+ All Categories
Home > Documents > Impact of vegetation fraction from Indian geostationary satellite on short-range weather forecast

Impact of vegetation fraction from Indian geostationary satellite on short-range weather forecast

Date post: 23-Nov-2016
Category:
Upload: prashant-kumar
View: 219 times
Download: 5 times
Share this document with a friend
11
Agricultural and Forest Meteorology 168 (2013) 82–92 Contents lists available at SciVerse ScienceDirect Agricultural and Forest Meteorology jou rn al h om epa ge: www.elsevier.com/locate/agrformet Impact of vegetation fraction from Indian geostationary satellite on short-range weather forecast Prashant Kumar a,, Bimal K. Bhattacharya b , P.K. Pal a a Atmospheric and Oceanic Sciences Group, EPSA, Space Applications Centre (ISRO), Ahmedabad 380015, India b Agro-Ecosystems Division (AED), ABHG, EPSA, Space Applications Centre (ISRO), Ahmedabad 380015, India a r t i c l e i n f o Article history: Received 22 December 2011 Received in revised form 20 July 2012 Accepted 16 August 2012 Keywords: Data assimilation Numerical weather prediction Vegetation Geostationary satellite Sensible heat fluxes Scintillometer a b s t r a c t Indian economy is largely depending upon the agricultural productivity and thus influences the trade among the SAARC countries. High-resolution and good-quality regional weather forecasts are necessary for planners, resource managers, insurers and national agro-advisory services. In this study, high reso- lution updated land-surface state in terms of vegetation fraction (VF) from operational vegetation index products of Indian geostationary satellite (INSAT 3A) sensor (CCD) was utilized in numerical weather pre- diction (NWP) model (e.g. WRF) to investigate its impact on short-range weather forecast over the control run. Results showed that the updated vegetation fraction from INSAT 3A CCD improved the low-level 24 h temperature (18%) and moisture (10%) forecast in comparison to control run. The 24 h rainfall forecast was also improved (more than 5%) over central and southern India with the use of updated vegetation fraction compared to control experiment. INSAT 3A VF based experiment also showed a net improve- ment of 27% in surface sensible heat fluxes from WRF in comparison to control experiment when both were compared with area-averaged measurements from Large Aperture Scintillometer (LAS). This trigg- ers the need of more and more use of realistic and updated land surface states through satellite remote sensing data as well as in situ micrometeorological measurements to improve the forecast quality, skill and consistency. © 2012 Elsevier B.V. All rights reserved. 1. Introduction Weather and climate are the biggest risk factors that have impacts on agricultural planning and management. Agriculture is the backbone of Indian Economy which influences international trade among SAARC (South Asian Association for Regional Cooper- ation) countries. About 65% of Indian population depends directly on agriculture and it approximately accounts for 22% of GDP (gross domestic product). Any abnormalities in the short-range weather forecast would seriously affect the growth and final yield of the crops. The one of the foremost aims of weather forecasting is to ren- der timely advice to farmers on the actual and expected weather, and its likely impact on the various day-to-day farming opera- tions. Especially for the nations (like India) largely depending on the agrarian economy, reliable high-resolution (up to 15 km) regional weather forecasts are long-pending demands from crop forecasters, planners, policy makers and for management of water resources, Corresponding author at: Atmospheric Sciences Division, Atmospheric and Oceanic Sciences Group, Earth, Ocean, Atmosphere, Planetary Sciences and Appli- cations Area, Space Applications Centre (ISRO), Ahmedabad 380015, India. Tel.: +91 79 26916052; fax: +91 79 26916075. E-mail addresses: [email protected], [email protected] (P. Kumar). agricultural disasters such as drought, flood, severity of certain pests and diseases (Strand, 2000). Land-surface processes through Soil–Plant–Atmosphere Continuum (SPAC) are important drivers for weather and climate systems over the tropics and particularly over the Indian sub-continent (Van der Tol et al., 2008). Realistic representation of land-surface states over the Indian region would help accurate simulations of environmental processes at micro, meso, and regional climate scales (Matsui and Lakshmi, 2005). However, in order to achieve these potential benefits, it is necessary to develop a strategy which will address and overcome the different challenges associated with the representation of the land-surface states over the Indian monsoon region. The problem of determining a physically consistent and accu- rate snapshot of the atmosphere is the central theme of numerical weather prediction (NWP). In succeeding decades, with progress in both computing power and optimization strategies, more sophis- ticated constraints and more diverse observations have been included in NWP (Kumar et al., 2011). Generating an accurate ini- tial state is recognized as one of the biggest challenges in NWP model prediction of weather events. The rapid increasing computer power has led to higher (1–15 km) resolution NWP models, which are able to resolve mesoscale features and thus to give more pre- cise forecasts. These high resolution forecasts have the potential to drastically change the forecasting procedure to determine the 0168-1923/$ see front matter © 2012 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.agrformet.2012.08.009
Transcript
Page 1: Impact of vegetation fraction from Indian geostationary satellite on short-range weather forecast

Iw

Pa

b

a

ARRA

KDNVGSS

1

ittaodfcdatawp

OcT

0h

Agricultural and Forest Meteorology 168 (2013) 82– 92

Contents lists available at SciVerse ScienceDirect

Agricultural and Forest Meteorology

jou rn al h om epa ge: www.elsev ier .com/ locate /agr formet

mpact of vegetation fraction from Indian geostationary satellite on short-rangeeather forecast

rashant Kumara,∗, Bimal K. Bhattacharyab, P.K. Pala

Atmospheric and Oceanic Sciences Group, EPSA, Space Applications Centre (ISRO), Ahmedabad 380015, IndiaAgro-Ecosystems Division (AED), ABHG, EPSA, Space Applications Centre (ISRO), Ahmedabad 380015, India

r t i c l e i n f o

rticle history:eceived 22 December 2011eceived in revised form 20 July 2012ccepted 16 August 2012

eywords:ata assimilationumerical weather predictionegetationeostationary satellite

a b s t r a c t

Indian economy is largely depending upon the agricultural productivity and thus influences the tradeamong the SAARC countries. High-resolution and good-quality regional weather forecasts are necessaryfor planners, resource managers, insurers and national agro-advisory services. In this study, high reso-lution updated land-surface state in terms of vegetation fraction (VF) from operational vegetation indexproducts of Indian geostationary satellite (INSAT 3A) sensor (CCD) was utilized in numerical weather pre-diction (NWP) model (e.g. WRF) to investigate its impact on short-range weather forecast over the controlrun. Results showed that the updated vegetation fraction from INSAT 3A CCD improved the low-level 24 htemperature (∼18%) and moisture (∼10%) forecast in comparison to control run. The 24 h rainfall forecastwas also improved (more than 5%) over central and southern India with the use of updated vegetation

ensible heat fluxescintillometer

fraction compared to control experiment. INSAT 3A VF based experiment also showed a net improve-ment of 27% in surface sensible heat fluxes from WRF in comparison to control experiment when bothwere compared with area-averaged measurements from Large Aperture Scintillometer (LAS). This trigg-ers the need of more and more use of realistic and updated land surface states through satellite remotesensing data as well as in situ micrometeorological measurements to improve the forecast quality, skilland consistency.

. Introduction

Weather and climate are the biggest risk factors that havempacts on agricultural planning and management. Agriculture ishe backbone of Indian Economy which influences internationalrade among SAARC (South Asian Association for Regional Cooper-tion) countries. About 65% of Indian population depends directlyn agriculture and it approximately accounts for 22% of GDP (grossomestic product). Any abnormalities in the short-range weatherorecast would seriously affect the growth and final yield of therops. The one of the foremost aims of weather forecasting is to ren-er timely advice to farmers on the actual and expected weather,nd its likely impact on the various day-to-day farming opera-ions. Especially for the nations (like India) largely depending on the

grarian economy, reliable high-resolution (up to 15 km) regionaleather forecasts are long-pending demands from crop forecasters,lanners, policy makers and for management of water resources,

∗ Corresponding author at: Atmospheric Sciences Division, Atmospheric andceanic Sciences Group, Earth, Ocean, Atmosphere, Planetary Sciences and Appli-ations Area, Space Applications Centre (ISRO), Ahmedabad 380015, India.el.: +91 79 26916052; fax: +91 79 26916075.

E-mail addresses: [email protected], [email protected] (P. Kumar).

168-1923/$ – see front matter © 2012 Elsevier B.V. All rights reserved.ttp://dx.doi.org/10.1016/j.agrformet.2012.08.009

© 2012 Elsevier B.V. All rights reserved.

agricultural disasters such as drought, flood, severity of certainpests and diseases (Strand, 2000). Land-surface processes throughSoil–Plant–Atmosphere Continuum (SPAC) are important driversfor weather and climate systems over the tropics and particularlyover the Indian sub-continent (Van der Tol et al., 2008). Realisticrepresentation of land-surface states over the Indian region wouldhelp accurate simulations of environmental processes at micro,meso, and regional climate scales (Matsui and Lakshmi, 2005).However, in order to achieve these potential benefits, it is necessaryto develop a strategy which will address and overcome the differentchallenges associated with the representation of the land-surfacestates over the Indian monsoon region.

The problem of determining a physically consistent and accu-rate snapshot of the atmosphere is the central theme of numericalweather prediction (NWP). In succeeding decades, with progress inboth computing power and optimization strategies, more sophis-ticated constraints and more diverse observations have beenincluded in NWP (Kumar et al., 2011). Generating an accurate ini-tial state is recognized as one of the biggest challenges in NWPmodel prediction of weather events. The rapid increasing computer

power has led to higher (1–15 km) resolution NWP models, whichare able to resolve mesoscale features and thus to give more pre-cise forecasts. These high resolution forecasts have the potentialto drastically change the forecasting procedure to determine the
Page 2: Impact of vegetation fraction from Indian geostationary satellite on short-range weather forecast

Fores

aHisnatflepa2eimgtoIvvl(spsri(I(aSiraaq

o(lie(ivVufi(dvfldcec

tscs(

P. Kumar et al. / Agricultural and

tmospheric structures in both space and time (James et al., 2009).igh-resolution observations are obligatory to provide accurate

nitial conditions for high-resolution forecast models. The neces-ity of high-resolution observations to initialize high-resolutionumerical model remains a significant challenge. High-resolution,ccurate representations of surface properties such as sea-surfaceemperature (SST), soil temperature and moisture content, grounduxes, and vegetation are necessary to better understand thearth-atmosphere interactions and improve the numerical weatherrediction (NWP). The high-resolution data source that is presentlyvailable from INSAT-3A geo-stationary satellites (Nigam et al.,011) but not used by and large, is a real-time estimate of veg-tation characteristics. A crude representation of land surfacenformation is used for routine generation of outputs from NWP

odel over Indian region (Das et al., 2008). The percentage ofreen fraction which is based on a 5-year average (or more) ofhe normalized difference vegetation index (NDVI) for each monthf the year at roughly 13 km2 horizontal resolution (Gutman andgnatov, 1998; James et al., 2009) within a model grid cell is used foregetation fraction. Horizontal and vertical distributions of plantegetation specified by the Green Vegetation Fraction (GVF) andeaf area index (LAI) are used to represent the model vegetationGutman and Ignatov, 1998). Evapotranspiration from vegetativeurfaces is an important variable that mainly impacts the trans-ort of moisture into the atmosphere during the warm period. Thepatial and temporal changes in vegetation are important to accu-ately predicting boundary layer structures as they have a largenfluence on partitioning the surface sensible and latent heat fluxesSegal et al., 1995; Crawford et al., 2001; Kurkowski et al., 2003).ncreased latent heat flux humidifies the planetary boundary layerPBL) and increases the moist static energy (MSE) of near-surfaceir and also increases the potential for precipitation (Eltahir, 1998;hukla and Mintz, 1982; Sud and Fennessy, 1982). There is emerg-ng evidence that vegetation cover change has a radiative impact onegional weather forecasts (James et al., 2009; Roy et al., 2003). As

result of the important role that vegetation plays in land-surfacend land–atmosphere interactions, it needs to be represented ade-uately in NWP models.

Vegetation fraction (VF) changes considerably within the periodf a week and season having considerable inter-annual variabilityJames et al., 2009; Kurkowski et al., 2003). The operational Noahand surface model (Chen and Dudhia, 2001; Ek et al., 2003) coupledn the Weather Research and Forecasting (WRF) model (Skamarockt al., 2008) holds the LAI fixed for all vegetation classes. James et al.2009) integrated the VF generated from polar orbiting satellite tomprove the short-range weather forecast in the case of severe con-ection and showed the significance of real time VF. However, theF from operational normalized difference vegetation index (NDVI)sing geostationary satellite data has certain advantages than thoserom polar orbiting satellite. (1) The possibility of cloud-free NDVIncreases from multiple acquisitions from geostationary satelliteFensholt et al., 2006); (2) the opportunity of getting practically noata gaps in the continental or country-scale NDVI increases from aalid single snapshot with data acquisition from geostationary plat-orm; Mosaicing from tiles to produce such large coverage generallyeads to inconsistent data gaps in case of polar orbiting satellitesue to less swath; (3) turn-around-time (TAT) from acquisition toountry-scale or continental-scale operational NDVI product gen-ration becomes less in case of geostationary satellite. This is veryrucial and pertinent to short-range weather forecasts.

Indian geostationary satellite (INSAT 3A) sensor (CCD) observeshe earth surface with continental (Asia) coverage at 1 km × 1 km

patial resolution and high temporal frequency (half-an-hour) atonstant view direction (Nigam et al., 2011). No other existing geo-tationary satellite missions in the world except INSAT 3A CCD1 km spatial resolution) of India and MSG SEVIRI (3 km spatial

t Meteorology 168 (2013) 82– 92 83

resolution) have payloads that take multiple observations per dayin multispectral optical bands. INSAT 3A is the only geostationarysatellite which scans Asia with multi-spectral bands. INSAT 3A waslaunched in 2003 with sub-satellite longitude at 93.5◦E. It coversone-fourth of the globe in a single snapshot mainly the Asia con-tinent (44.5◦E to 105.3◦E, 9.8◦S to 45.5◦N). It has CCD payload thatwas specifically designed to monitor vegetation and snow coverconditions over Asia regularly at spatial resolution of 1 km × 1 km.It has three unique optical bands in red (0.62–0.68 �m) also calledvisible, near infrared (0.77–0.86 �m), short-wave infrared (SWIR)(1.55–1.69 �m) wavelength regions.

The INSAT-3A provides a higher spatial and temporal resolu-tion data than the AVHRR-based climatology data currently usedin WRF. These high-resolution, near real-time INSAT-3A data arehypothesized to be more accurate to reflect variations in vege-tation characteristics. A limitation that the climatological datasetpresents is that the annual cycle of GVF always represents the samein models from one year to the next. In reality, the response ofvegetation to meteorological and climate conditions varies fromshort-range (up to 48 h) to long-range forecast (15–30 days) basedon anomalous weather. The objective of this study is to improvethe land surface states obtained through forcing of vegetation frac-tion which leads to more accurate short-term forecasts of regionalweather using WRF model. The vegetation fraction generated fromoperational NDVI product using multi-spectral observations fromINSAT-3A CCD camera were incorporated in WRF model and werecompared to control run prepared from USGS (United States Geo-logical Survey) vegetation fraction.

2. Methodology

2.1. Vegetation fraction from INSAT-3A CCD NDVI product

Green or active vegetation fraction closely approximates frac-tion of photosynthetically active radiation absorbed by greencanopy (Chen et al., 2008; Norato et al., 2006) denoted by frac-tional absorbed photosynthetically active radiation (FPAR), used assurrogate for vegetation fraction. It can be determined from theformulation (Eq. (1)) given by Sellers et al. (1994)

FPARi,t = (NDVIi,t − NDNIi,min) · (FPARi,max − FPARi,min)(NDVIi,max − NDNIi,min)

+ FPARi,min (1)

using pre-defined upper and lower limits of FPAR and NDVIon land use/land cover basis and NDVI time-series data. Theupper (FPARi,max) and lower (FPARi,min) limits of FPAR for 12plant functional types (PFTs) were determined from twenty years’NOAA Pathfinder AVHRR Land (PAL) FPAR time series data. The1 km × 1 km International Geosphere Biosphere Programme (IGBP)global land use land cover map was used to assign the abovelimits for 12 PFTs. Here, spatially variable upper (NDVIi,max) andlower (NDVIi,min) limits of NDVI at ith pixel was determined fromannual time series of temporally smoothed dekadal (ten-day)NDVI composites of INSAT 3A CCD. Surface reflectances in red(0.62–0.68 �m) and near-infrared (0.76–0.82 �m) bands of INSAT3A CCD camera were computed from TOA (top-of-atmosphere)band reflectances after atmospheric correction using SMAC (Sim-ple Model for Atmospheric Correction) code (Rahman and Dedieu,1994) and average atmosphere constituted through global clima-tology of 1◦ × 1◦ aerosol at 550 nm, precipitable water and ozone.Ten-day maximum value composites are available as regular prod-

uct at 1-km spatial resolution from IMDPS (INSAT MeteorologicalData Processing System). The cross-comparison of eight-day CCDNDVI with MODIS eight-day TERRA NDVI showed an overall neg-ative bias (−0.07) in CCD NDVI (Nigam et al., 2011) which are
Page 3: Impact of vegetation fraction from Indian geostationary satellite on short-range weather forecast

84 P. Kumar et al. / Agricultural and Forest Meteorology 168 (2013) 82– 92

F tion in2

aiitataCtfJga

faatr(s∼V

2

a(

ig. 1. Vegetation fraction data from (a) USGS and (b) INSAT-3A VF (vegetation frac009 used in WRF model.

ttributed to mainly differences in band intervals, BRDF correctionn MODIS, etc. The same bias is assumed for dekadal compos-tes also. The bias-removed dekadal NDVI composites were put toemporal smoothening filter using Harmonic analysis (for detailedlgorithm, see Menenti et al., 1993; Roerink et al., 2000) to removehe spikes in the NDVI time series due to consistent clouds, highererosol load than the average. The reconstructed smoothed dekadalCD NDVI at 0700 GMT for 2009 were then used in the above equa-ion as NDVIi,t to determine dekadal FPAR vis-a-vis green vegetationraction. Monthly averages were obtained from dekadal values forune and July 2009. The 1-km averages were resampled to 0.15◦

rids and the gridded vegetation fraction was used to WRF models input.

Before using the INSAT-3A CCD camera generated vegetationraction (INSAT-3A VF) in WRF model, we have compared the USGSnd INSAT-3A VF over Indian region. The considerable changesre observed between two VF (Fig. 1a–d) over Indian regions inwo consecutive months of June and July 2009. The north-westernegion of India is represented with very less VF (∼5%) in USGSFig. 1a and Fig. 1c), although in INSAT-3A VF (Fig. 1b and d) data,ame region was represented with low VF but ranging between15% and 25% for both the months. Similar trend in differences inF were also evident in western as well as central India.

.2. WRF model

The forecast model used in this study is the Weather Researchnd Forecasting (WRF) (Skamarock et al., 2008) Model version 3.1http://www.wrf-model.org). This mesoscale numerical model is

percent) data during June 2009 and (c) USGS and (d) INSAT-3A VF data during July

designed to serve both operational forecasting and atmosphericresearch needs. It is a limited-area, non-hydrostatic primitiveequation model with multiple options for various physical parame-terization schemes. The present version of the WRF model employsArakawa C-grid staggering for the horizontal grid and a fully com-pressible system of equations. The terrain following hydrostaticpressure with vertical grid stretching was followed for the verticalgrid. The time-split integration uses the third-order Runge–Kuttascheme with a smaller time step for acoustic and gravity wavemodes. Physics options used in this study include the Kain–Fritsch(Kain and Fritsch, 1990, 1993) cumulus parameterization schemeand the WRF Single-Moment 6-class graupel (WSM6) microphysicsscheme. The planetary boundary layer is parameterized using theYonsei University (YSU) planetary boundary layer scheme (Hongand Dudhia, 2003; Hong and Pan, 1996) and, for the soil model,the multi-layer Noah land surface model (LSM) model is used. Thelong-wave radiation scheme is based on the rapid radiative trans-fer model (RRTM), and the short-wave radiation scheme is based onDudhia (1989). The role of vegetation fraction in WRF is to deter-mine three components of evapo-transpiration (soil evaporation,wet canopy evaporation, canopy transpiration) and surface energybalance in Noah land surface model. This also has influence ondetermining the displacement height and roughness length in thesurface layer above land surface to determine exchange coefficientswithin Planetary Boundary Layer (PBL) (Chen and Dudhia, 2001).

This, in turn, modulates surface heat fluxes (sensible and latentheat). Generally, seasonal albedo is used for different land use/landcover categories. But model parameterization makes it dynamicwithin a season from real-time soil moisture and vegetation
Page 4: Impact of vegetation fraction from Indian geostationary satellite on short-range weather forecast

P. Kumar et al. / Agricultural and Forest Meteorology 168 (2013) 82– 92 85

F durinv

ffttoeaafttbte

2

6wleeAglc

ig. 2. Scatter plot of INSAT-3A and CNT (before assimilation) vegetation fractionegetation fraction during (c) June and (d) July 2009.

raction. Albedo governs net radiation at surface and thereby sur-ace energy balance. Yang et al. (1999) used two land surface modelso investigate the sensitivity of the ground heat flux to the vegeta-ion cover fraction. They observed that reducing the areal coveragef vegetation results in an increase in day time soil heat flux. Strackt al. (2008) noted that changes in land surface properties such aslbedo, roughness length, stomatal resistance, and leaf area indexlter the surface energy balance, leading to differences in near sur-ace temperatures. Noah LSM performance in the Eta model effectshe physical parameterizations of radiation and clouds, which affecthe amount of available energy at the surface, and stability ofoundary layer and surface layer processes, which affect surfaceurbulent heat fluxes and ultimately the surface energy budget (Ekt al., 2003).

.3. Design of simulation experiment

All experiments were conducted in single domain (Lon:1.9–98.7◦E, Lat: 5.5–38.5◦N) consisting of 269 × 269 grid pointsith 15 km horizontal grid resolution. The model had 36 vertical

evels with the top of the model atmosphere located at 10 hPa. Twoxperiments (CNT: with USGS VF and EXP: with INSAT-3A gen-rated VF) were performed daily at 0000 UTC during 12 June–07

ugust 2009. National Centers for Environmental Prediction (NCEP)lobal data assimilation system (GDAS) analysis with 1◦ × 1◦ reso-ution were used directly as first guess (FG) and lateral boundaryondition for all the experiments. A 48 h forecast was made daily

g (a) June and (b) July 2009 and scatter of INSAT-3A and EXP (after assimilation)

from 0000 UTC during this period (57 cases) with CNT and EXPinitial conditions.

2.4. Assimilation approach

The objective analysis technique is used here for the assimila-tion of INSAT-3A vegetation fraction to improve first-guess griddedanalysis (USGS VF) by incorporating additional observational infor-mation (INSAT-3A VF). The Cressman (Cressman, 1959) objectiveanalysis scheme is used here for assimilation to improve the firstguess. The Cressman scheme allows for a circular radius of influ-ence is implemented here and it is also known as the successivecorrection method. The model state is set equal to the USGS valuesin the vicinity of available observations. If we represent a back-ground state (which is USGS VF) by xb and a set of n observations ofthe same parameter by y(i), where i = 1, 2, . . ., n, then the analyzedfield of the parameter xa using Cressman technique at each modelgrid point j can be represented as,

xa(j) = xb(j) +∑n

i=1w(i, j){y(i) − xb(i)}∑ni=1w(i, j)

where, w(i, j) = max

{0,

R2 − d2i,j

}.

R2 + d2i,j

Here di,j is the distance between the model grid point j and obser-vation point i and xb(i) is the background value interpolated to the

Page 5: Impact of vegetation fraction from Indian geostationary satellite on short-range weather forecast

86 P. Kumar et al. / Agricultural and Forest Meteorology 168 (2013) 82– 92

F nctioi , 9, 15F

o2i

tcptVt

2

2

(u

ig. 3. Temporal profiles of dekadal (10-day) vegetation fraction over major plant fun the third dekad in those months ending with 31 days corresponding to number 3ebruary corresponding to number 6 in X-axis.

bservation point. R is the radius of influence which is defined here5 km, and w(i, j) is a measure of weight which decreases with

ncreasing di,j, becoming zero if di,j = R.Scatter distribution (Fig. 2) of USGS and INSAT-3A VF showed

hat the less to very less VF value was observed in USGS VF inomparison to INSAT-3A VF over Indian landmass during the studyeriod. Very less correlation was observed in two sources of vege-ation fraction from USGS and INSAT-3A. After assimilation of CCDF using Cressman method, analyzed VF was found to be closer to

he actual VF. This showed the successful assimilation of VF.

.5. Validation strategy

.5.1. Weather variablesHere, we have considered the root-mean-square difference

RMSD) in the forecast fields as a standard measurement for eval-ating the performance of WRF model forecast. The NCEP GDAS

nal types in India. Each month has three dekads. However, eleven days are counted, 21, 24, 30, 36 in X-axis but eight or nine days (for leap year) in the third dekad of

analysis was used to validate the synoptic features of average(based on all sample days) of model simulated parameters. NCEPGDAS analysis (1◦ × 1◦) has been used for computing the RMSDin the CNT and EXP experiments predicted temperature, humid-ity, etc. The NCEP GDAS analysis has been widely used because itis easily available. The strength of NCEP analysis products is thatthey provide gridded data with global coverage at high temporal(6-hourly) resolution. Goswami and Sengupta (2003) assessed thequality of NCEP reanalysis with surface observations over Indianregion. Roads and Betts (1999) compared the NCEP and ECMWFreanalysis over the Mississippi River basin. They found that boththe reanalysis produced similar seasonal energy components. Theinter-annual variations from NCEP and ECMWF were also compa-rable.

Precipitation was validated against the rainfall product fromTRMM (Tropical Rainfall Mapping Mission) satellite having 25 kmspatial resolution. Limited and sparse network of rain gauge

Page 6: Impact of vegetation fraction from Indian geostationary satellite on short-range weather forecast

Fores

oa2id2seSnANl(CtThtsimua

obTm

ı

wiofıI

2

at21mwdrueraMuvcamsm

P. Kumar et al. / Agricultural and

bservations over Indian region are unable to provide a reli-ble spatial representation of precipitation (Gruber and Levizzani,008). Remote sensing techniques are advantageous for monitor-

ng rainfall over large regions. They provide more homogeneousata quality compared with ground observations (Schulz et al.,009). The rainfall data used in this study (TRMM 3B42) con-ists of: (1) TRMM High Quality combined microwave precipitationstimates of TRMM Combined Instrument (TCI), TMI, Specialensor Microwave/Imager (SSM/I), Advanced Microwave Scan-ing Radiometer for EOS (AMSR-E) onboard EOS Aqua, and thedvanced Microwave Sounding Unit B (AMSU-B) onboard theational Oceanic and Atmospheric Administration (NOAA) satel-

ites; (2) The TRMM Variable Rain Rate precipitation algorithmVAR) microwave-calibrated Infrared (IR); (3) Global Precipitationlimatology Centre (GPCC) or Climate Analysis and Monitoring Sys-em (CAMS) gauge analyses (Huffman et al., 2007). The detailedRMM 3B42 estimation procedure is described at the websitettp://trmm.gsfc.nasa.gov/3b42.html. The TRMM 3B42 precipita-ion output comprises 0.25◦ × 0.25◦ grid cells for every 3 h, withpatial extent covering a global belt (−180◦W to 180◦E) extend-ng from 50◦S to 50◦N latitude. This is the main source of remotely

easured real-time rainfall data over tropics. Mishra et al. (2010)sed the TRMM 3B42 rainfall as benchmark to validate the rainfalllgorithm over Indian region.

Bilinear interpolation is used to make both model forecast andbservations at the same scale. We have analyzed the spatial distri-ution of RMSD in the model predicted temperatures, and moisture.he improvement parameter (ı) was used to quantify the improve-ent in the forecast as given below.

=

[(1/N)

∑Ni=1(Oi − FC

i)2]1/2

−[

(1/N)∑N

i=1(Oi − FEi

)2]1/2

[(1/N)

∑Ni=1(Oi − FC

i)2]1/2

× 100

(2)

here FC is the CNT forecast produced using USGS VF data and FE

s the EXP forecast produced with the INSAT-3A VF and O is thebservation (NCEP analysis/TRMM) and N is the total number oforecasts, which are 57 in this case. A positive (negative) value of

indicates improvement (degradation) in the EXP forecast due toNSAT-3A VF data as compared to CNT.

.5.2. Land surface fluxes: sensible heat fluxArea averaged sensible heat flux (H) over a footprint of 1.5 km2

gricultural landscape representing was recorded using Large Aper-ure Scintillometer (LAS) (Kipp and Zonen LAS 150) during October006 to November 2010 maintaining a clear diagonal path length of.5 km between mounted transmitter and receiver. The measure-ents were made at Nawagam (22◦46′45′′N, 72◦34′31′′E) situatedithin large (approx. 5 km × 5 km) agricultural region of Khedaistrict, Gujarat state of North-West India representing irrigatedice–wheat system in semi-arid climate. The LAS system meas-res the amount of scintillations at 880 nm wavelength that arexpressed as the optical turbulence structure parameter (C2

n ) of theefractive index of air. The ‘H’ was derived from C2

n measurementsnd ancillary data from the associated weather assembly using theonin–Obukhov Similarity Theory (MOST). The details of LAS set-

p over a single site in Indian semi-arid climate and its use foralidation of satellite based energy budget have already been dis-ussed by Bhattacharya et al. (2011). These data were recorded at

n interval of 10 min. In present study, hourly averaged ‘H’ from LASeasurements were used to compare WRF simulated land surface

ensible heat flux at 6 h after 00 GMT during June, July and Augustonths of the study year (2009).

t Meteorology 168 (2013) 82– 92 87

3. Results and discussion

3.1. Temporal profiles of VF over major PFTs

The dekadal (ten-day) temporal profiles of annual variationof VF of major plant functional types (PFTs) from INSAT 3A CCDNDVI product are shown in Fig. 3 for croplands, grassland andforests. Clearly two crop growth cycles were evident because ofrice–rice and wheat–rice cropping systems prevalent over easternand northern India, respectively.

The peak VFs were 0.85 at around band number 10 (2nd weekof April) and 0.75 at band number 30 (fourth week of October) forrice–rice system. These were 0.83 at band number 6 (2nd week ofFebruary) and 0.72 at band number 25 (2nd week of September)for wheat–rice system. In grassland of Thar Desert, the VF attainedits minima (<0.2) around band number 15 (2nd week of May) andmaxima (0.6) at band number 25 (3rd week of September) afterrecession of south-west monsoon. The VF of evergreen forest overWestern Ghat remained high throughout the year, between 0.55and 0.9 with peak at band number 30 (fourth week of October)and with a little dip during April. But the dry deciduous forest overcentral India showed distinct annual cycle with the maxima (0.83)at band number 27 (2nd week of October) and minima (0.28) atband number 15 (fourth week of May). These profiles distinguishedthe wide variability of growth behavior of major vegetation typesover India. It is expected that this would trigger the continuousresponse-feedback mechanism between land and atmosphere.

3.2. Regional evaluation of forecast weather

3.2.1. Air temperature and moistureThe model forecasted (CNT and EXP) temperature was vali-

dated with NCEP analyzed temperature for the period of study (12June 2009 to 31 July 2009). The geographical distribution of per-centage improvement parameter (ı) of surface temperature (2 m)was given in Fig. 4. We have compared the 12 hourly predictedsurface (2 m) temperatures from CNT and EXP experiments withthe NCEP analyzed 2 m surface temperature. As compared to CNTrun, the INSAT-3A VF based experiment (EXP) positively (∼18%)impacted (positive area is more) over the Indian landmass in the12 h prediction (Fig. 4) of surface temperature whereas few pock-ets of degradation are observed in Southern India and Eastern India.Degradation was also observed over the hilly terrains and Tibetanplateau. INSAT 3A CCD NDVI product was not corrected for surfaceBRDF (Bi-directional Distribution Function). The effect could havebeen substantial at higher altitudes in the hilly terrains of India aswell as Tibetan plateaus. This seemed to have led to substantiallyhigher VF.

Similarly, the geographical distribution of improvement param-eter (ı) up to 48 h forecast showed the similar results as obtainedin 12 h prediction with slightly less positive percentage improve-ment mainly over Central India. In general, spatial distributionof improvement parameter (ı) of surface temperature showedthat the INSAT-3A generated VF have positive impact in the sur-face temperature (2 m) forecast over Central India and most ofthe Indian regions. Fig. 5 showed the geographical distribution ofimprovement parameter (ı) for 12 h predicted lower level (850 hPa)temperature. The RMSD in lower level temperature forecast wasincreased with the forecast length from 12 h to 48 h forecast. Incomparison to CNT run, the modify VF based experiment (EXP)positively impacted (positive area is more) over Central India inthe 12 h (Fig. 5) prediction of lower level temperature whereas few

pockets of degradation were also observed in the Indo-Gangeticregion and southern India.

Similar spatial distribution of improvement parameter (ı) oftemperature were observed in other higher levels with less

Page 7: Impact of vegetation fraction from Indian geostationary satellite on short-range weather forecast

88 P. Kumar et al. / Agricultural and Forest Meteorology 168 (2013) 82– 92

Fa

plgvmvfIs

ai(t

FR

ig. 4. Spatial distribution of improvement parameter (ı) for 12 h temperature (2 mir) forecast.

ercentage improvement. It is interested to note that 48 h lowerevel temperature forecast was improved over the Bay of Ben-al. In the tropical region, planetary boundary layer (PBL) heightaries approximately from 1 to 2 km which corresponds to approxi-ately 850 hPa. Any changes in the land surface characteristics (e.g.

egetation fraction, soil moisture, etc.) will have impact on the sur-ace layer fluxes as well as on atmospheric parameters in the PBL.mprovement in both the near-surface as well as at 850 hPa levelhowed the consistency in the forecast impact.

Temporal distribution of domain average surface (2 m) temper-ture from NCEP analysis, CNT and EXP experiments are shown

n Fig. 6a during 11 June to 01 August 2009. Less temperatureFig. 6a) was observed in NCEP analysis in comparison to bothhe simulation experiments. CNT experiment overestimated the

ig. 6. (a) Time series of domain average surface (2 m) air temperature (◦C) from CNT anMSD (◦C) in CNT and EXP for 24 h forecast.

Fig. 5. Spatial distribution of improvement parameter (ı) for 12 h temperature(850 hPa) forecast.

surface temperature around 2–3 ◦C and this overestimation wasreduced (∼1.2 ◦C) with the use of INSAT-3A vegetation fraction.This improvement was more prominent during the month of July.Vertical profile of domain average temperature (Fig. 6b) showedthat the positive improvement was observed up to 700 hPa andabove this, very less to no or negative improvement was seen afterthe assimilation of INSAT-3A vegetation fraction (based on 50 dayssimulation).

The predicted specific humidity (g kg−1) from the WRF exper-iments was verified with the NCEP analyzed specific humidity.Similar to temperature validation, we have compared the 12 hourly

predicted surface (2 m) and lower (850 hPa) level specific humid-ity from CNT and EXP experiments with the NCEP analysis. Thegeographical distribution of percentage improvement (ı) of surface

d EXP with NCEP analysis and (b) vertical profile of domain average temperature

Page 8: Impact of vegetation fraction from Indian geostationary satellite on short-range weather forecast

P. Kumar et al. / Agricultural and Fores

Fig. 7. Spatial distribution of improvement parameter (ı) for 12 h specific humidity(

sot(owsIoi(ic

Fh

2 m air) forecast.

pecific humidity was given in Fig. 7. Less than 1 g kg−1 RMSD wasbserved over the Western Ghat and Western India in 12 h forecastime. In comparison to control run, the INSAT-3A VF data positively∼10%) impacted over central India in the 12 h (Fig. 7) predictionf surface specific humidity whereas some pockets of degradationere also observed in northern and southern parts of India. Overall,

patial distribution of improvement parameter (ı) showed that theNSAT-3A VF have very less to no impact in the 12 h (Fig. 7) forecastf surface specific humidity over India. Similar features, as shown

n Fig. 7 are observed in lower level (850 hPa) specific humidityfig. not shown) forecast. Interestingly, some pockets of positivempacts were observed in central region which illustrated the majorhanges in VF data in this region. In general, spatial distribution of

ig. 8. (a) Time series of domain average surface (2 m) air relative humidity (%) from CNumidity RMSD (◦C) in CNT and EXP for 24 h forecast.

t Meteorology 168 (2013) 82– 92 89

improvement parameter (ı) of surface specific humidity showedthat the INSAT generated VF have very less to no impact in the 12 hpredicted surface specific humidity.

Similar to temperature, temporal distribution of domain aver-age surface (2 m) relative humidity from NCEP analysis, CNT andEXP experiments are shown in Fig. 8a for the same period. USGSbased experiment produced the less domain average surface (2 m)relative humidity in comparison to NCEP analyzed relative humid-ity (2 m). Similar to temperature improvement, surface relativehumidity is also improved after the utilization of INSAT-3A veg-etation fraction and this improvement is ∼10% in surface moistureduring the month of July.

Similar to temperature error profile, vertical profile of domainaverage relative humidity (Fig. 8b) showed that the positiveimprovement was observed up to 650 hPa and above this, very lessto no or negative improvement was seen with the use of INSAT-3Avegetation fraction.

3.2.2. RainfallTropical Rainfall Measuring Mission (TRMM) 3B42 V6 product

(Haddad et al., 1997; Adler et al., 2000) was used for the verificationof the model predicted rainfall over land as well as Oceanic region.The TRMM rainfall used for validation is at 3-h temporal resolutionand 0.25◦ × 0.25◦ spatial resolution. The model-predicted rainfallwas re-sampled to 25 km (using bilinear interpolation) in order tocompare it with the TRMM rainfall. TRMM 3B42 algorithm’s 3-hrain rate was converted to accumulated rainfall assuming constantrain rate over 3 h. The monthly (July-2009) total rainfall from TRMMshowed that the maximum monsoon rainfall was observed over thewest coast of peninsular India and eastern region. For the quanti-tative assessment of improvement (degradation) due to the use ofINSAT-3A VF data as compared to CNT, we have computed the spa-tial distribution of improvement parameter. It was clear from Fig. 9that the rainfall prediction was improved (relative to the controlsimulation) 5–10% by incorporation of INSAT-3A VF data.

Use of INSAT-3A VF data improved the 24-h rainfall predic-tion over the central India and southern India coast. This impactwas mainly due to the improvement in the lower level tempera-ture fields. Over southern India, central India and north-east India,compared to the control simulation, the rainfall was improved inEXP experiment. Some pockets of degradation were also seen inthe rainfall prediction in central India. Temporal distribution of

domain average percentage improvement parameter in 24 h rain-fall forecast is shown in Fig. 10. It showed that for most of thesimulation experiments, INSAT 3A based vegetation fraction hadpositive impact over the control experiments. There was systematic

T and EXP with NCEP analysis and (b) vertical profile of domain average relative

Page 9: Impact of vegetation fraction from Indian geostationary satellite on short-range weather forecast

90 P. Kumar et al. / Agricultural and Forest Meteorology 168 (2013) 82– 92

Fig. 9. Spatial distribution of improvement parameter (ı) for 24 h rainfall (mm) forecast.

Fr

icptAwoiotUsrp

3

b

Fig. 11. Comparison of WRF simulated sensible heat fluxes (H) using USGS andINSAT 3A CCD vegetation fraction and area averaged measurements of sensible heat

ig. 10. Temporal distribution of domain average percent improvement in 24 hainfall forecast.

mprovement from 16 June to 18 July 2009 except one day inase of domain average 24 h rainfall forecast. Large changes inercent improvement were noticed during 19–25 July. After that,here was less consistency in the improvement pattern up to 6ugust. Improvement consistency in rainfall forecast up to 25 Julyould significantly help farmers in deciding to set time of sowing

perations, transplanting and spraying schedules for crops dur-ng summer-monsoon season through improvement in the qualityf agro-advisories. The pockets of central India showing degrada-ion typically represent rainfed agriculture in semi-arid climate.pdation of surface and sub-surface soil moisture instead of using

oil moisture climatology would be the wise choice for properepresentation of land surface albedo and surface energy balanceartitioning in addition to vegetation fraction.

.3. Impact on land surface sensible heat fluxes

Sensible heat flux, the most critical component of surface energyalance, determines the partitioned net radiation to surface leaving

flux using Large Aperture Scintillometer (LAS) over a semi-arid agricultural land-scape in North-West India during June, July and August 2009. Slanting straight linesare trend lines.

latent heat fluxes through convection. The simulated output of sen-sible latent heat fluxes (H) at 6 h forecast from 0000 UTC frombuilt-in Noah land surface scheme in WRF was compared with coin-cident area-averaged sensible heat flux measurements from LargeAperture Scintillometer (LAS). The comparison was made for threemonths (June, July, and August 2009) during south-west monsoonperiod over a semi-arid landscape in Gujarat state of North-WestIndia. The plot of comparison and error statistics over all the threemonths is presented in Fig. 11 and Table 1, respectively. In general,the temporal variation of simulated ‘H’ is in line with LAS mea-sured ‘H’. However, there was always large overestimation of ‘H’

from WRF as compared to measurements (Fig. 11). The improve-ment in simulated ‘H’ from WRF was noticed with the impact ofINSAT 3A CCD vegetation fraction as compared to that of USGS
Page 10: Impact of vegetation fraction from Indian geostationary satellite on short-range weather forecast

P. Kumar et al. / Agricultural and Forest Meteorology 168 (2013) 82– 92 91

Table 1Error statistics of comparison of WRF forecast sensible heat flux at 6 h forecast from 0000 UTC with area averaged measurements from Large Aperture Scintillometer (LAS).

Months in 2009 WRF (USGS) WRF (CCD) % RMSE of measured meanfor WRF (USGS)

% RMSE of measured meanfor WRF (CCD)

MAE (W m−2) RMSE (W m−2) MAE (W m−2) RMSE (W m−2)

June 195 201 149 155 107 83July 121 136 88 106 216 169August 141 154 132 150 421 412

Mean 152.3 163.7 123 137 248 221

WRF (USGS): WRF CNT (control) run with USGS vegetation fraction.WRF (CCD): WRF EXP (experimental) run with INSAT 3A CCD vegetation fraction.

RMSE (root mean square error) =

√∑i[(Pi )−(Oi )]

2

N .

MAE (mean absolute error) =∑

iABS[(Pi )−(Oi )]

N .Where Pi = WRF simulated ‘H’ at ith case.Oi = measured ‘H’ from LAS at ith case.N

(tailtfleeir(8tow2fthR1tJaoaorfdwhtom3arscctswt

= number of paired datasets.

control experiment). Trend lines in Fig. 11 showed decreasingrend from June to August both in case of WRF simulated ‘H’nd measured ‘H’. It is also true from measurements that rainfallncreases from June to August in the north-western India due toate arrival of south-west monsoon. This is coupled with increasingrend in cloud cover and decreasing incident short-wave radiativeuxes but increasing surface wetness. These could lead to gen-ral decreasing trend in sensible heat fluxes. The mean absoluterror (MAE) and root mean square error (RMSE) of WRF ‘H’ var-ed from 121 to 195 W m−2 and from 136 to 201 W m−2 (Table 1),espectively during June to August 2009 for control experimentUSGS vegetation fraction). The errors were substantially less (MAE:8–149 W m−2, RMSE: 106–155 W m−2) with INSAT 3A CCD vege-ation fraction. The mean RMSE of ‘H’ at 6 h forecast from 0000 UTCver three months from the impact of INSAT vegetation fractionas found to be closer (137 W m−2) to the reported (Hong et al.,

009) range of mean RMSE of daily average ‘H’ (38–100 W m−2)rom WRF with respect to IHOP (International H2O project) fluxower observation sites spread over eastern Kansas and Okla-oma Panhandle in the US. It is to be noted that the per centMSE with respect to measured mean gradually increased from07 to 421% for WRF with USGS vegetation fraction and from 83o 412% for WRF with INSAT 3A CCD vegetation fraction duringune to August. This indicated that surface heat fluxes gradu-lly became less sensitive to vegetation fraction over the passagef time or with increasing vegetation growth. Vegetation gener-lly attains its peak growth by August utilizing monsoon rainfallr through availability of irrigation water when surface wetnessemains at its peak. This leads to a condition of near invariant sur-ace albedo. Then, the flux partitioning seems to be increasinglyependent on accuracy estimates of incident short-wave and long-ave radiation fluxes within Noah energy balance scheme underighly dynamic cloud cover conditions. It is interesting to notehat decrease in improvement of sensible heat flux with passagef time is exactly in line with loss of consistency in the improve-ent of rainfall forecast from July end to early August (refer Section

.2.2). The scale mismatch between WRF grid resolution (25 km2)nd LAS measurements (∼1.5 km2), lack of assimilation of othereal-time land surface input variables such as surface and sub-urface soil moisture, leaf area index (LAI), lack of revision ofoefficients embedded in the parameterization of land surface pro-esses (LSP) and relevant to sub-tropics are some of the lacunae

hat could lead to such a large overestimation in simulated sen-ible heat fluxes from WRF. However, in the present study, thereas net improvement in WRF sensible heat flux estimates to the

une of 27% in terms of RMSE with updated vegetation fraction

from INSAT as compared to USGS climatology when area aver-aged sensible heat flux measurements from scintillometer overa semi-arid agricultural landscape were used as validation refer-ence.

4. Summary and conclusions

Comparison of numerical forecasts of the WRF model weremade, using both climatic (USGS) and satellite derived updatedfractional vegetation cover from the INSAT-3A CCD data, in orderto examine the potential importance of real-time or updated veg-etation fraction information to 48 h forecasts of 2-m temperatures,moisture as well as sensible heat flux. Quality and its consistency offorecasts from NWP models is extremely valuable in order to settleclaims under crop insurance and reinsurance schemes offered byprivate and government agencies based on certain weather basedderivatives. The present study clearly demonstrated the signifi-cant improvement (15–20%) in the quality of regional short-rangeweather forecasts of temperature and moisture, and 5–10% forrainfall with the use of updated near real-time vegetation frac-tion from Indian geostationary satellite sensor over climatology.Overall, there was net improvement in WRF sensible heat flux esti-mates to the tune of 27% in terms of RMSE with updated vegetationfraction from INSAT as compared to USGS climatology when LASmeasurements of sensible heat flux over a semi-arid agriculturallandscape were used as validation reference. Temporal distribu-tion of domain average per cent improvement in rainfall forecastshowed positive impact of INSAT 3A VF in short-range weatherforecast. However, loss of consistency in rainfall improvementtowards July end to early August must have been triggered throughdecay in improvement in sensible heat fluxes thereby showing lessimpact of vegetation fraction on rainfall forecast with the passageof time. However, large overestimation in sensible heat flux wasnoticed in both the cases that call for further evaluation throughexperiments with other real-time land surface inputs to WRF. Sys-tematic increase in per cent RMSE of ‘H’ with passage of timeemphasized the need to evaluate and revise the radiation balanceparameterization scheme in cloudy skies relevant to sub-tropics.While the uncertainties can be minimized through assimilationof more and more use of other satellite remote sensing basedland surface inputs such as albedo, leaf area index (LAI) and soilmoisture, the regional-specific biases of other inputs such as soil

temperatures can be removed from the network of in situ microm-eteorological measurements. Future work would attempt to furtherreduce the uncertainties in regional short-range weather forecastsfrom WRF to improve the forecast quality.
Page 11: Impact of vegetation fraction from Indian geostationary satellite on short-range weather forecast

9 Fores

A

SeCaEta

R

A

B

C

C

C

C

D

D

E

E

F

G

G

G

H

H

H

H

H

2 P. Kumar et al. / Agricultural and

cknowledgments

The authors would like to thank Mr. A. S. Kiran Kumar, Director,AC and Dr J. S. Parihar, Deputy Director, EPSA/SAC for constantncouragement and guidance. Authors are thankful to Nationalenter for Atmospheric Research (NCAR) for WRF model and thenalysis. The global analyzed data provided by National Centers fornvironmental Prediction (NCEP) are acknowledged with sincerehanks. The authors are grateful to comments and suggestions bynonymous reviewer that helped improve the quality of this paper.

eferences

dler, R.F., Bolun, D.T., Curtis, S., Nelkin, E.J., 2000. Tropical rainfall distributionsdetermined using TRMM combined with other satellite and rain gauge informa-tion. J. Appl. Meteor. 39, 2007–2023.

hattacharya, B.K., Mallick, K., Nigam, R., Dakore, K., Shekh, A.M., 2011. Effi-ciency based wheat yield prediction in a semi-arid climate using surfaceenergy budgeting with satellite observations. Agric. Forest Meteorol. 151,1394–1408.

hen, F., Dudhia, J., 2001. Coupling an advanced land-surface–hydrology model withthe Penn State–NCAR MM5 modeling system. Part I. Model description andimplementation. Mon. Weather Rev. 129, 569–585.

hen, F., Lemone, P., Trier, S., Manning, K., Barlage, M., 2008. Impacts of Land/PBLinteraction on short-term prediction of precipitation for the IHOP 2002. 10–15June episode (http://www.mmm.ucar.edu.step/files/FEI STEP Mar08.ppt).

rawford, T.M., Stensrud, D.J., Mora, F., Merchant, J.W., Wetzel, P.J., 2001. Value ofincorporating satellite-derived land cover data in MM5/PLACE for simulatingsurface temperatures. J. Hydrometeorol. 2, 453–468.

ressman, G.P., 1959. An operational objective analysis system. Mon. Weather Rev.87, 367–374.

as, S., Ashrit, R., Iyengar, G.R., Mohandas, S., Gupta, M.D., George, J.P., Rajagopal, E.N.,Dutta, S.K., 2008. Skills of different mesoscale models over Indian region duringmonsoon season: forecast errors. J. Earth Syst. Sci. 117 (October (5)), 603–620.

udhia, J., 1989. Numerical study of convection observed during the winter mon-soon experiment using a mesoscale two-dimensional model. J. Atmos. Sci. 46,3077–3107.

k, M.B., Mitchel, K.E., Lin, Y., Rogers, E., Grunmann, P., Koren, V., Gayno,G., Tarpley, J.D., 2003. Implementation of Noah land surface modeladvances in the National Centers for Environmental Prediction oper-ational mesoscale Eta model. J. Geophys. Res. 108 (D22), 8851,http://dx.doi.org/10.1029/2002JD003296.

ltahir, E.A.B., 1998. A soil moisture rainfall feedback mechanism. 1. Theory andobservations. Water Resour. Res. 34, 765–776.

ensholt, R., Sandholt, I., Stisen, S., Tucker, C., 2006. Analysing NDVI for the Africancontinent using the geostationary meteosat second generation SEVRI sensor.Remote Sens. Environ. 101, 212–229.

oswami, B.N., Sengupta, D., 2003. A note on the deficiency of NCEP/NCAR reanaly-sis surface winds over the equatorial Indian Ocean. J. Geophys. Res. 108, 3124,http://dx.doi.org/10.1029/2002JC001497.

ruber, A., Levizzani, V., 2008. Assessment of Global Precipitation Products, A projectof the World Climate Research Programme Global Energy and Water CycleExperiment (GEWEX) radiation panel. WCRP Report No. 128, WMO/TD No. 1430,50 pp.

utman, G., Ignatov, A., 1998. The derivation of the green vegetation fractionfrom NOAA/AVHRR data for use in numerical weather prediction models. Int.J. Remote Sens. 19, 1533–1543.

addad, Z.S., Smith, E.A., Kummerow, C.D., Iguchi, T., Farrar, M.R., Durden, S.L.,Alves, M., Olson, W.S., 1997. The TRMM day-1 radar/radiometer combined rain-profiling algorithm. J. Meteor. Soc. Jpn. 75, 799–809.

ong, S.Y., Dudhia, J., 2003. Testing of a new nonlocal boundary layer vertical diffu-sion scheme in Numerical weather prediction applications. In: 20th Conferenceon Weather Analysis and Forecasting/16th Conference on Numerical WeatherPrediction, Seattle, WA.

ong, S., Lakshmi, V., Small, E.E., Chen, F., Tewari, M., Manning, K.W., 2009. Effects ofvegetation and soil moisture on the simulated land surface processes from the

coupled WRF/Noah model. J. Geophys. Res. 114 (D18118), 1–13.

ong, S.Y., Pan, H.L., 1996. Nonlocal boundary layer vertical diffusion in a mediumrange forecast model. Mon. Weather Rev. 124, 2322–2339.

uffman, G.J., Adler, R.F., Bolvin, D.T., Gu, G., Nelkin, E.J., Bowman, K.P., Stocker,E.F., Wolff, D.B., 2007. The TRMM Multi satellite Precipitation Analysis (TMPA):

t Meteorology 168 (2013) 82– 92

quasiglobal, multiyear, combined-sensor precipitation estimates at fine scales.J. Hydrometeorol. 8, 38–55.

James, K.A., Stensrud, D.J., Yossouf, N., 2009. Value of real-time vegetation fractionto forecasts of severe convection in high-resolution models. Weather Forecast.24, 187–210.

Kain, J.S., Fritsch, J.M., 1990. A one-dimensional entraining/detraining plume modeland its application in convective parameterization. J. Atmos. Sci. 47, 2784–2802.

Kain, J.S., Fritsch, J.M., 1993. Convective parameterization for mesoscale models: theKain–Fritcsh scheme. In: Emanuel, K.A., Raymond, D.J. (Eds.), The Representationof Cumulus Convection in Numerical Models. Amer. Meteor. Soc., 246 pp.

Kumar, P., Singh, R., Joshi, P.C., Pal, P.K., 2011. Impact of additional surface observa-tion network on short range weather forecast during summer monsoon 2008over Indian subcontinent. JESS 120, 1–12.

Kurkowski, N.P., Stensrud, D.J., Baldwin, M.E., 2003. Assessment of implemen-ting satellite-derived land cover data in the Eta model. Weather Forecast. 18,404–416.

Matsui, T., Lakshmi, V., 2005. The effects of satellite-derived vegetation cover vari-ability on simulated land–atmosphere interactions in the NAMS. J. Climate 18,21–40.

Menenti, M., Azzali, S., Verhoef, W., Van Swol, R., 1993. Mapping agro-ecologicalzones and time lag in vegetation growth by means of Fourier analysis of timeseries of NDVI images. Adv. Space Res. 5, 233–237.

Mishra, A., Gairola, R.M., Varma, A.K., Agarwal, V.K., 2010. Remote sensing of precip-itation over Indian land and oceanic regions by synergistic use of multi satellitesensors. J. Geophys. Res. 115, D08106, http://dx.doi.org/10.1029/2009JD012157.

Nigam, R., Bhattacharya, B.K., Gunjal, K.R., Padmanabhan, N., Patel, N.K., 2011. Con-tinental scale vegetation index from Indian geostationary satellite: algorithmdefinition and validation. Curr. Sci. 100, 1184–1192.

Norato, M., Li, Z., Williams, J.W., 2006. Observed Vegetation-Climate feedbacks inthe United States. J. Climate 19, 763–786.

Rahman, H., Dedieu, G., 1994. SMAC: a simplified method for the atmospheric cor-rection of satellite measurements in the solar spectrum. Int. J. Remote Sens. 15,123–143.

Roads, J., Betts, A., 1999. NCEP-NCAR and ECMWF reanalysis surface water andenergy budgets for the Mississippi River Basin. J. Hydrometeorol. 1 (1), 88.

Roerink, G.J., Menenti, M., Verhoef, W., 2000. Reconstructing cloud-free NDVIcomposites using Fourier analysis of time series. Int. J. Remote Sens. 21 (9),1911–1917.

Roy, S.B., Weaver, C.P., Nolan, D.S., Avissar, R., 2003. A preferred scale for landscapeforced mesoscale circulations. J. Geophys. Res., 108.

Schulz, J., Albert, P., Behr, H.D., Caprion, D., Deneke, H., Dewitte, S., Durr, B., Fuchs,P., Gratzki, A., Hechler, P., Hollmann, R., Johnston, S., Karlsson, K.G., Manni-nen, T., Muller, R., Reuter, M., Riihela, A., Roebeling, R., Selbach, N., Tetzlaff, A.,Thomas, W., Werscheck, M., Wolters, E., Zelenka, A., 2009. Operational climatemonitoring from space: the EUMETSAT Satellite Application Facility on ClimateMonitoring (CM-SAF). Atmos. Chem. Phys. 9, 1687–1709.

Segal, M., Arritt, R., Clark, C., Rabin, R., Brown, J., 1995. Scaling evaluation of the effectof surface characteristics on potential for deep convection over uniform terrain.Mon. Weather Rev. 123, 383–400.

Sellers, P.J., Tucker, C.J., Collatz, G.J., 1994. A global 10 × 10 NDVI dataset for climatestudies. Part 2. The generation of global field of terrestrial biophysical parameterfrom the NDVI. Int. J. Remote Sens. 15 (7), 3519–3545.

Shukla, J., Mintz, Y., 1982. Influence of land-surface evapotranspiration on the earth’sclimate. Science 215, 1498–1501.

Skamarock, W.C., Klemp, J.B., Dudhia, J., Gill, D.O., Barker, D.M., Duda, M.G., Huang,X.Y., Wang Wand Powers, J.G., 2008. A Description of the Advanced ResearchWRF Version 3. NCAR/TN-475 STR; NCAR Technical Note, Mesoscale andMicroscale Meteorology Division, National Center of Atmospheric Research,June 2008, 113 pp.

Strack, J.E., Pielke Sr., R.A., Steyaert, L.T., Knox, R.G., 2008. Sensitivity of June near-surface temperatures and precipitation in the eastern United States to historicalland cover changes since European settlement. Water Resour. Res. 44, W11401,http://dx.doi.org/10.1029/2007WR006546.

Strand, J.F., 2000. Some agrometeorological aspects of pest and disease managementfor the 21st century. Agric. Forest Meteorol. 103, 73–82.

Sud, Y.C., Fennessy, M., 1982. A study of the influence of surface albedo onJuly circulation in semi-arid regions using the GLAS GCM. J. Climatol. 2,105–125.

Van der Tol, C., Meesters, A.G.C.A., Dolman, A.J., Waterloo, M.J., 2008. Opti-mum vegetation characteristics, assimilation, and transpiration during

a dry season. 1. Model description. Water Resour. Res. 44, W03421,http://dx.doi.org/10.1029/2007WR006241.

Yang, Z.L., Dai, Y., Dickinson, R.E., Shuttleworth, W.J., 1999. Sensitivity of ground heatflux to vegetation cover fraction and leaf area index. J. Geophys. Res. 104 (D16),http://dx.doi.org/10.1029/1999 JD900230, 19,505–19,514.


Recommended