+ All Categories
Home > Documents > Relationship between temperature trend magnitude ... · PDF fileRelationship between...

Relationship between temperature trend magnitude ... · PDF fileRelationship between...

Date post: 21-Mar-2018
Category:
Upload: dohanh
View: 215 times
Download: 1 times
Share this document with a friend
10
Relationship between temperature trend magnitude, elevation and mean temperature in the Tibetan Plateau from homogenized surface stations and reanalysis data Qinglong You a,b,f , Shichang Kang a,c, , Nick Pepin d , Wolfgang-Albert Flügel b , Yuping Yan e , Houshang Behrawan b , Jie Huang a,f a Laboratory of Tibetan Environment Changes and Land Surface Processes, Institute of Tibetan Plateau Research, Chinese Academy of Sciences (CAS), Beijing 100085, China b Department of Geoinformatics, Friedrich-Schiller University Jena, Jena 07743, Germany c State Key Laboratory of Cryospheric Science, CAS, Lanzhou 730000, China d Department of Geography, University of Portsmouth, PO1 3HE, UK e National Climate Center, Beijing 100081, China f Graduate University of Chinese Academy of Sciences, Beijing 100049, China abstract article info Article history: Received 13 April 2009 Accepted 5 January 2010 Available online 21 January 2010 Keywords: warming trend elevation dependency Tibetan Plateau reanalysis Temperature trend magnitudes at 71 homogenized surface stations with elevations above 2000 m asl in the eastern and central Tibetan Plateau (TP) and 56 grid points from surface NCEP and ERA-40 reanalyses in the TP's vicinity are examined. Both the surface meteorological stations and ERA-40 show general warming trends at the majority of locations, especially in winter. NCEP fails to identify this. Compared with the surface stations, both NCEP and ERA-40 reanalysis data underestimate air temperature trends in the TP, but ERA-40 is better than NCEP. There are no simple linear relationships between elevation and temperature trend magnitudes on an annual or seasonal basis in the surface data or ERA-40, and in NCEP this relationship is inconsistent. Instead there are signicant correlations between mean annual and seasonal temperatures and temperature trend magnitudes in the surface dataset and NCEP data (but not ERA-40). We suggest this is due to cryospheric feedback since trends are enhanced when mean annual temperatures are near freezing. The absence of any simple elevation dependency in temperature trends suggests that the rapid warming rate derived from high elevation ice-cores in this region should be interpreted with caution. In addition, more attention should be given to the selection of reanalysis to represent surface climate in the TP, since topographical differences between grid points and stations, and other reanalysis model differences such as surface land schemes, cause differences in trend identication and patterns in this critical region. © 2010 Elsevier B.V. All rights reserved. 1. Introduction The Fourth Assessment Report of the Intergovernmental Panel on Climate Change (IPCC) shows an increase in global mean temperature of approximately 0.74 °C during the latest century (IPCC, 2007). This warming will have important consequences for the hydrological cycle, particularly in regions where the water supply is currently dominated by melting snow or ice. More than one-sixth of the Earth's population relies on glaciers and seasonal snow packs for their water supply (Barnett et al., 2005). The Tibetan Plateau (TP), with an average eleva- tion of over 4000 m asl and an area of approximately 2.5×10 6 km 2 , is the highest and most extensive highland in the world. It has the largest glacier area in the mid-latitude regions and has therefore been called the Asian water tower(Yeh and Gao, 1979). The TP provides a crucial link for the water resources for most of the Asian continent through melting glaciers that feed the rivers, thereby impacting the livelihood of over 3.7 billion people (Zhang, 2007). For example, 23% of the population of China lives in the western regions where glacial melt provides the principal dry season water source (Barnett et al., 2005). Hydrological changes resulting from glacial retreat, such as increased discharge, rises in lake level, more frequent glacial lake outbursts leading to ooding, enhanced glacial debris ows, and changes in water resources have been the focus of many studies (Li et al., 2008). Unfortunately, the environmental consequences of global warming are already evident in the TP, including increasing air (Liu and Chen, 2000), permafrost temperature and permafrost degradation (Cheng and Wu, 2007; Wu and Zhang, 2008; Zhao et al., 2004), the accelerating melting of glaciers (e.g.,Kang et al., 2007; Tian et al., 2006), thickening of the active layer (Zhang, 2007), and increasing tem- perature extremes (You et al., 2008a). Therefore it is important to quantify the rate of warming and examine its spatial pattern. Global and Planetary Change 71 (2010) 124133 Corresponding author. Laboratory of Tibetan Environment Changes and Land Surface Processes, Institute of Tibetan Plateau Research, Chinese Academy of Sciences (CAS), Beijing 100085, China. Tel.: +86 10 6284 9681; fax: +86 10 6284 9681. E-mail address: [email protected] (S. Kang). 0921-8181/$ see front matter © 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.gloplacha.2010.01.020 Contents lists available at ScienceDirect Global and Planetary Change journal homepage: www.elsevier.com/locate/gloplacha
Transcript
Page 1: Relationship between temperature trend magnitude ... · PDF fileRelationship between temperature trend magnitude, elevation and mean temperature in the Tibetan Plateau from homogenized

Global and Planetary Change 71 (2010) 124–133

Contents lists available at ScienceDirect

Global and Planetary Change

j ourna l homepage: www.e lsev ie r.com/ locate /g lop lacha

Relationship between temperature trend magnitude, elevation and meantemperature in the Tibetan Plateau from homogenized surface stationsand reanalysis data

Qinglong You a,b,f, Shichang Kang a,c,⁎, Nick Pepin d, Wolfgang-Albert Flügel b, Yuping Yan e,Houshang Behrawan b, Jie Huang a,f

a Laboratory of Tibetan Environment Changes and Land Surface Processes, Institute of Tibetan Plateau Research, Chinese Academy of Sciences (CAS), Beijing 100085, Chinab Department of Geoinformatics, Friedrich-Schiller University Jena, Jena 07743, Germanyc State Key Laboratory of Cryospheric Science, CAS, Lanzhou 730000, Chinad Department of Geography, University of Portsmouth, PO1 3HE, UKe National Climate Center, Beijing 100081, Chinaf Graduate University of Chinese Academy of Sciences, Beijing 100049, China

⁎ Corresponding author. Laboratory of Tibetan EnvSurface Processes, Institute of Tibetan Plateau Research(CAS), Beijing 100085, China. Tel.: +86 10 6284 9681;

E-mail address: [email protected] (S. Kang

0921-8181/$ – see front matter © 2010 Elsevier B.V. Adoi:10.1016/j.gloplacha.2010.01.020

a b s t r a c t

a r t i c l e i n f o

Article history:Received 13 April 2009Accepted 5 January 2010Available online 21 January 2010

Keywords:warming trendelevation dependencyTibetan Plateaureanalysis

Temperature trend magnitudes at 71 homogenized surface stations with elevations above 2000 m asl in theeastern and central Tibetan Plateau (TP) and 56 grid points from surface NCEP and ERA-40 reanalyses in theTP's vicinity are examined. Both the surface meteorological stations and ERA-40 show general warmingtrends at the majority of locations, especially in winter. NCEP fails to identify this. Compared with the surfacestations, both NCEP and ERA-40 reanalysis data underestimate air temperature trends in the TP, but ERA-40is better than NCEP. There are no simple linear relationships between elevation and temperature trendmagnitudes on an annual or seasonal basis in the surface data or ERA-40, and in NCEP this relationship isinconsistent. Instead there are significant correlations between mean annual and seasonal temperatures andtemperature trend magnitudes in the surface dataset and NCEP data (but not ERA-40). We suggest this is dueto cryospheric feedback since trends are enhanced when mean annual temperatures are near freezing. Theabsence of any simple elevation dependency in temperature trends suggests that the rapid warming ratederived from high elevation ice-cores in this region should be interpreted with caution. In addition, moreattention should be given to the selection of reanalysis to represent surface climate in the TP, sincetopographical differences between grid points and stations, and other reanalysis model differences such assurface land schemes, cause differences in trend identification and patterns in this critical region.

ironment Changes and Land, Chinese Academy of Sciencesfax: +86 10 6284 9681.).

ll rights reserved.

© 2010 Elsevier B.V. All rights reserved.

1. Introduction

The Fourth Assessment Report of the Intergovernmental Panel onClimateChange (IPCC) shows an increase in globalmean temperature ofapproximately 0.74 °C during the latest century (IPCC, 2007). Thiswarming will have important consequences for the hydrological cycle,particularly in regions where the water supply is currently dominatedby melting snow or ice. More than one-sixth of the Earth's populationrelies on glaciers and seasonal snow packs for their water supply(Barnett et al., 2005). The Tibetan Plateau (TP), with an average eleva-tion of over4000 m asl and anarea of approximately 2.5×106 km2, is thehighest and most extensive highland in the world. It has the largestglacier area in the mid-latitude regions and has therefore been called

the “Asian water tower” (Yeh and Gao, 1979). The TP provides a cruciallink for the water resources for most of the Asian continent throughmelting glaciers that feed the rivers, thereby impacting the livelihoodof over 3.7 billion people (Zhang, 2007). For example, 23% of thepopulation of China lives in the western regions where glacial meltprovides the principal dry season water source (Barnett et al., 2005).Hydrological changes resulting from glacial retreat, such as increaseddischarge, rises in lake level, more frequent glacial lake outburstsleading to flooding, enhanced glacial debris flows, and changes inwater resources have been the focus of many studies (Li et al., 2008).Unfortunately, the environmental consequences of global warmingare already evident in the TP, including increasing air (Liu andChen, 2000), permafrost temperature and permafrost degradation(Cheng and Wu, 2007; Wu and Zhang, 2008; Zhao et al., 2004), theacceleratingmelting of glaciers (e.g.,Kang et al., 2007; Tian et al., 2006),thickening of the active layer (Zhang, 2007), and increasing tem-perature extremes (You et al., 2008a). Therefore it is important toquantify the rate of warming and examine its spatial pattern.

Page 2: Relationship between temperature trend magnitude ... · PDF fileRelationship between temperature trend magnitude, elevation and mean temperature in the Tibetan Plateau from homogenized

125Q. You et al. / Global and Planetary Change 71 (2010) 124–133

Previous research based on a combination of surface instrumentalstations, reanalysis data and numerical models (Chen et al., 2003;Frauenfeld et al., 2005; Liu and Chen, 2000) has shown a significantwarming in the TP during the last half century, in phase with theglobal trends. Liu and Chen, 2000 revealed a more pronouncedwarming at high elevation stations and when compared withsurrounding regions in the TP, which was confirmed by numericalexperiments (Chen et al., 2003). Such a tendency may continue infuture climate change scenarios (Liu et al., 2009). This elevationdependency has been widely cited (e.g., Chen et al., 2003; Frauenfeldet al., 2005) and used to explain the rapid warming derived fromice cores (Kang et al., 2007; Tian et al., 2006). However, our recentpaper (You et al., 2008b) fails to find an elevation dependency in thetrends of temperature extremes in the eastern and central TP. On aglobal scale enhanced warming with elevation is not universallyaccepted, and some observational studies show contrasting patterns orno patterns, such as in the tropical Andes (Vuille and Bradley, 2000),South America (Vuille et al., 2003), and in a global analysis of1084 homogenous stations from the GHCN (Global Historical ClimateNetwork) and CRU (Climate Research Unit) datasets (Pepin andLundquist, 2008; Pepin and Seidel, 2005).

Despite theseprevious studies (Liu andChen, 2000;Youet al., 2008b),there is no work examining elevation dependency of temperaturetrend magnitudes in the TP which uses homogenized air temperaturedata (Li et al., 2004a). This study therefore uses a homogenized datasetof 71 surface stations.

The National Centers for Environmental Prediction/NationalCenter for Atmospheric Research (NCEP/NCAR) Reanalysis (NCEPhereafter) (Kalnay et al., 1996) and European Center for Medium-Range Weather Forecasts (ECMWF) reanalysis (ERA-40 hereafter)(ECMWF, 2002) have been widely used in climate studies. Thereforewe also investigate the relationship between trends in monthly NCEPand ERA-40 reanalysis data and elevation. This is partly becausehomogenous meteorological stations are relatively sparse, but alsosince we wish to assess how strongly the reanalysis data and surfacestations agree in their patterns of trend variability.

2. Dataset and method

Monthly mean temperature of 71 stations and their elevationswere provided by the National Climate Center, China MeteorologicalAdministration (CMA). The datasets are the China HomogenizedHistorical Temperature Datasets (1951–2004 period) (version 1.0),released originally on December 14, 2006. The data has beenhomogenized to account for station relocations. Many stations,especially those in big cities, have been relocated since 1951. Someurban stations like Beijing and Shanghai have been relocated severaltimes between the suburbs and urban sites (Li et al., 2004b). In thisdataset, the main inhomogeneities resulting from station relocationhave been identified and discontinuities caused by confirmed causeshave been adjusted (Li et al., 2004a). Detailed descriptions of dataquality control and homogenization are available from the abovepapers.

Seventy-one stations were selected (Fig. 1) from the datasetsaccording to selection procedures as described in our recent papers(You et al., 2008a; You et al., 2008b). Most stations distributed in theeastern and central TP were set up in the 1950s. Elevations are allabove 2000 m asl. The period of 1961–2004 was selected for analysis.The monthly mean surface air temperature NCEP reanalysis datawere provided by the NOAA/OAR/ESRL PSD, Boulder, Colorado, USA,from their website at http://www.cdc.noaa.gov/. This data coverJanuary 1948 to the present with a spatial resolution of 2.5°×2.5(Kalnay et al., 1996). Themonthlymean surface air temperature ERA-40reanalysis data were obtained from the ECMWF website (http://www.ecmwf.int/). ERA-40 temperatures are available from September 1957to August 2002 with a spatial resolution of 2.5°×2.5° (ECMWF, 2002).

Periods of 1961–2004 and 1961–2001 were selected from NCEP andERA-40 data, respectively. The slightly shorter period for ERA-40, due tolack of more current data, has limited influence on our results(Section 3.4). The ERA-40 analysis of temperature at a height of 2 mwere produced every 6 h as part of the data assimilation but not directlyby its primary three-dimensional analysis of atmospheric fields. Thebackground 2 m temperature was derived from 6 h backgroundforecasts of skin temperature and temperature at the lowest modellevel (located at a height of 10 m), consistent with the model'sparameterization of the surface layer (Simmons et al., 2004).While screen-level temperature measurements were not used inthe NCPE/NCAR reanalysis, and the surface air temperature productwas derived instead from analyzed atmospheric values that wereconstrained primarily by observations of upper air variables andsurface pressure (Simmons et al., 2004).

To compare with surface stations, we selected grid points containedwithin the rectangle outlined in Fig. 1, thus covering the same spatialdomain. The surface elevations of the 56 grid points are extracted fromGTOPO30 digital elevation data (available from http://eros.usgs.gov).

The Mann–Kendall test for a trend and Sen's slope estimates wereused to detect and estimate trends in annual, seasonal temperatureseries (Sen, 1968). In this paper, a trend is considered to be statisticallysignificant if it is significant at the 5% level.

3. Results

3.1. Relationship between trend magnitudes at surface stations andreanalysis grid points

The average annual temperature trend at the surface stations has amagnitude of 0.25°C/decade, and most stations have significantincreasing trends, respectively (Table 1). On a seasonal basis, thegreatest mean warming rate is found in winter, with most stationsshowing significant warming. Warming occurs in all seasonsdominated by warmer winters and autumns. Mean temperaturedifferences between surface stations and reanalysis on an annual andseasonal basis are shown in Table 2. Compared with surface stations,both NCEP and ERA-40 reanalysis data have cold biases on an annualand seasonal basis. These are much larger for NCEP (mean bias of4.81°C) than for ERA-40 (0.70°C). Surface stations shows generalwarming trends on an annual and seasonal basis. For the NCEPreanalysis data, the average annual temperature trend shows only aslight decrease (−0.02°C/decade) (Fig. 2) and most grid points in thesoutheastern TP have decreasing trends. This is confirmed by theannual mean surface air temperature differences between 1983–2001and 1961–1982 (Fig. 3(A)). For the ERA-40 reanalysis, the averageannual temperature trend is 0.22 °C/decade (Fig. 2), and most gridpoints distributed in the southwestern TP have larger increasingtrends, which is consistent with the annual mean surface air tem-perature differences between 1983–2001 and 1961–1982 (Fig. 3(B)).In contrast to the surface stations, NCEP reanalysis fails to identifywarming in the TP with the exception of winter, because the surfaceland are covered by snow and ice in most regions in winter. ERA-40represents the warming trends better than NCEP and the patterns ofwarming are similar to the surface stations on an annual and seasonalbasis (Table 1). Thus the warming in the surface stations is on averagemuch stronger than in both reanalyses, in accordance with Pepin andSeidel (2005) although they compared surface data trends with thosein free-air reanalysis.

3.2. Relationship between trend magnitudes and elevation from surfacestations and reanalysis grid points

Fig. 4 (left column) shows the relationship between the temper-ature trend magnitudes at individual surface stations and elevationon an annual and seasonal basis, together with their correlation

Page 3: Relationship between temperature trend magnitude ... · PDF fileRelationship between temperature trend magnitude, elevation and mean temperature in the Tibetan Plateau from homogenized

Fig. 1. Distribution of 71 stations (dot) and 56 reanalysis grid points (cross in rectangle) in the eastern and central Tibetan Plateau and its adjacent territories.

126 Q. You et al. / Global and Planetary Change 71 (2010) 124–133

coefficients. There are slight correlations between temperature trendmagnitude and elevation for annual, summer, autumn andwinter, butthey are not significant. The only strong positive relationship occurs inspring. Categorizing the stations into 6 elevation bands with aninterval of 500 m, the largest mean trend magnitudes appear at arelatively low elevation of 2500–3000 m in most cases (highlighted inbold) (Table 3(a)). Thus the relationship between trend magnitudesand elevation is not clear.

Fig. 4 (middle column) shows similar graphs for NCEP reanalysisgrid points. Again the relationships between trend magnitudes andelevation are not systematic. There are significant positive relation-ships on an annual basis and in spring, but the opposite relationship inautumn. The elevation band with the maximum trend magnitude(highlighted in bold) is very varied (Table 3(b)). Thus, the patternsare stronger but less systematic in NCEP reanalysis data in comparison

Table 1Descriptive statistics and number of stations, NCEP and ERA-40 reanalysis grid pointswith positive (significant at the 0.05 level) and negative (significant at the 0.05 level)temperature trend magnitudes on an annual and seasonal basis. The study periodsfor stations, NCEP, ERA-40 are during 1961–2004, 1961–2004 and 1961–2001,respectively.

Mean Std. deviation Positive trends Negative trends

Stations Annual 0.25 0.11 70 (64) 1 (1)Spring 0.17 0.10 68 (31) 3 (1)Summer 0.20 0.11 70 (53) 1 (0)Autumn 0.26 0.13 69 (54) 2 (0)Winter 0.40 0.17 70 (62) 1 (0)

NCEP Annual −0.02 0.11 20 (3) 36 (8)Spring −0.14 0.20 11 (2) 45 (18)Summer −0.05 0.10 20 (4) 36 (9)Autumn −0.03 0.11 25 (0) 31 (2)Winter 0.13 0.23 41 (19) 15 (0)

ERA-40 Annual 0.23 0.14 53 (41) 3 (0)Spring 0.11 0.16 44 (14) 12 (0)Summer 0.17 0.15 50 (23) 6 (0)Autumn 0.27 0.13 56 (38) 0 (0)Winter 0.36 0.25 52 (35) 4 (1)

with the surface stations. For ERA-40 reanalysis data, the relationshipsbetween temperature trend magnitude and elevation are not clear(Fig. 4 (right column)). However, the largest mean trend magnitudesall emerge at 2000–2500 m asl (Table 3(c)), which is more similar tothe surface stations. Overall there is very little consistent elevationalsignal in all datasets.

3.3. Relationship between trend magnitudes and mean temperature fromsurface stations and reanalysis grid points

Relationships between trend magnitudes and mean annualtemperature are stronger. For surface stations, there are strongnegative correlations between temperature trend magnitude andmean temperature on an annual and seasonal basis (Pb0.05). This istrue in all seasons (Fig. 5 (left column)). Faster warming trends occurat lower temperature. It is clear that mean annual temperaturesaround 0°C have large warming trends, and negative relationshipsbetween trend magnitude and mean temperature appear abovefreezing. Although the scarcity of station below 0°C limits the study,there is a suggestion that this relationship breaks down at meantemperatures below freezing at least in the annual, spring and autumndata.

For NCEP reanalysis data, the negative relationships are alsoevident on an annual basis and in spring and winter, but disappear insummer and autumn (Fig. 5 (middle column)). ERA-40 (Fig. 5 (rightcolumn)) fails to show any relationships, apart from an enhancedwarming at lower temperatures similar to the surface stations inwinter.

Table 2Mean temperature differences between surface stations and reanalysis on an annualand seasonal basis. Units are degree.

Annual Spring Summer Autumn Winter

Surface stations minus NCEP 4.81 5.55 3.74 4.49 5.49Surface stations minus ERA-40 0.70 0.67 0.48 1.04 0.65

Page 4: Relationship between temperature trend magnitude ... · PDF fileRelationship between temperature trend magnitude, elevation and mean temperature in the Tibetan Plateau from homogenized

Fig. 2. Average regional trends for surface stations, NCEP and ERA-40 (left plot) and comparisons between temperature of surface stations and temperature from NCEP and ERA-40reanalysis data (right plot) on an annual basis. The straight lines are linear fits and Lin stands for trends per decade, and R and P for correlation coefficients and statistical significance,respectively. The study periods are same with Table 1.

127Q. You et al. / Global and Planetary Change 71 (2010) 124–133

3.4. Influence of data period and spatial domain

We chose to demonstrate trends for the surface stations and NCEPup to 2004 to use the most up to data records available. Howeverbecause ERA-40 was only available up to 2001, it has a slightly shorterperiod. To assess the importance of this on our comparisons, we

Fig. 3. Annual mean surface 2 m temperature differences between 1983–20

recalculated results for NCEP and the surface stations for 1961–2001and the differences in trends were negligible (not shown). Certainly incomparison to the similarities/differences discussed earlier betweendatasets, the influence of data period is minimal in this case. We alsorecalculated trends for sub-periods 1961–1982 and 1983–2004(1983–2001 for ERA-40) and although there were a few differences

01 and 1961–1982: A for NCEP/NCAR and B for ERA-40 reanalysis data.

Page 5: Relationship between temperature trend magnitude ... · PDF fileRelationship between temperature trend magnitude, elevation and mean temperature in the Tibetan Plateau from homogenized

Fig. 4. Temperature trend magnitudes (°C/decade) of stations (left), NCEP (middle) and ERA-40 (right) reanalysis grid points versus elevation (m asl) on an annual and seasonalbasis. Others are same with Fig. 3.

128 Q. You et al. / Global and Planetary Change 71 (2010) 124–133

Page 6: Relationship between temperature trend magnitude ... · PDF fileRelationship between temperature trend magnitude, elevation and mean temperature in the Tibetan Plateau from homogenized

Table 3Mean trend magnitudes of temperature for (a) stations, (b) NCEP and ERA-40 reanalysis grid points in categorized elevation bands. The elevation band with maximummagnitude ishighlighted in bold. Units are degrees per decade. The study periods are the same with Table 1.

Altitude(m)

2000–2500 2500–3000 3000–3500 3500–4000 4000–4500 4500–5000

(a)No. of stations 9 19 16 13 9 5Mean of elevation (m) 2289 2810 3261 3753 4199 4605Annual 0.15 0.27 0.25 0.26 0.25 0.27Spring 0.06 0.18 0.16 0.21 0.19 0.22Summer 0.12 0.23 0.20 0.22 0.19 0.21Autumn 0.16 0.28 0.26 0.29 0.28 0.28Winter 0.34 0.46 0.43 0.36 0.32 0.38

(b)No. of grids 3 2 3 6 6 6Mean of elevation (m) 2208 2780 3249 3784 4237 4787Annual −0.10 −0.04 −0.07 −0.05 −0.03 0.06Spring −0.29 −0.17 −0.31 −0.18 −0.02 0.06Summer −0.06 0 −0.10 −0.16 −0.06 −0.01Autumn −0.06 −0.03 −0.04 −0.08 −0.06 −0.05Winter 0.03 0.04 0.18 0.30 0.06 0.17

(c)No. of grids 3 2 3 6 6 6Mean of elevation (m) 2208 2780 3249 3784 4237 4787Annual 0.46 0.19 0.28 0.25 0.09 0.19Spring 0.35 0.06 0.11 0.17 −0.01 0.09Summer 0.45 0.03 0.17 0.26 0.05 0.20Autumn 0.50 0.32 0.35 0.32 0.15 0.25Winter 0.52 0.40 0.49 0.33 0.26 0.28

129Q. You et al. / Global and Planetary Change 71 (2010) 124–133

in seasonal relationships, the main findings of our work remainedunchanged.

In our analysis, we chose reanalysis grid points from the samespatial domain as the surface stations. This leads to some relativelylow elevation grid points (b2000 m) being included on the edges ofthe area. Subsequent removal of these points does not make asignificant difference to the main patterns of trends identified,including their magnitudes or relationships with elevation or meanannual temperature.

4. Discussion and conclusions

We have analyzed the relationships between temperature trendmagnitudes and elevation using surfacemeteorological stations, NCEPand ERA-40 reanalysis grid points in the eastern and central TP.Surface stations indicate a general warming trend, mostly due towinter and autumn,which is consistentwith previous studies (Liu andChen, 2000) and the reported dramatic glacier shrinking in the region(Kang et al., 2007; Tian et al., 2006; Zhang, 2007). Reanalysis datashows less marked warming but ERA-40 captures the surfacewarming trend better than NCEP. There are no simplistic linearrelationships between temperature trend magnitudes and elevationin any dataset, and therefore a suggestion of elevation dependencybased on rapid warming rates derived from ice cores at highelevations could be misleading, even though earlier studies(Chen et al., 2003; Liu and Chen, 2000) did suggest enhancedsensitivity. The lack of a simplistic elevation increase of warmingrates in this study agrees with our findings concerning therelationship between the trend magnitudes of temperatureextremes and elevation in this same region (You et al., 2008b). Inthat study, elevation effects were overridden by topographic varia-tions and the degree of urbanization at stations.

To investigate whether topography influences trend magnitudesin our datasets, each of the 71 stations was classified into 3 topo-graphic types: summit, flat and valley, according to a topographicalindex derived from GTOPO30 digital elevation data (available fromhttp://eros.usgs.gov). We calculated a topographical index by over-laying a grid centered on each station and calculated the elevation

difference between the station and the eight surrounding cells(You et al., 2008b). Table 4 shows the number of stations andtemperature trend magnitudes in differing topographical classes andin categorized elevation ranks in the eastern and central TP on anannual and seasonal basis. For summit stations, there are few stationsand stations at higher elevations have largest trend magnitudes. Bothflat and valley stations have equal numbers. In categorized elevationranks on an annual basis, temperature trend magnitudes for flatstations are 0.16, 0.34, 0.25, 0.25, 0.23, 0.29 degree/decade for annual,spring, summer, autumn, winter, respectively, which are larger thanthose for valley stations in most cases. Stations between 2500 and3000 m have larger temperature trend magnitudes for flat stationswhile are larger for flat stations and the elevation bands with themaximum trend magnitude for valley stations are very varied. Forboth flat and valley stations, there are larger temperature trendmagnitudes in winter and autumn. Our analysis shows thattopography does influence the trend magnitudes. In particularbecause most flat stations (with steeper trends) are at lowerelevations, this acts to reduce elevational dependency somewhat inthis dataset.

On a global scale it has been shown that there is no simplerelationships between elevation and warming rates (Pepin andSeidel, 2005), partly because of the role of cryospheric feedback(Pepin and Lundquist, 2008). Our study suggests that in the TP thereare significant negative correlations between trend magnitudes andmean temperature on an annual and seasonal basis at the surfacestations. Enhanced warming occurs where the mean annual temper-ature is currently near freezing point in the surface data. We suggestthe closer the current mean annual temperature to freezing themore potential for enhanced cryospheric feedback (Pepin andLundquist, 2008). This does not transfer to a clear relationshipbetween warming rate and elevation because in this topographicallycomplex landscape with frequent temperature inversions meanannual temperatures do not simplistically decrease with elevation.Regions near this critical temperature should be given more attentionin terms of environmental consequences on ecosystems. Althoughthe increased sensitivity as temperatures approach freezing fromabove is extremely clear in the surface data, below 0 °C it is less

Page 7: Relationship between temperature trend magnitude ... · PDF fileRelationship between temperature trend magnitude, elevation and mean temperature in the Tibetan Plateau from homogenized

Fig. 5. Temperature trend magnitudes (°C/decade) of stations (left), NCEP (middle) and ERA-40 (right) reanalysis grid points versus mean temperature on an annual and seasonalbasis. Others are same with Fig. 3.

130 Q. You et al. / Global and Planetary Change 71 (2010) 124–133

Page 8: Relationship between temperature trend magnitude ... · PDF fileRelationship between temperature trend magnitude, elevation and mean temperature in the Tibetan Plateau from homogenized

Table 4Number of stations and temperature trend magnitudes in differing topographical classes and in categorized elevation ranks in the eastern and central TP on an annual and seasonalbasis. The elevation band with maximum magnitude is highlighted in bold. Units are degrees per decade.

2000-2500 2500–3000 3000–3500 3500–4000 4000–4500 4500–5000

Summit station Number 2 0 3 0 0 1Annual 0.15 – 0.25 – – 0.30Spring 0.16 – 0.17 – – 0.19Summer 0.09 – 0.18 – – 0.11Autumn 0.14 – 0.27 – – 0.32Winter 0.27 – 0.42 – – 0.58

Flat station Number 3 11 6 4 4 4Annual 0.16 0.34 0.25 0.25 0.23 0.29Spring 0.09 0.22 0.14 0.18 0.15 0.25Summer 0.13 0.28 0.22 0.24 0.18 0.23Autumn 0.15 0.35 0.27 0.30 0.24 0.29Winter 0.34 0.58 0.44 0.36 0.34 0.35

Valley station Number 4 8 6 9 5 1Annual 0.14 0.23 0.24 0.27 0.26 0.22Spring 0.01 0.16 0.17 0.22 0.22 0.17Summer 0.11 0.21 0.17 0.21 0.20 0.23Autumn 0.16 0.25 0.26 0.29 0.32 0.23Winter 0.33 0.38 0.42 0.36 0.31 0.25

131Q. You et al. / Global and Planetary Change 71 (2010) 124–133

apparent whether trends continue to amplify as mean annualtemperatures decrease further. In the transition seasons and on anannual basis this does not appear to be the case. However in winter,when there is more data, the relationship appears to hold well belowfreezing. In theory the cryospheric feedback loop will be strongest at0 °C when snow and ice melt and diminish at lower temperatures.Further research is required into the form of this forcing in extremelycold climates.

This sensitivity to temperature is also a feature of NCEP reanalysisgrid point data on an annual basis, and in winter and spring, whencryospheric feedback is likely to be especially important. Thus NCEPdata, although showing slight cooling in the region, does capture trendvariability in relation to mean temperature to an extent. Less coolingcould be because of large biases in reanalysis “surface” temperatures,which is consistent with previous research comparing reanalysis andsurfacedata in theTP (ZhaoandFu,2006). To investigate the influenceofelevation differences on our temperature comparisons for bothreanalyses,we analyze the correlations between annual air temperaturedifferences (station minus reanalysis, dT) and elevation differences(mean surface station minus reanalysis model elevation, dH) forNCEP (top plot) and ERA-40 (middle plot) during 1961–2004(Fig. 6). We also show the correlation between annual reanalysistemperature differences (ERA-40 minus NCEP) and the elevation dif-ference (ERA-40 minus NCEP). There are significant negative correla-tions for NCEP (R=−0.39, Pb0.05) and ERA-40 (R=−0.77, Pb0.05),but the relationship is much stronger in ERA-40, annually andseasonally. In most cases, the elevation differences (model elevationminus surface stations elevation) are positive because surface stationsare situated in the flat and valley bottoms lower than the reanalysismodel topography. Thus most of the temperature bias in ERA-40 dueto the elevation difference. This is not the case for NCEP where therelationship is confused by some anomalously cold locations atrelatively lower elevations. Thus assimilating surface station elevationsinto the ERA-40 reanalysis elevation model would minimize tempera-ture difference. However such a method would be of limited use inNCEP which has more serious deficiencies.

ERA-40 data captures the general warming better than NCEP,because NCEP does not assimilate surface observations such astemperature, moisture, wind, while ERA-40 does when available(Frauenfeld et al., 2005). ERA-40 also assimilates land surface schemesin the models. In order to know whether the land surface schemesin reanalysis system influence trend magnitudes, we select lowvegetation and surface roughness to analyze. Table 5 shows the meantrend magnitudes of temperature in categorized low vegetation and

surface roughness ranks on the annual and seasonal basis. For lowvegetation type, stations in semi-desert and short grass have largertemperature trend magnitudes, with rates of 0.35 and 0.25 degree/decade on an annual basis respectively. Those trends are morepronounced in winter and autumn. It has the same patterns on theseasonal basis. For surface roughness type, stations in the rankbetween 5 and 10 m have the largest trend magnitudes on the annualand seasonal basis. There are larger temperature trend magnitudes inwinter and autumn. It is clear that land-atmosphere interaction has animportant role in climate change. For the TP, the heterogeneouslandscape effects the aerodynamic and thermodynamic roughnesslength and changes in surface heat sources caused by the underlyinglandscape could contribute to the climate change in the TP. Theheterogeneous landscape, such as desert and Gobi, has lower surfaceheat capacity than grassland and farmland, which indicates that thoseregions have more rapid warming. On the other hand, due to humanactivities and climate change together, parts of the TP are undergoinggrassland and permafrost degradation, desertification and deforesta-tion, and other ecological phenomena(Cui and Graf, 2009), which tosome extent changed the original surface landscape and influencedthe balance radiation between land and atmosphere. All above factorscontribute to the increasing warming and accelerates the warming inthe TP. This indicates that assimilating surface observations in themodels can influence the results, which is consistent with our findingsthat ERA-40 is more realistic than NCEP.

In summary, which reanalysis is “better” at representing surfacetrend patterns and variability depends on the criterion used. Comparedwith the surface stations, the trend magnitudes from NCEP have morenegative values, but the relationship between trend magnitudes andmean annual temperature is picked up to an extent. We suspect thatthe underestimation of trends is partly caused by inconsistent modeltopography in the reanalysis system (Ma et al., 2008). Zhao andFu (2006) suggest that NCEP reanalysis data underestimate themagnitude of air temperature trends in the TP and that instantaneoustemperature differences between NCEP and surface stations areinconsistent, being much larger in western than in eastern China.Our comparisons substantiate the general unreliability of NCEP data inthis region, particularly when climate trendmagnitudes are concerned.ERA-40 is much more similar to surface stations in terms of warmingrates. This is probably because ERA-40 temperature analysis wasderived by analyzing surface synoptic observation and NCEP was not(Simmons et al., 2004). Previous comparison of air temperature fromERA-40 with those from meteorological stations in the TP has shownthat any temperature bias is almost exclusively due to differences in

Page 9: Relationship between temperature trend magnitude ... · PDF fileRelationship between temperature trend magnitude, elevation and mean temperature in the Tibetan Plateau from homogenized

Table 5The mean trend magnitudes of temperature in categorized low vegetation and surfaceroughness ranks on the annual and seasonal basis (Unit: □ per decade).

Annual Spring Summer Autumn Winter

Low vegetation Crops, mixedfarming

0.16 0.05 0.12 0.21 0.36

Short grass 0.25 0.18 0.21 0.26 0.38Irrigated crops 0.17 0.10 0.15 0.19 0.31Semi-desert 0.35 0.20 0.27 0.41 0.59Evergreenshrubs

0.19 0.15 0.12 0.19 0.35

Surfaceroughness (m)

0–5 0.25 0.14 0.21 0.28 0.455–10 0.28 0.19 0.23 0.30 0.45N10 0.22 0.18 0.17 0.22 0.31

The maximum trend magnitude is highlighted in bold.

Fig. 6. Correlations between annual air temperature difference (station minus reanalysis,dT) andelevationdifferences (mean stationminusDEMelevations, dH) forNCEP (topplot),ERA-40 (middle plot) during 1961–2004. The bottom plot shows the correlation betweenannual temperature differences between ERA-40 and NCEP and elevation differencebetween ERA-40 and NCEP DEM. Others are same with Fig. 3.

132 Q. You et al. / Global and Planetary Change 71 (2010) 124–133

elevation between the grid point and the station (Frauenfeld et al.,2005). However the relationships with mean annual temperature, soevident in the surface data, are non-existent. Differences in surface landscheme in the models are likely to be partly responsible and requirefurther research. Our results strengthen the case for being critical when

selecting a reanalysis to represent surface climate in the TP and thisshould be acknowledged when using reanalysis data as input forregional climate modeling.

Acknowledgments

This study is supported by the National Natural Science Foundationof China (40771187, 40830743), National Basic Research Program ofChina (2005CB422004), Key Program of CAS (KZCX2-YW-145), andSixth Framework Program Priority (036952). The authors thank theNational Climate Center, China Meteorological Administration, forproviding the meteorological data for this study. Qinglong You thanksthe ‘DAAD-CAS Joint Scholarship Program’ and the ‘graduate innova-tion research for science and society practice special imbursementactivity of CAS in 2008’.

References

Barnett, T.P., et al., 2005. Potential impacts of a warming climate on water availability insnow-dominated regions. Nature 438, 303–309.

Chen, B., et al., 2003. Enhanced climatic warming in the Tibetan Plateau due to doublingCO2: a model study. Clim. Dyn. 20, 401–413.

Cheng, G.D., Wu, T.H., 2007. Responses of permafrost to climate change and theirenvironmental significance, Qinghai–Tibet Plateau. J. Geophys. Res.-Earth Surf. 112.

Cui, X.F., Graf, H.F., 2009. Recent land cover changes on the Tibetan Plateau: a review.Climatic Change 94, 47–61.

European Center for Medium-Range Weather Forecasts (ECMWF), 2002. Workshopon re-analysis, 5–9 November 2001. ERA-40: Proj. Rep. Ser., vol. 3. 443 pp.,Reading. U.K.

Frauenfeld, O.W., et al., 2005. Climate change and variability using European Centre forMedium-Range Weather Forecasts reanalysis (ERA-40) temperatures on theTibetan Plateau. Journal of Geophysical Research-Atmospheres 110, D02101.

IPCC, 2007. Summary for Policymakers of Climate Change 2007: The Physical ScienceBasis. Contribution of Working Group I to the Fourth Assessment Report ofthe Intergovernmental Panel on Climate Change. Cambridge University Press,Cambridge, UK. M.

Kalnay, E., et al., 1996. The NCEP/NCAR 40-year Reanalysis Project. Bulletin of theAmerican Meteorological Society 77 (3), 437–471.

Kang, S.C., et al., 2007. Recent temperature increase recorded in an ice core in the sourceregion of Yangtze River. Chinese Science Bulletin 52, 825–831.

Li, Q.X., et al., 2004a. Detecting and adjusting temporal inhomogeneity in Chinese meansurface air temperature data. Advances in Atmospheric Sciences 21, 260–268.

Li, Q.X., et al., 2004b. Urban heat island effect on Annual Mean Temperature duringRecent 50 Years in China. Theoretical and Applied Climatology 79 (4), 165–174.

Li, X., et al., 2008. Cryospheric change in China. Glob. Planet. Change 62, 210–218.Liu, X.D., Chen, B.D., 2000. Climatic warming in the Tibetan Plateau during recent

decades. International Journal of Climatology 20, 1729–1742.Liu, X.D., et al., 2009. Elevation dependency of recent and future minimum surface air

temperature trends in the Tibetan Plateau and its surroundings. Glob. Planet.Change 68, 164–174.

Ma, L.J., et al., 2008. Evaluation of ERA-40, NCEP-1, and NCEP-2 reanalysis airtemperatures with ground-based measurements in China. Journal of GeophysicalResearch-Atmospheres 113.

Pepin, N.C., Lundquist, J.D., 2008. Temperature trends at high elevations: patternsacross the globe. Geophysical Research Letters 35.

Pepin, N.C., Seidel, D.J., 2005. A global comparison of surface and free-air temperaturesat high elevations. Journal of Geophysical Research-Atmospheres 110, D03104.

Sen, P.K., 1968. Estimates of regression coefficient based on Kendall's tau. Journal of theAmerican Statistical Association 63, 1379–1389.

Page 10: Relationship between temperature trend magnitude ... · PDF fileRelationship between temperature trend magnitude, elevation and mean temperature in the Tibetan Plateau from homogenized

133Q. You et al. / Global and Planetary Change 71 (2010) 124–133

Simmons, A.J., et al., 2004. Comparison of trends and low-frequency variability in CRU,ERA-40, and NCEP/NCAR analyses of surface air temperature. Journal ofGeophysical Research-Atmospheres 109, D24115, doi:10.1029/2004JD005306.

Tian, L.D., et al., 2006. Recent rapid warming trend revealed from the isotopic record inMuztagata ice core, eastern Pamirs. Journal of Geophysical Research-Atmospheres111, D13103.

Vuille, M., Bradley, R.S., 2000. Mean annual temperature trends and their verticalstructure in the tropical Andes. Geophysical Research Letters 27, 3885–3888.

Vuille, M., et al., 2003. 20th century climate change in the tropical Andes: observationsand model results. Climatic Change 59, 75–99.

Wu, Q.B., Zhang, T.J., 2008. Recent permafrost warming on the Qinghai–Tibetan plateau.Journal of Geophysical Research-Atmospheres 113.

Yeh, T.C., Gao, Y.X., 1979. Meteorology of the Qinghai–Xizang (Tibet) Plateau (in Chinese).Science Press, Beijing.

You, Q.L., et al., 2008a. Changes in daily climate extremes in the eastern and centralTibetan Plateau during 1961–2005. Journal of Geophysical Research-Atmospheres113, D07101.

You, Q.L., et al., 2008b. Relationship between trends in temperature extremes andelevation in the eastern and central Tibetan Plateau, 1961–2005. GeophysicalResearch Letters 35, L04704.

Zhang, T.J., 2007. Perspectives on environmental study of response to climatic and landCover/Land use change over the Qinghai-Tibetan plateau: an introduction. ArcticAntarctic and Alpine Research 39, 631–634.

Zhao, T.B., Fu, C.B., 2006. Preliminary comparison and analysis between ERA-40, NCEP-2reanalysis and observations over China. Clim. Environ. Res. 11, 14–32.

Zhao, L., et al., 2004. Changes of climate and seasonally frozen ground over thepast 30 years in Qinghai–Xizang (Tibetan) Plateau, China. Glob. Planet. Change 43,19–31.


Recommended