1 Location of the study area.

Inference of landslide susceptible areas by Landsat 85

easily be observed by remotely sensed data, and could be utilized in predicting the state of ground water. This follows that investigation on canopy moisture content could lead to understand the ground water distribution, hence landslide proven areas. Also, spatial limitations and the difficulty of continuous investigations could be answered with the development of remote sensing techniques.

Tucker reported (Tucker, 1977) that 0.63 to 0.69 Mm is the best spectral region for estimating low level of chlorophyll and leaf water content, and spectral response of this region has an inverse relationship with chlorophyll and leaf moisture. Tucker demonstrated (Tucker, 1980) that, 1.55-1.75 Mm spectral window is superior than the other spectral transmission windows in 0.7-2.5 Mm spectral region in monitoring moisture stress in plant canopy by satellite remote sensing. Cohen illustrated (Cohen, 1991) that leaf water potential is highly correlated with vegetation indices than any single Landsat Thematic Mapper band, and the middle-infrared to near-infrared ratio with the highest correlation. Thematic Mapper band 4 has the lowest response, and band 5 has the highest response to change in leaf water content.

These are some of the research works that show the possible use of spectral investigations in detecting moisture stress in vegetation. In this study, attempt was made to estimate the spatial locations of vegetation that have relatively high moisture level irrespective of seasonal changes using Landsat Thematic Mapper (TM) data, and their correlation with landslide areas mapped by other means. The research concept was developed on following assumptions: (a) After a lengthy rainy period the vegetational cover moisture density will be

relatively high due to presence of excess soil moisture. (b) A prolonged dry season could lower ground water table, hence reducing moisture

in the soil, which could induce water stress in vegetation. (c) The areas that have excess ground water may not develop appreciable water

deficiency with annual precipitation patterns.

MATERIAL AND STUDY AREA

Investigation was carried out in the upper stream of Yoshino river basin in Shikoku island, Japan. The total catchment area of the river basin is about 3750 km2. A study area with about 1000 km2 was selected for the present study as depicted in Fig. 1. Geology of the river basin is notified by the Median tectonic line and the Mikabu tectonic line. The area between Median and Mikabu tectonic lines are designated as Sanbagawa region, and the area to the south of Mikabu tectonic line is referred to as Chichibu region, as shown in Fig. 1. The most important lithologies in the Sanbagawa area are crystalline schist including black, green and quarts, and in the Chichibu region is Mikabu green schist. All of these bedrocks contain several weak places due to influence of tectonic movements, and this area is identified as one of the Japanese top fracture zones. Landslides have been in active in this area for centuries and presently under active landslides prevention projects. It is said that the course for the landslides in this area is mainly due to water retention characteristic of the soil. The mean annual rainfall of the area is about 2500 mm. The land cover types of the study area are dominated by forests, paddy fields, crop lands, orchards and few water bodies. According to published land use maps, the forest cover of the study area is about 85% of the total area.

86 L. Samarakoon et al.

Rainfall data of the area from 1984 to 1990, and information of TM data available for the same period was studied to select two scenes that meets the requirements of the research concept. The scene acquired on 3 September 1986, which was after a considerable rainy period, and 6 November 1986, which was in the dry season, were selected for the analysis. Table 1 shows rainfall within the study area for the year, 1986.

Black and white photographs acquired in 1985 were obtained for selection of reference data, and to verify satellite data classification results. Also, landslide map of the area was acquired for assessment of classification accuracy.

IMAGE REGISTRATION AND GEOCODTNG

1:50 000 topographical maps were acquired and used as base-maps for georeferencing satellite data. This was carried out in two steps. Initially the two scenes were overlaid

Table 1 Precipitation observed within the study area during 1986.

1 2 3 4 5 6 7 8 9

10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31

Total

Jan

35.5

0.5

1.5 0.5

38.0

Feb

0.5

0.5 3.5

2.5 67.5

1.5

76.0

Mar

6.5 6.5 1.5

37.5 8.5

1.0

30.0 9.0 7.5 0.5

13.5 25.0

1.0 6.5

Apr

9.5

105.0 5.0 0.5

45.0 15.5

17.0 7.5

20.0 5.0

2.5 21.0 6.0

0.5

May 11.0 6.5

21.0 2.0

22.0 17.0

28.5 67.0

70.0

19.5

June

3.0 153.0 136.5

1.0 29.0 10.0 5.5

26.5

1.5

59.5

July

0.5 1.0

12.5 6.5 4.5 8.0 6.0

2.5 209.0 31.0

35.5 6.0 9.5 3.5

0.5

2.0

Aug

18.0

2.0

5.0

30.5

17.5

6.5 2.5

0.5

4.5 161.5

154.0 260.0 264.5 436.5 339.0 248.5

Sep

gM 5.5

16.5 22.0 11.0

35.0

6.0 15.0 30.0 8.0

23.0 10.5

109.5 35.0

327.0

Oct 22.0

8.5

9.0

0.5

8.0

2.5

50.5

Nov

7.0

MN

17.5

0.5

10.0

35.0

Dec

2.0 7.0

4.0 22.5 0.5

96.0 1.0

0.5

5.0 2.0 3.5 2.0 1.5

147.5

Represents the dates that Landsat TM data acquired for the study

Inference of landslide susceptible areas by Landsat 87

by taking September scene as the master image. This was followed by the georeferencing of master image with topographical maps. Subsequently, the November image, which was transformed onto September scene in earlier step also registered over the same base-map. A bilinear polynomial using least square adjustment was used for geometric transformation, and pixels in both of the scenes were resampled into 30 x 30 m pixels using the nearest neighborhood method. The standard error for the prediction of control points of map from September image was less than 10 m in both of the arbitrary selected X and Y directions.

METHOD OF ANALYSIS

NDVI images of the georeferenced TM data sets were generated using the bands given in the following formula:

NDVI = (band4 - band3)/(band4 + band3)

The TM band 3 and 4 in the above equation belongs to the red and nearinfrared regions in the electromagnetic spectrum, respectively. Subsequently, a new image was created with the difference of NDVI images of two dates to extract the unchanged forest cover in both of the dates. The term unchanged stands for forest cover with uniform aerial coverage in both dates, and that have not subjected to any seasonal changes. Sensible comparison of two dates satellite data for any phenological or moisture change can be carried out if the above mentioned factors remain same. The NDVI was used here as it reduces the influence of the background on the canopy composite reflectance (Clevers, 1988), and partly compensate for illumination conditions and atmospheric effects (Kaneniasu et al., 1979 & Tucker et al., 1979). Samples were chosen randomly over the unchanged forest land, and digital counts of Landsat TM band 3, 4 and 5 of both dates were extracted. These samples were analyzed by factor analysis, which identifies a relatively small number of factors that can be used to represent relationship among a set of many, interrelated variables (Davis, 1973). Also, this helps to identify underlying factors and permits a way to understand what the data are really measuring. Instead of digital counts itself, band ratios 4/5 and 4/3 of both of the dates were used in factor analysis to minimize atmospheric effect, illumination differences on temporal images, and to enhance the interpretation of the condition of vegetation. On the results of factor analysis the selected samples were grouped into two; high moisture in both dates and relative deficit in moisture, in the following manner.

(a) The physical meaning of the most significant two factors were interpreted according to the factor loadings and spectral characteristics of the TM bands used.

(b) Having determined the physical meaning of each factor, the positions of each sample on the factor score plot were examined and grouped into above mentioned classes Concerning the physical representation of the factors. The statistical parameters of the estimated classes were used to classify whole area

into high moisture forest cover using maximum likelihood technique. Finally, distribution of high moisture areas were compared with available landslides maps of the area.

88 L. Samarakoon et al. t

RESULTS AND DISCUSSION

Rainfall data in Table 1 shows that there were more rainy days than clear skies for about two months before the acquisition of the September scene. Specially 162 mm. precipitation five days before the satellite pass would have effectively contributed to soil moisture and ground water in the study area. From September 22 to November 6 there was no effective rainfall except for less intensive and sporadic rainfall. Therefore, it would be reasonable to assume that this period could have introduced an appreciable water deficiency in soil as well as in forest canopy.

Theoretically, the difference of two dates NDVI values of the unchanged land cover should be zero. The resultant image showed a deviation from this expectation. This could be due to variation in atmospheric condition and difference in illumination in two dates, which may not have fully compensated using NDVI. Therefore, referring to aerial photographs samples were selected, and probable range of the NDVI of unchanged forest cover was established. Applying this range, unchanged forest cover was established and random samples were selected for further analysis.

The factor loadings for the randomly selected samples over unchanged forest cover is shown in Table 2. It shows ratio 4/3 of both dates has a high contribution to factor 1. Ratio 4/5 of both dates has a high correlation with factor 2. Figure 2 shows

Table 2 Factor loading and the distribution of Eigen values for the first two factors.

September 4/5 September 4/3 November 4/5 November 4/3

Eigen value

Factor 1

0.0541 0.7163 0.5089 0.9243

2.3940

Factor 2

0.5931 .. 0.0850

0.4208 0.0878

0.9870

a laboratory experiment conducted on maize leaves (Woolley, 1971). It can be deduced from this figure that ratio 4/5 decreases with reduction of leaf water. This gives the behavior of red reflectance with near-infrared reflectance and canopy leaf area index. Also, it can be deduce that ratio 4/3 is sensitive to leaf area index or the presence of green biomass. On these explanation and properties of ratio 4/5 and 4/3 the factor axes of this research could be interpreted as follows:

Factor 1: Correlation with green biomass or amount of leaf cover.

Factor 2: Represents moisture in two dates. High value depicts relatively high moisture in both dates.

The factor scatter diagram of the samples is shown in Fig. 3. On the above interpretations the samples in quadrant 1 and 2 could be interpreted as samples from areas with relatively high moisture in both dates when compared to samples in third and fourth quadrants. Samples in first and fourth quadrants are high in leaf cover than samples in other two quadrants. Therefore, samples in first quadrant could be classified

Inference of landslide susceptible areas by Landsat 89

1.0

0.5-

a

- 13114!

• ''7 1 / t / 1/

•

1 1 1 . 1

7^ i

ï . ,

1 5 1 Landsat TM

\ dried V ^ \ / Y r leaves

\ 1« ^ fresh \ i leaves v-^\

1 A ! ' *

, i , i , i i , i i , i , 0 0.5 1.0 1.5 2.0 2.5

wavelength |im

Fig. 2 Reflectance characteristics of fresh and dried Maize leaves on a laboratory experiment (modified from Woolley, 1971).

as relatively high in moisture and leaf cover. Samples in second quadrant could be identified as high in moisture but relatively low leaf cover than samples in first quadrant. On this interpretation the study area was classified into high moisture areas in two dates using the statistics of samples in first and second quadrants. The classification was carried out using maximum likelihood method. In the classification stage samples in two quadrants were not combined into one group to avoid possible multimodel distribution of training samples.

The resultant image of high moisture areas was compared with the landslides map of the area produced by Ministry of Construction upon field surveys. The surveys have been confined only to areas where the destruction has a severe impact on inhabitant of the area. The comparison found that most of the estimated high moisture areas by satellite data interpretation were distributed over landslides proven areas. Therefore,

--1

- i«

FACTOR 2

'2

ft A __ . A V

-\ • FACTOR 1 e

• c

Fig. 3 Factor score scatter diagram of the random samples selected over the unchanged forest cover in the study area.

90 L. Samarakoon et al.

it could be said, areas that indicated high moisture could be landslide hazard regions, or the potential of occurring landslides in estimated high moisture areas is greater than other areas.

CONCLUSION

In this study it was found that satellite data could be used to estimate soil moisture through interpreting the state of the vegetation. As there is a high correlation of soil moisture to ground water in landslide proven areas in the Yoshino river basin, this facilitates an indirect approach to estimate landslide susceptible regions. Classification result of this study was centered on factor analysis, which gives relative characteristics of the sampled being used. Therefore, it is important to choose samples that have similar land cover to detect moisture differences. This research was carried out on unchanged forest cover in both dates, which could be considered as coniferous forest, but selecting suitable satellite images other areas could be analyzed by the same procedure. The observed high moisture areas in this study will assist in easy location of sites for detailed field investigation for identifying areas that are highly susceptible for landslides. Finally, it could be said that this method could be used as a preliminary approach in landslide studies where the detailed investigation could be restricted to a confined area.

Acknowledgement This report was funded by the Yoshinogawa office of the Ministry of Construction. Thank goes to all the officer who extend their services in carrying out this research work.

REFERENCES

Davis, J. C.(1973) Statistics and Data Analysis in Geology. John Wiley , New York. Clevers, J. G. P. W. (1988) The derivation of a simplified reflectance model for the estimation of leaf area index.

Remote Sensing of Environment 25, 53-69.

Cohen, W. (1991) Response of vegetation indices to changes in three measures of water stress. Photogr.ammetric Engineering & Remote Sensing 57(2), 195-202.

Kanemasu, E. T., Demetriades-Shah, T. H. & Su, H. (1990) Estimating grassland biomass using remotely sensed data. Chapter 12 in: Application of Remote Sensing in Agriculture. Butterworths Press, London.

Takagi, M. & Murai, S. (1991) Inference of landslide area from LandsatTM and DTM data. Proceedings of 1 lth Asian Conference on Remote Sensing J4, 1-5.

Tucker, C. J. (1977) Asymptotic nature of grass canopy spectral reflectance. Applied Optics 16(5), 1151-1156. Tucker, C. J. (1980) Remote sensing of leaf water content in the near infrared. Remote Sensing ofEnvironment 10,

23-32. Woolley, J. T. (1971) Reflectance and transmittance of light by leaves. Plant Physiol. 47, 656-662.