搜档网
当前位置:搜档网 › Impacts of climate warming on vegetation in Qaidam Area

Impacts of climate warming on vegetation in Qaidam Area

Impacts of climate warming on vegetation in Qaidam Area from1990to2003

Biao Zeng&Tai-Bao Yang

Received:28March2007/Accepted:19September2007/Published online:27October2007

#Springer Science+Business Media B.V.2007

Abstract The observed warming trend in the Qaidam area,an arid basin surrounded by high mountains,has caused land surface dynamics that are detectable using remotely sensed data.In this paper,we detected land-cover changes in the Qaidam Area between1990 and2003in attempt to depict its spatial variability. The land-cover changes were categorized into two trends:degradation and amelioration,and their spatial patterns were examined.Then we estimated the corre-lation coefficients between growing-season NDVI and several climatic factors with the consideration of duration and lagging effects.The results show that the inter-annual NDVI variations are positively correlated with May to July precipitations,but not significantly correlated with sunshine duration.We observed no obvious trend in precipitation or sunshine duration from1990to2003.Thus,the authors suggest that their slight fluctuations may not be responsible to the decade-scaled land-cover changes.However,our results indicate a good positive relationship between the NDVI trend and climate warming in the amelio-rated areas,but a negative one in the degraded areas. By statistical analyses,we found that degradations mainly occurred at the oasis boundaries and at lower elevations in the non-oasis regions where effective soil moisture might have been reduced by the warming-caused increase in evapotranspiration.At higher eleva-tions where thermal condition acts as a major limiting factor,ameliorations were unequivocally detected, which is attributable to the direct facilitation by tem-perature increases.We suggest that the impacts of the observed climate warming on vegetation are spatially heterogeneous,depending on the combinations of thermal condition and moisture availability. Keywords Spatial variability.NDVI.

Land-cover change.Degradation.Amelioration Introduction

In recent decades,land surfaces over the Northern Hemisphere experience a general warming,which was modeled to continue in the following future (IPCC2001).This climate warming can cause and have caused detectable ecosystem changes(Hughes 2000;Walther et al.2002),because climatic fluctua-tions are usually of great importance to vegetation dynamics at both global and regional scales.Detailed knowledge about the ecological impacts of increased temperatures and their spatial variability will contrib-

Environ Monit Assess(2008)144:403–417

DOI10.1007/s10661-007-0003-x

B.Zeng

:T.-B.Yang(*)

College of Resource and Environmental Science, Lanzhou University,

Lanzhou730000,China

e-mail:yangtb_lzu@https://www.sodocs.net/doc/e212175063.html,

B.Zeng

:T.-B.Yang

Key Laboratory of Western China’s Environmental Systems(Ministry of Education),Lanzhou University, Lanzhou730000,China

ute to more robust projections of future scenarios of environmental changes.Researches on these issues, at present,are gaining a major focus from the scientific community(Braswell et al.1997;Hughes2000; Myneni et al.1997;Stenseth et al.2002;Walther et al.2002).In recent studies,it is generally concluded that terrestrial vegetation activities in the Northern Hemisphere have increased over the past two decades owing to elevated temperatures,especially in the high latitudes(Lucht et al.2002;Nemani et al.2003; Slayback et al.2003;Tucker et al.2001).

However,vegetations in arid lands may have different responses to climate change when compared to those in other regions with different thermal and hydrological condition.They prove most sensitive to climate changes(Roerink et al.2003).In particular, vegetation in cold drylands can be closely linked with both temperature and moisture.Thripathy et al. (1996)indicated that environmental change is appar-ent in the arid and semi-arid regions of central Asia due to little protection from the sparse vegetation cover.But great uncertainties still exist in predicting the regional and local responses to climate changes in these regions(Lioubimtseva et al.2005).Based on meteorological measurements,IPCC report(2001) pointed out a likely1–2°C/century warming in Asia interior,which must have great impacts on the regional ecosystem.Unfortunately,inadequate atten-tion has been currently paid to this area(Fang et al. 2004;Lioubimtseva et al.2005;Shi et al.2003;Yu et al.2003,2004).

The Qaidam Area lies in the eastern part of the arid Asia interior,having fewer population and human activities.It provides a unique environment for dis-cerning how warming climates influence arid vegeta-tion,though anthropogenic perturbation becomes another responsible factor for vegetation dynamics (Leemans and Zuidema1995;Lioubimtseva et al. 2005).Land-use changes can directly lead to vegeta-tive alterations.Furthermore,local human activities can significantly disturb hydrologic and nutrient cycles in arid lands,which in turn influence floristic growth and distribution(IPCC2001;Lioubimtseva et al.2005).Therefore,it is vital to distinguish climate impacts from human disturbances,in order to recog-nize the natural mechanism of vegetation dynamics. Fortunately,the landscapes in the Qaidam Area are mostly kept natural in recent decades.The researches of this region will facilitate a comprehensive under-standing of the environmental changes in developed drylands with considerable human impacts.

Multi-temporal analysis is a traditional method to detect vegetation dynamics.It relies on accurate mon-itoring of land surface attributes(Coppin et al.2004). However,field data currently available are generally disqualified to serve this purpose because they are commonly collected at rather narrow spatial and temporal scales and may vary in type and reliability (Curran and Williamson1986;Pettorelli et al.2005). In view of these defects,remote sensing is a feasible and efficient alternative to retrieve vegetation distribu-tion and coverage at large scales(Eidenshink and Faundeen1994;Kerr and Ostrovsky2003).

Many researchers have attempted to retrieve re-gional land-covers by satellite data with high temporal frequency.In these studies,the normalized difference vegetation index(NDVI,Rouse et al.1974)is commonly used to quantify the biophysical property of land surface.NDVI is established as the difference of the spectral reflectance measurements between the near-inferred and red bands normalized by their sum, based on the fact that chlorophyll absorbs red radiation whereas the mesophyll leaf structure scat-ters near-inferred radiation(Rouse et al.1974). DeFries et al.(1994,1995)have developed multi-temporal phenological metrics to derive land-cover classifications from a time series of NOAA(National Oceanic and Atmospheric Administration)A VHRR (Advanced Very High Resolution Radiometer)NDVI data.Loveland and Belward(1997)have mapped an IGBP-DIS global land-cover product using unsu-pervised classification of multi-temporal NDVI data with post-classification refinement.The temporal length and frequency of A VHRR NDVI also make it possible to observe and model the relationship be-tween vegetation and climatic factors(Lucht et al. 2002;Yu et al.2004;Nezlin et al.2005).

The goal of this study are to retrieve the spatial and temporal patterns of vegetation dynamics in the Qaidam Area between1990and2003,and to examine the decadal trends,and then to analyze the impacts of warming climate on vegetation cover.We compared the vegetation maps that were independently classified by the seasonal NDVI sequences of two time periods (the early1990s and the early2000s).The vegetation trends were extracted and their distributions were examined by elevation.Then the authors estimated the correlations between growing-season NDVI and

several climatic factors in lands with different dynamic trends.Our results show that vegetation responses to climate warming can be spatially heterogeneous in such an arid environment,owing to the multiple im-pacts of increased temperature.

Materials and methods Study area

The Qaidam Basin,surrounded by the Altun,Qilian and Qimantagh Mountains,lies in the northeastern margin of Qinghai –Tibet Plateau (Fig.1).It is one of the largest arid inter-mountain basins in the Asia continent.The Qaidam Area concerned by this paper includes the basin interior and the surrounding moun-tains.It has an area of approximately 260,000km 2between latitudes 35°–40°N and longitudes 90°–99°E.The average elevation of the basin interior is about 2,700m above the sea level and the mountaintops even rise to over 5,000m.

Generally,the growing season continues from May through September in all of the Qaidam area.Annual mean temperature is lower than 5°C,and growing-season mean temperature is about 12°C.Owing to the cloudless climate,annual sunshine duration in this

basin is generally sufficient for vegetation growth.In contrast,mean annual precipitation is insufficient and even below 50mm in some areas of the basin interior while the annual evaporation is over 25times greater than precipitation.These generate an arid temperate climate in the basin interior.However,climate be-comes colder and relatively wetter in the high-mountain areas.Recent meteorological measurements indicate that the Qaidam Area is undergoing a warming on the order of 1–2°C.Especially this trend becomes ever strong in the recent decade (see Fig.2).

Owing to hyper-aridity,the main land-covers in the basin interior are barrenland and shrub desert with extremely low vegetation coverage.Short semi-shrubs and fast-springing grasses are the dominant vegetations,most of which belong to the halophytic and xeric plant species.However,some oases,with water supplied by fault springs,locally distribute below the piedmont zone.Masses of grassland establish in the inner parts of the oases,which are surrounded by shrubland and shrub-desert in the outer parts.Vegetation cover can change dramatically along moisture gradient within the oases.With elevation rising,temperature declines and moisture availability increases in the mountain areas.There zonally appear shrubland and grassland,whose vegetation coverage changes from an intermediate density in the eastern mountains to a low density in the

western.

Fig.1Study area

The Qaidam Area,limited by the bad geo-location, has not experienced any extensive exploitation in the recent history.At present,there is still only a small quantity of agricultural activities in some oases in the eastern part.Thus,the non-disturbed ecosystem in the Qaidam Area can provide natural information on vegetation dynamics under warming climates in both arid and high-elevation regions.

Datasets

1-km monthly NDVI data for land-cover classifications

We used1-km A VHRR NDVI dataset in calculating monthly NDVI for the classification in the early1990s. The original data that we used were parts of the1-km A VHRR Global Land Dataset with a10-day composite period.They were provided through the National Center for Earth Resources Observation&Science (EROS)by the U.S.Geological Survey(USGS).This dataset has been noise-removed,radiometric-calibrat-ed and corrected for ozone and Rayleigh scattering (Eidenshink and Faundeen1994).In monthly NDVI calculations,we adopted the maximum-value com-positing method(Holben1986).Terra MODIS1-km monthly NDVI products(MOD13A3)were used for the early2000s.They were downloaded through National Aeronautics and Space Administration (NASA)EOS Gateway.MODIS Vegetation Index Algorithm Theoretical Basis Document Version3.0 provides detailed information on processing this dataset(Huete et al.1999).All the image data were geometrically transformed into the Lambert Azimuthal Equal Area projection using control point approach and the nearest neighbor algorithm was adopted to preserve quality assurance.

Because of different sensor characteristics and pro-cessing methods,there have some differences between the MODIS and A VHRR NDVI products(Vermote et al.2002;Huete et al.1999).However,the1-km NDVI data would be only used in the independent land-cover classifications,which mainly rely on the NDVI profiles and the relative values from a same dataset,rather than the absolute values across sensors. Thus,we may not need to do more corrections and inter-calibrations for the NDVI datasets.

A successive NDVI sequence

Time sequences of NDVI data are now commonly used in vegetation monitoring and studies focusing on the effects of climate changes on plants(Pettorelli et al. 2005;Running and Nemani1988;Wang et al.2003

;

Yu et al.2003).In this study,we used a successive 14-year(1990–2003)NOAA A VHRR NDVI series to observe the interannual variations and the decadal trend in vegetation.This dataset is15-day-maximum composite with a spatial resolution of8km.It is processed and archived by the Global Inventory Mapping and Monitoring Study(GIMMS)group at NASA/Goddard Space Flight Center(Tucker et al. 2005).The GIMMS group uses Empirical Mode Decomposition to identify and remove parts of the NDVI noise that are most related to the satellite drift (Pinzon2002).Stratospheric aerosol corrections have been applied to remove the effects of Mt.Pinatubo eruptions from1991to1993(Vermote et al.1997).In addition,calibrations based on invariant desert targets have also been applied to the original data to minimize the effects of sensor degradation(Los1993).

Using the15-day NDVI composites,we calculated monthly NDVI data by the maximum-value composi-ting procedure to further minimize effects of cloud contamination,varying solar zenith angles and sur-face topography(Holben1986).NDVI values range from?1.0to1.0,where increasing positive values indicate increasing green vegetation.Negative values indicate non-vegetated features such as water,barren areas,ice,snow or clouds.

Other data

The1-km gridded monthly precipitation and temper-ature data from1989to2003were acquired from the Chinese Natural Sources Database(CNSD,http:// https://www.sodocs.net/doc/e212175063.html,).They were generated by spatial interpolations using the measured data from 680meteorological stations in China.The images of the Qaidam Area were clipped and used as the basic climatic data.Sunshine durations were simply aver-aged by the meteorological records inside the Qaidam Area to reflect the mean status of the corresponding period.A vegetation map of the Qaidam Area was also obtained from the CNSD.This vegetation map contains flora distributions that were used as ancillary information in the land-cover classifications.The climatic and vegetation maps were both originally generated by Institute of Geographic Sciences and Natural Resources Research,Chinese Academy of Sciences.The1:250,000topographic maps(mapped by State Bureau of Surveying and Mapping,China in 1999)were processed in the GIS platform to generate a digital elevation model that contained topographic and elevation information.

Land-cover classification

According to field observations and the vegetation map,we can divide the Qaidam Area mainly into four land-cover classes,namely barrenland(B.),shrub-desert(D.),shrubland(S.),and grassland(G.).These specific classification units were adapted from IGBP land cover scheme and are directly related to surface vegetation(Belward1996).Barrenland is defined as land that never has more than5%vegetated cover during growing season.It is often covered by exposed soil,sand,rock,water,or snow and ice.Shrub-desert is defined as land with shrub canopy cover within5–20%and with none or very sparse herbaceous vegetation.Shrubland is defined as land with shrub canopy cover above20%.Grassland refers to land dominated by herbaceous vegetations and with shrub cover less than20%.Grassland often has a high vegetation cover and mainly establishes in areas with good moisture condition,particularly in the oases and high elevations.These four land-cover classes are different in combination of vegetation cover,spatial distribution,and phenological characteristics.This makes it possible to classify land-cover by multi-temporal NDVI interpretations.

We performed a similar classification used in gen-erating the IGBP DIS-Cover global1-km land cover product.It can be described as a multi-temporal un-supervised classification of NDVI data with post-classification refinement using ancillary data(Loveland and Belward1997).The processes and theoretical basis can be found in detail in Loveland et al.(2000). The classified object in this paper was restricted only to the Qaidam Area,which is far smaller than the extent of the DIS-Cover product.Consequently,a better accuracy can be expected owing to involving less classification units and complexity.

NDVI variations in growing season include most of phenological information in a year,so we only utilized the monthly NDVI from May through September in the classifications.This operation can reduce the overall data volume and remove the noise coming from snow and ice in the winter.The months of May through September were experientially select-ed to represent the average start and end of a growing season.

The preliminary results by clustering may represent multiple disparate classes.Therefore,some heteroge-neous clusters need to be split into more homogenous patches using post-classification refinement.The ancillary data we used in this operation were the digital elevation model and vegetation map stated in “Other data.”Then the unsplit clusters and the new split patches were manually checked and labeled based on their comprehensive attributes,such as NDVI,seasonality,elevation,vegetation community and spatial adjacent relation.However,shrubland and shrub-desert have very similar phenology,so the NDVI value and the spatial adjacent relation were experientially emphasized in their identifications.In addition,woodland and cropland that rarely exist in the study area were regrouped into the surrounding patches in the refinement process.

Land-cover class is suggested to be stable and consistent within neighboring years under a natural condition.Thus,multi-seasonal data with a prolonged time span of classification period were expected to reduce errors and remove the high-frequency vari-ability so as to highlight the decadal trend.We sep-arately selected three phases of growing seasons for land-cover classifications of the early1990s(years 1992,1993and1995)and the early2000s(years2001, 2002and2003).The main consideration of choosing these years is image quality.Year1994was skipped because of NOAA-11malfunction and the failure of NOAA-13shortly after launch(see http://www2.ncdc. https://www.sodocs.net/doc/e212175063.html,/docs/klm/).Years1992,1993and1995 locate in a relatively cold period while years2001, 2002and2003representing a warming one.Both periods cover a wet year,a dry year and an average year compared to the mean precipitation amount.

The classification of each growing season was performed independently.If the preliminary results of an areal pixel are all different in the three maps of a same period,the authors re-consult reference materi-als in order to give a final decision.But the account of these pixels occupies only about1percent of the total. Where the preliminary results are completely or2/3 consistent,the final decision is given as the majority.In this way,we generated the Qaidam land-cover maps of the early1990s and the early2000s,separately.

In addition,we also produced a land-cover map using the1-km A VHRR data of year2001.The results were compared with those generated from the contemporary MODIS data to examine the impacts of sensor difference on the classification.The two classified maps show a good agreement.The overall consistency was93%with a Kappa coefficient of0.90.

Statistical analyses of land-cover changes

We directly compared the1990s’land-cover map with that of the2000s’,and detected their changes.We focused our statistics only on the lands that were ever vegetated in the two time periods.The land-cover conversions were categorized into two trends:degra-dation and amelioration,according to the general vegetation cover of the four classes(Table1).Then we calculated the area percentages of each land-cover class,each type of land-cover conversion and each dynamic trend.In the statistics,we separately ana-lyzed the oasis and non-oasis regions,due to their significant differences in hydrological condition.

In the oases,surface water is mainly provided by the streams and springs sitting in the central part. Thus along the perpendicular direction of the oasis boundaries form gradients of moisture availability, which greatly affect vegetation growth and distribu-tion there.To observe the diversity of land-cover changes along these gradients,we calculated the distances of the degraded and ameliorated pixels to the oasis boundary.A cumulative plot by distance was generated to present the spatial patterns of the degraded and ameliorated lands.The distances were classified every500m,ranging from500to25,000m.

Similarly,we classified the non-oasis region into vertical sections of every100m,ranging from2,700to 5,100m.Statistical analyses were performed separate-ly in each elevation section and their results were compared to examine the vertical diversity in land-Table1Degradation and amelioration processes

Trends Processes

Degradation Grassland to shrub-desert

Grassland to Barrenland

Shrubland to shrub-desert

Shrubland to Barrenland

Shrub-desert to Barrenland

Change between Grassland

and shrubland and NDVI

obviously decreased Amelioration Reverse processes to the

items above

cover changes instead of the horizontal.In the mountain areas above3,500m,we also categorized pixels by their surface slope and calculated the area percentage of degraded lands to the total in each slope degree.

Estimates of correlation between NDVI and climatic factors

The growing-season NDVI values during the14years (1990–2003)were linearly regressed to sunshine duration,precipitation and temperature integrated over different time intervals.Growing-season NDVI, in this paper,is defined as the summation of the monthly NDVI values from May through September in a year.As a coarse approximation of biomass, growing-season NDVI proves sensitive in coupling vegetative variation and climate change(Pettorelli et al.2005).Many researches used NDVI data in monitoring arid and semi-arid vegetation and demon-strated the relationship between NDVI and climate variations in a wide range of arid lands on seasonal to interannual time scales(Nezlin et al.2005;Shi et al. 2003;Weiss et al.2004;Yu et al.2003,2004). However,it is believed that the ecological meaning of NDVI change is ambiguous in non-vegetated areas. Thus we only utilized the NDVI values of vegetated lands to represent vegetative condition in the corre-sponding areas.That is,we only selected the pixels having a14-year monthly average NDVI greater than 0.1and within3δ(standard deviation)of the monthly average.This guarantees that non-vegetation-related variations are not included in the analyses(Myneni et al.1997).In addition,calibration based on invariant desert targets has been applied to the NDVI dataset and insures a slope with respect to time of0.00in barren surface(Pinzon2002;Tucker et al.2005).

The growing-season NDVI values were calculated from the processed8-km GIMMS NDVI sequence, and then spatially averaged over the stable,degraded and ameliorated lands,respectively.Similarly,the temperature and precipitation series of the corre-sponding areas were calculated using the1-km gridded climatic data.With the consideration of the study scale, we only used the large-area image patches and discarded the independent pixels.For the oases,we only analyzed the NDVI variations of the stable vegetated lands because the coarse resolution of NDVI image was not enough to separate the fragmentary amelioration patches from the degradation ones.

Then we performed correlation analyses between growing-season NDVI and climatic factors of a series of time intervals to evaluate during what period climatic factor(s)significantly influence vegetation growth in a given landscape(Wang et al.2003).The time interval is controlled by altering the combination of duration and lag time in unit of a month.It involves not only the concurrent growing season(May to September),but also the preseason months extending back to the end of the preceding growing season. Results and discussions

Land-cover changes under climate warming

In the oases

As shown in Fig.3a,land degradations are more extensive than ameliorations in the oases.The

most

intensive process that40.1%of shrub-desert retro-gresses into barrenland,results in the great shrinkage of shrub-desert and drastic expansion of barrenland during1990to2003(Fig.3b).In contrast,16.7and 1.2%of shrub-deserts locally evolve into shrubland and grassland,respectively.However,these positive land-cover conversions are relatively weak when compared to the negative ones(Fig.3a).Although shrubland area keeps balanced in the land-cover conversions,it doesn’t mean that shrubland really remain stable in the oases under the warming climates.In fact,24.3and13.0%of shrublands, respectively,retrogress into shrub-desert and barren-land while rewarding compensation from grassland and shrub-desert in some other regions.But,corre-spondingly,the area percentages of grassland and shrub-desert reduce.As a result of all the processes stated above,the total vegetated area in the oases decreases by17.0%during this period(Fig.3b).

Furthermore,the degradation and amelioration processes are not homogeneously distributed in the oases.The degradations concentrate near the oasis boundaries(within5km)and the ameliorations centralize between the degraded zone and the stable oasis-core(Fig.4).That is,the lands within the marginal areas tend more towards degradation than other oasis areas.

In the non-oasis region

The non-oasis region is believed to better reflect zonality,because they are less affected by local hydrological condition than the oases.As shown in Fig.5,there are some differences in the land-cover conversions and outcomes in the non-oasis region when compared to those in the oases.For example, there are less grassland and shrubland directly turning into barrenland and1%of barrenland locally recov-ered by shrub-desert(Fig.5a).Especially,the total area percentage of shrubland has increased by4% owing to22.6%of grassland and15.4%of shrub-desert turning into this class(Fig.5b).However,the overall trend and main land-cover conversions are consistent.In detail,grassland and shrub-desert also lose their area percentages by3and7%(Fig.5b), respectively,which are separately equal to17and 31%of their original amounts.Shrub-desert retro-gressing into barrenland still presents the dominant conversion.As a result,the degradation processes are still more intensive than the ameliorations at the basin scale(Fig.5a),as far as their involved area is concerned.

According to zonal statistics,the dynamic scenar-ios change with elevation in the non-oasis region.The conversion rates of the main land-covers show great differences between the alpine and the relatively lower-elevation areas(Fig.6a).At lower elevations, shrubland and shrub-desert,respectively,widely retrogress into shrub-desert and barrenland.But the situation is reversed in the alpine areas(Fig.6a). These lead to the drastic shrinkage of shrub-desert at lower elevations and expansion of shrubland at higher elevations(Fig.6b).

Synthetically,Fig.7shows that degradation inten-sity generally decreases with elevation(R2=0.619, P<0.0001)while amelioration intensity significantly increases,as far as area percentage is concerned. Especially,the linear relation is very strong between elevation and the area percentage of amelioration (R2=0.90,P<0.0001).Although the degradations are stronger than ameliorations at lower elevations,the latter are enhanced with elevation rising and become the dominant trend in the alpine mountains.

In the locally populated area

The Qaidam Basin has an average population density of only1.5/km2according to the national census in year2000.Most people,however,centralize in several towns.Thus,there locally formed a few populated areas where anthropogenic influence on vegetation might be more important than other regions.

To

examine their differences in land-cover change,we analyzed in particular three major towns and their surrounding areas within 50km distance,namely Ge ’ermu,Delingha,and Dulan (see Fig.1).The conversion trends were shown in Table 2.

Between Ge ’ermu and Delingha,the intensities of the dynamic trends are almost identical.The degrada-tions appear stronger than the ameliorations.The main process is still shrub-desert retrogressing into barren-land,as well as that on the regional scale (Figs.3and 5).In Dulan,the ameliorations are very limited when compared to those in other regions,while the degradations remain dominant.The main conversion process becomes 35%of shrubland retrogressing into shrub-desert.In all the three areas,however,vegetated

lands reduced the same as those in the non-populated regions.The results stated above indicate that anthro-pogenic impacts may somehow modify the local scenarios of land-cover changes in populated areas but the overall trend keeps unchanged and the spatial extents of these impacts are limited in the Qaidam area.Climate forcing

We emphasized climatic trends in the causal analyses of land-cover changes at regional scale because of few human activities in the Qaidam Area.As shown in Fig.8a,precipitation only fluctuates around the mean value with a big variability but no obvious trend.Similarly,sunshine duration oscillates at even

a

Fig.6a Land-cover conversion rates in each elevation section.The rate is defined as the ratio of the converted area to the original.b Changes in area percentage of each land-cover class at different

elevations

smaller degree,and also has no significant trend during this time period(Fig.8b).By contrast,temperature in the study area increases significantly in the past decade,characterized by a relative cold period from 1990though1993and a persistent warm period after 1998(see Figs.2and8c).This warming occurs almost in each month and peaks in July(Fig.8d).

Temperature,as an important climatic factor,can affect vegetation growth and distribution in the cold arid areas,directly and indirectly.First,temperature is a primary physiological limiting factor for vegetation in cold regions.Rising temperature can directly facilitate floral exuberances there.Second,elevated temperature can also cause a higher surface evapo-transpiration.Increases in temperature of0.5–2.0 degrees centigrade can raise evapotranspiration by 0.2–2.0mm/day in the desert,unless is accompanied by increased rainfall(Mabutt1989;Greco et al. 1994).The warming without more precipitation in arid lands can reduce moisture availability and cause intensified surface drought,which are especially disadvantageous for areas short of water supply. Thus,although there may be other inducing factors, the decadal warming appears important to the land-cover changes in the Qaidam Area.

NDVI variations,trends and their linkages

with climate changes

The growing-season NDVI values of different regions were respectively linear-regressed on the year from 1990to2003to examine the NDVI trend and,by inference,the vegetation trend.The results show that there is no significant NDVI trend in the stable vegetated lands.The r2values are only0.017for the stable vegetated lands in the oasis-regions and0.037 for those in the non-oasis regions(far lower than0.1 significant level).We analyzed,in particular,the NDVI trends in the stable vegetated lands based on land-cover class.The r2values are0.026,0.036,and 0.016,respectively for grassland,shrubland and shrub-desert.By contrast,the NDVI trends appear statistically significant both in the degraded and ameliorated non-oasis regions(Fig.9).However,the NDVI variations of the degraded and ameliorated lands within the oases are not available because they are mostly located in a10-km-wide belt surrounding the oasis core.Although the degradation patches lie closer to the boundary than the ameliorations(Fig.4), they cannot be clearly distinguished by the coarse resolution(8km)of NDVI sequence.

Correlation analyses show that the interannual variations in growing-season NDVI are not signifi-cantly related to those in sunshine duration of any time interval in a year.It indicates that vegetation growth in the Qaidam Area is not influenced by the little fluctuations in sunshine duration.We suggest that this is due to the sufficient sunlight under cloudless weather in this area.

As shown in Table3,the growing-season NDVI values in all lands are primarily related to precipita-tion in May to July during the current growing season.Increased June precipitation appears always a positive signal for vegetation growth in the study area.The results also imply that the interannual vegetation variations are not necessarily linked with the entire annual or growing-season precipitation amount.In fact,they are most correlated to precipi-tation within a relatively short and specific duration in a year.Monthly precipitation amounts of May to July were separately compared year-by-year and no sig-nificant trend was found during this period(Fig.9

). Table2Intensities of the trends in land-cover changes within three locally populated areas

Degradation(%)Amelioration(%)Stable(%) Ge’ermu23.710.865.5 Delingha23.610.665.8 Dulan27.0 1.671.4

This suggests that precipitation may not be responsi-ble to the land-cover conversions that generally result from long-term climatic forcing.

The linkages between growing-season NDVI and temperature are more complex than those between NDVI and precipitation.Increases in monthly temper-atures during growing season appear a positive driving force for the ameliorated lands,in contrast to a negative one for the degraded (Table 3).This differ-ence is suggested to result from the balance between changes in thermal condition and moisture availability under climate warming,both of which can influence vegetation growth.The degraded lands concentrating at oasis boundaries and lower elevations have better thermal condition but less moisture availability,and thus are more depressed by the intensified drought due to rising temperatures.However,the ameliorated lands mostly locate in the alpine areas,where vegetation establishment and growth are more limited by thermal condition for the wetter but colder climates.Mean-while,in the stable lands (mostly at middle elevations),there were no statistical correlations observed between growing-season NDVI and temperature within any time interval.But it does not mean that the temperature increases have no effect on vegetation there.The authors suggest that the active influences by climate warming in these areas may be removed or reduced by the concurrently intensified drought.

In addition,no relationship between NDVI and the preseason climate was observed in any regions.That is,no obvious lag-time effect exists across seasons and vegetation growth in the Qaidam Area is

mainly

Fig.8Climate anomalies during 1990–2003.a Annual and seasonal precipitation anomalies (percentage departures to the 14-year mean).b Annual and seasonal sunshine duration anomalies (percentage departures to the 14-year mean).c Annual and

seasonal temperature anomalies.d Monthly temperature anoma-lies (subtract the averaged monthly values of the former 7years from those of the latter)

controlled by climatic factors of the concurrent growing season.

Spatial variability of climate-warming impacts

on vegetation

Owing to the“bi-directional”effects of increased temperature described above,the climate-warming impacts on natural vegetation vary with combinations of background thermal condition and moisture avail-ability.Lower background values strengthen the significances of temperature increment and warming-caused moisture loss.Thus,the increase in tempera-ture can be important and positive for lands with wetter but colder climates,when depress lands with better thermal condition but less moisture availability.

As a result,vegetation dynamics under climate warming appears greatly different between the verti-cal sections because mountain topography effectively affects the temperature and precipitation patterns that mainly determine the background thermal condition and moisture availability at large scales.Generally, temperature declines with elevation rising but precip-itation may be increased for orographic rain.Conse-quently,the relative moisture increases upwards owing to less evaporation and elevated precipitation. That is,the significance of elevated temperature increases at higher elevations and that of the warm-ing-reduced moisture turns smaller correspondingly. Therefore,higher temperatures facilitate vegetation establishment and recovery in the alpine areas,where the negative impact of climate warming is limited and the positive is much enhanced.This is supported by the good linear-relation between elevation and area per-centage of the ameliorated lands(Fig.7).But lands at lower elevations are more depressed by the

intensified

drought because of a warmer and drier background. Elevated temperatures without more precipitation turns into bad signals for these areas.As a result of the decadal warming,lands drastically degrade there in land-cover class and vegetation cover(see Figs.6, 7and9).

It is noteworthy that degradation and amelioration coexist at many elevations.This indicates that elevation is not the single factor influencing the background thermal and moisture condition,which, in turn,determine the ecological impacts of climate warming.Especially,many factors can affect local hydrology that controls moisture availability.For example,degradation processes may be restricted to some extent by local water supply.As shown in Fig.7,fewer lands degrade at elevations ranging from 2,900to3,200m,where many fluvial fans distribute around the basin interior.Contrarily,the worst situa-tion occurs in the basin interior(2,700–2,900m)and piedmont zone(3,300–3,500m),where water avail-ability is very small.In the mountain areas(above 3,500m),degradation intensity increases with surface slope rising that makes against water conservation (Fig.10).Similarly,more lands in the oases degrade along the oasis boundaries far from surface water.The better hydrological condition seems to make lands resistive to degradation under climate warming,and vice versa.

In summary,climate-warming impacts on vegeta-tion have a great spatial variability in the Qaidam Area,due to the differences in background temper-ature and moisture condition that are mainly con-trolled by elevation and local hydrology.Suffering more from the intensified drought,lands mostly degrades in areas with lower elevation and fewer water supplies.In contrast,they ameliorate in areas with higher elevation and better hydrological condi-tion,where vegetation growth is directly facilitated by elevated temperatures.

In previous studies,it is concluded that the Northern Hemisphere becomes greener especially in the higher latitudes during the past two decades(Myneni et al. 1997;Nemani et al.2003;Tucker et al.2001).After comparing the annual NDVI,Fang et al.(2004)found an increase trend in vegetation activity in Northwest China.However,as shown above,vegetation dynam-ics in the Qaidam Area from1990to2003are not very consistent with these trends.NDVI in about60% of the vegetated lands shows no significant increase under the climate warming(Figs.3and5).Although vegetation growth is locally enhanced at higher ele-vations,their area percentages are limited when compared to the total amount.Shrinkage of the vegetated lands shows an overall desertification under the climate warming in the Qaidam Area and especial-ly in the basin interior,which is also supported by Shi

Table3Maximum values in significant correlation coefficients (P<0.05)between growing-season NDVI and climatic factors of different durations in a year

Lands Temperature Precipitation

Nonoasis-degradation Duration June–

September

June–July R value?0.6560.632

Nonoasis-amelioration Duration July–August June R value0.6270.601

Nonoasis-stable Duration June

R value0.810

Oasis-stable Duration June

R value0.543

Stable grassland Duration May–June

R value0.739

Stable shrubland Duration June

R value0.718

Stable shrub-desert Duration May–June

R value0.676

“Nonoasis-degradation”means the analysis is specified for the

degraded lands in the non-oasis regions.The blank item means

no significant correlation(P<0.05)observed.

Fig.10Degradation intensities in lands with different slope

et al.(2003).In particular,large quantities of shrub-deserts and shrublands at lower elevations are less limited by thermal condition but are extremely lack of water supply.Their growing-season NDVI appears negatively correlated to temperature increment.How-ever,considering the study resolution and the short time-span in this paper,the authors suggest further researches at multiple spatial and temporal scales. Conclusions

Our analyses demonstrate that about40%of the veg-etated lands in the Qaidam Area respond to the climate warming during the period from1990to2003.In detail,lands mostly degrade at lower elevations and oasis boundaries,but ameliorate at higher elevations on regional scale,although land-cover change scenar-ios can be locally modified in the populated areas.The great spatial variability may be caused by the differ-ences in background thermal condition and moisture availability.The trends in growing-season NDVI appear significant in the ameliorated and degraded lands,but are ambiguous in the classified stable lands. Correlation analyses show that the interannual varia-tions of vegetation in the Qaidam Area are positively related to precipitation within a specific duration,rather than the entire annual or seasonal amounts.But precipitation may not be responsible to the land-cover conversions because no decadal trend exists in May to July precipitation amounts,which prove to have highest correlations to growing-season NDVI.Moreover,the interannual relationship between NDVI and sunshine duration was not examined.By contrast,the concurrent-seasonal temperature increases seem responsible to the growing-season NDVI changes in the degraded and ameliorated lands.However,impacts of climate warm-ing on the classified stable lands are profound and no statistically significant correlations are there observed at interannual scale between growing-season NDVI and temperature within any time interval.Overall,our results show that climate-warming impacts on vegeta-tion in the Qaidam Area are spatially heterogeneous and may not coincide with the general“greening”trend at global or continental scales.

Acknowledgement This research is supported by Innovation Team Project(No.40421001)of National Natural Science Foun-dation,and Outstanding Young Teacher Project(No.20022031)of Ministry of Education and Project(No.2001626)of Ministry of Finance in China.The authors would like to thank the MODIS Land Discipline Group and National Center for Earth Resources Observation&Science by the U.S.Geological Survey for kindly sharing the NDVI data.We also acknowledge Institute of Geographic Sciences and Natural Resources Research,Chinese Academy of Sciences for providing the climatic and vegetation data.The authors are grateful to Dr.Yong-Tao Yu for assistance in language and thank Dr.Wan-Qin Guo and two anonymous reviewers for their constructive suggestions.

References

Belward,A.S.(1996).The IGBP-DIS global1km land cover data set(DISCover):Proposal and implementation plans.

IGBP-DIS Working Paper,IGBP-DIS Office,Meteo-France,42Av.G.Coriolis,F-31057Toulouse,France. Braswell,H.,Schimel,S.,Linder,E.,&Moore III,B.(1997).

The response of global terrestrial ecosystems to interan-nual temperature variability.Science,278,868–872. Climate Change(2001).Third assessment report of the intergovernmental panel on climate change IPCC,WG I (2001).Cambridge:Cambridge University Press. Coppin,P.,Jonckheere,I.,Nackaerts,K.,Muys,B.,&Lambin,E.

(2004).Digital change detection methods in ecosystem monitoring:A review.International Journal of Remote Sensing,25,1565–1596.

Curran,P.J.,&Williamson,H.D.(1986).Sample size for ground and remotely sensed data.Remote Sensing of Environment,20,31–41.

DeFries,R.,Hansen,M.,&Townshend,J.(1995).Global discrimination of land cover types from metrics derived from A VHRR pathfinder data.Remote Sensing of Envi-ronment,54,209–222.

DeFries,R.S.,&Townshend,J.R.G.(1994).NDVI-derived land cover classifications at a global scale.International Journal of Remote Sensing,15,3567–3586. Eidenshink,J.C.,&Faundeen,J.L.(1994).The1km A VHRR global land data set:First stages in implementation.

International Journal of Remote Sensing,15,3443–3462. Fang,J.Y.,Piao,S.L.,He,J.S.,&Ma,W.H.(2004).

Increasing terrestrial vegetation activity in China,1982–1999.Science in China(Series C),47,229–240. Greco,S.,Moss,R.,Viner,D.,&Jenne,R.(1994).Climate scenarios and socioeconomic projections for IPCC WG II assessment(Washington,DC:IPCC,IPCC WMO and UNEP).

Holben, B.N.(1986).Characteristics of maximum-value composite images from temporal A VHRR data.Interna-tional Journal of Remote Sensing,12,1147–1163. Huete,A.R.,Justice,C.O.,&Leeuwen,W.V.(1999).MODIS vegetation index(MOD13),algorithm theoretical basis document version 3.0.https://www.sodocs.net/doc/e212175063.html,/data/ atbd/land_atbd.html.

Hughes,L.(2000).Biological consequences of global warming: Is the signal already.Trends in Ecology and Evolution,15, 56–61.

Kerr,J.T.,&Ostrovsky,M.(2003).From space to species: Ecological applications for remote sensing.Trends in Ecology and Evolution,18,299–305.

Leemans,R.,&Zuidema,G.(1995).Evaluating changes in land cover and their importance for global change.Trends in Ecology and Evolution,10,76–81.

Lioubimtseva,E.,Cole,R.,Adams,J.M.,&Kapustin,G.

(2005).Impacts of climate and land-cover changes in arid lands of Central Asia.Journal of Arid Environments,62, 285–308.

Los,S.O.(1993).Calibration adjustment of the NOAA

A VHRR normalized difference vegetation index without

recourse to channel1and2data.International Journal of Remote Sensing,14,1907–1917.

Loveland,T.R.,&Belward,A.S.(1997).The IGBP-DIS global1km land cover data set,DIS-Cover:First results.

International Journal of Remote Sensing,18,3289–3295. Loveland,T.R.,Reed,B.C.,Brown,J.F.,Ohlen,D.O.,Zhu,J., &Yang,L.,et al.(2000).Development of a global land cover characteristics database and IGBP DIS-Cover from1-km AVHRR Data.International Journal of Remote Sensing,21,1303–1330.

Lucht,W.,Prentice,L. C.,Myneni,R. B.,Sitch,S., Friedlingstein,P.,&Cramer,W.,et al.(2002).Climatic control of the high-latitude vegetation greening trend and Pinatubo effect.Science,296,1687–1689.

Mabutt,J.A.(1989).Impacts of carbon dioxide warming on climate and man in the semiarid tropics.Climatic Change, 15,191–221.

Myneni,R.B.,Keeling,C.D.,Tucker,C.J.,Asrar,G.,&Nemani, R.R.(1997).Increased plant growth in the northern high latitudes from1981to1991.Nature,386,698–702. Nemani,R.R.,Keeling,C.D.,Hashimoto,H.,Jolly,W.M., Piper,S.C.,&Tucker,C.J.,et al.(2003).Climate-driven increases in global terrestrial net primary production from 1982to1999.Science,300,1560–1563.

Nezlin,N.P.,Kostianoy,A.G.,&Li,B.(2005).Inter-annual variability and interaction of remote-sensed vegetation index and atmospheric precipitation in the Aral Sea region.

Journal of Arid Environments,62,677–700. Pettorelli,N.,Vik,J.O.,Mysterud,A.,Gaillard,J.M.,Tucker,

C.J.,&Stenseth,N.C.(2005).Using the satellite-derived

NDVI to assess ecological responses to environmental change.Trends in Ecology and Evolution,20,503–509. Pinzon,J.(2002).Using HHT to successfully uncouple seasonal and interannual components in remotely sensed data.SCI2002Conference Proceedings Jul14–18, Orlando,Florida,SCI International.

Roerink,G.J.,Menenti,M.,Soepboer,W.,&Su,Z.(2003).

Assessment of climate impact on vegetation dynamics by using remote sensing.Physics and Chemistry of the Earth, 28,103–109.

Rouse,J.W.,Haas,R.H.,Schell,J.A.,Deering,D.W.,& Harlan,J.C.(1974).Monitoring the vernal advancement of retrogradation of natural vegetation(p.371).Green-belt,MD:NASA/GSFC(Type III,Final Report). Running,S.W.,&Nemani,R.R.(1988).Relating seasonal patterns of the A VHRR vegetation index to simulated

photosynthesis and transpiration of forest in different climates.Remote Sensing of Environment,24,347–367. Shi,Q.D.,Chen,L.J.,Pan,X.L.,&Lv,G.H.(2003).

Characteristics of vegetation evolution in arid land of western China using remote sensing images from1982to 1990.Resources Science SINICA,25,84–88. Slayback, D.,Pinzon,J.,Los,S.,&Tucker, C.(2003).

Northern Hemisphere photosynthetic trends1982–1999.

Global Change Biology,9,1–15.

Stenseth,N.C.,Mysterud,A.,Ottersen,G.,Hurrell,J.W., Chan,K.S.,&Lima,M.(2002).Ecological effects of climate fluctuations.Science,297,1292–1296. Thripathy,G.K.,Ghosh,T.K.,&Shah,S. D.(1996).

Monitoring of desertification process in Karnataka state of India using multi-temporal remote sensing and ancillary information using GIS.International Journal of Remote Sensing,17,2243–3357.

Tucker,C.J.,Pinzon,J.E.,Brown,M.E.,Slayback,D.,Pak,E.

W.,&Mahoney,R.,et al.(2005).An Extended A VHRR 8-km NDVI Data Set Compatible with MODIS and SPOT Vegetation NDVI Data.International Journal of Remote Sensing,26,4485–4498.

Tucker, C.J.,Slayback, D. A.,Pinzon,J. E.,Los,S.O., Myneni,R.B.,&Taylor,M.G.(2001).Higher northern latitude normalized difference vegetation index and grow-ing season trends from1982to1999.International Journal of Biometeorology,45,184–190.

Vermote,E.,El Saleous,N.,Kaufman,Y.J.,&Dutton,E.

(1997).Data pre-processing:Stratospheric aerosol perturb-ing effect on the remote sensing of vegetation:Correction method for the composite NDVI after the Pinatubo eruption.Remote Sensing Reviews,15,7–21. Vermote, E.,El Salleous,N.Z.,&Justice, C.O.(2002).

Atmospheric correction of MODIS data in the visible to near infrared:first results.Remote Sensing of Environment, 83,97–111.

Walther,G.R.,Post,E.,Convey,P.,Menzel,A.,Parmesank,C., &Beebee,T.J.C.,et al.(2002).Ecological responses to recent climate change.Nature,416,389–395.

Wang,J.,Rich,P.M.,&Price,K.P.(2003).Temporal responses of NDVI to precipitation and temperature in the central Great Plains,USA.International Journal of Remote Sensing,24,2345–2364.

Weiss,J.L.,Gutzlera,D.S.,Allred Coonrodc,J.E.,&Dahm,C.

N.(2004).Long-term vegetation monitoring with NDVI in

a diverse semi-arid setting,central New Mexico,USA.

Journal of Arid Environments,58,249–272.

Yu,F.F.,Price,K.P.,Ellis,J.,&Shi,P.J.(2003).Response of seasonal vegetation development to climatic variations in eastern central Asia.Remote Sensing of Environment,87, 42–54.

Yu,F.,Price,K.P.,Ellis,J.,&Shi,P.(2004).Interannual variations of the grassland boundaries bordering the eastern edges of the Gobi Desert in central Asia.International Journal of Remote Sensing,25,327–346.

相关主题