GIS-based Modeling Of The Potential Distribution And The Potential Net Primary Productivity Of Deciduous Broadleaved Forest In The Western Part Of The Chinese Loess Plateau

Due to rapid growth of human population and ever increasing human demands on natural resources,the past century witnessed the most intensive and extensive alterations of carth's environments primarily through human-induced land degradation and associated ecological deterioration.These alterations not only threaten the economic future of human population at local and regional scales through reducing ecosystem productivity,but also adversely impact functional processes of carbon sequestration,energy balance and water cycles at hemispheric and global scales through destroying the natural equilibrium of land-atmosphere interactions.The Chinese Loess Plateau might be easily an extreme example of human-induced land degradation and associated ecological deterioration.During the past several decades,the Chinese government has made many attempts to restore the severely deteriorated ecological conditions,but cases of failure were numerous simply because the attempts were carried out without following the natural laws governing the relationships between vegetation and climate. Therefore,the potential distribution of main vegetation types and ecological characteristics of those potential vegetation have to be modeled as foundation for aforestation.The aims of the study are:1) to establish a feasible technology system for modeling the potential distribution of deciduous broadleaved forest(DBF),the main forest vegetation type in the Loess Plateau,with high resolution,2) to modeling the net primary productivity of the potential vegetation communities,and 3) to modeling the potential distribution of DBF under 12 future climate scenarios in the study area.We believe that the study can effectively facilitate the aforestation planning for this area.In addition,the result may function as a base against which to estimate the historic changes in land covers and land uses,to assess the ecological potentials for carbon sequestration,and to evaluate the climatic significance of land-air interactions.In earlier studies of vegetation distribution modeling,researchers would probably draw a line around the outermost spatial points of occurrence of the vegetation.Modern modeling methods improve on this simple approach by using correlations of occurrences with environmental factors such as climate.However,the usefulness of those modeled results was extremely limited because the spatial resolution is rather low and also because not all of the significant environmental variables were considered.In this study,with the advances of Geographical Information System(GIS) technologies and the availability of high-resolution Digital Elevation Model(DEM),several major environmental variables(e.g.,precipitation,air temperature,potential evapotranspiration,etc.) were spatially depicted with high resolution and employed into modeling of potential distribution of DBF.The main contents of the study are:1) 10 environmental factors influencing the DBF distribution were selected as the predictors,which are spring precipitation,summer precipitation, full-winter precipitation,spring potential evapotranspiration,summer potential evapotranspiration,full-winter potential evapotranspiration,spring precipitation variability, summer precipitation variability,full-winter precipitation variability,and biotemperature.Six methods were employed to spatially interpolate these 10 environmental factors.2) The current distribution of DBF was extracted from the Landsat TM\ETM images using supervised classification.3) Five methods(Border Function(BF),Logistic Regression(LR),Genetic Algorithm for Rule-Set Prediction(GARP),MAXENT and Holdridge Model) were employed to establish the mathematic model delineating the potential distribution of DBF.4) The net primary productivity of the potential DBF was calculated based on Carnegie-Ames-Stanford Approach. 5) The potential distributions of DBF under 12 future climate scenarios were modeled.The main conclusions are as follows:(1) According to root mean square error(RMSE),the optimal spatial interpolation methods were selected for various climate factors to ensure the results reliable.1) Monthly mean air temperature:the optimal spatial interpolation method was co-kriging(CK) for January and December,stepwise multiple linear regression(SMLR) for from February to October,ordinary kriging(OK) for November.The RMSE of temperature interpolation of every month was always under 1.5℃.2) Monthly precipitation:the optimal method was inverse distance weighted(IDW) for January,SMLR for February and December,SMLR combined with interpolated residual(SMLR+IR) for from March to August,universe kriging(UK) for from September to November.The RMSE of precipitation interpolation of every month was always less than 12 mm.(3) Precipitation variability:the optimal method was SMLR for spring and summer,IDW for full-winter.The RMSE was always less than 4%.(2) Based on FAO Penman-Monteith equation,with the temperature layer,relative sunshine duration layer,wind speed layer,actual vapour pressure layer and extraterrestrial radiation layer,the potential evapotranspiration in each raster of the study area was estimated. The result showed that the average potential evapotranspiration of the study area was 814 mm. The index was low in southeastern part and high in northern part.Topographic factors influenced potential evapotranspiration significantly.In addition,we assessed the sensitivity of the calculated potential evapotranspiration to change of temperature,relative sunshine duration, wind speed,actual vapour pressure,and extraterrestrial radiation.The result revealed that the most obvious influence on the potential evapotranspiration was from the error of the extraterrestrial radiation,and then temperature.The least influence was from the error of wind speed.(3) The current distribution area of DBF was extracted from Landsat TM\ETM images. The results showed that the area was 1850 km²,which was 3.73%proportion of the total study area.(4) Comparison of 5 methods showed:1) BF,GARP,MAXENT,HM were relatively accurate,while LR had low accuracy;2) LR tended to underestimate the potential distribution area;3) GARP and MAXENT were used to model the distribution of single species,and in this study,they proved to be suitable for modeling the distribution of vegetation type with high accuracy;4) Holdridge Model was rather briefly and efficient,which was suitable for analyzing the change of vegetation distribution responding to the change of climate.(5) The modeled results showed that the area of potential DBF in the study area was about 5203 km² which was 2.8 times of the area of current DBF.The approximate range of potential DBF was as follows:1) in the west of Maxian mountain,the north border of the potential distribution of DBF crossed through the north part of Jishishan county and Linxia district,and through the middle of Dongxiang county,and through the northeastern edge of Guanghe county. 2) DBF potentially distributed in Maxian Mountain and Xinglong Mountain while not existing in high elevation in Maxian Mountain.3) There was discontinuously distribution of potential DBF in from southern and southeastern Maxian Mountain to Weiyuan city.4) There was potential DBF in the southwest corner of Anding district.5) There was discontinuously distribution of potential DBF in Longxi county,Tongwei county,Gangu county,Qinan county, Zhuanglang county and the south edge of Jingning county.6) There was continuously distribution of potential DBF in the most part of Linxia county,Hezheng county,Kangle county, Weiyuan county,Zhangxian county,Qingshui county and Beidao district.(6) Based on CASA,we estimated the NPP of the potential DBF in our study area.The result showed that the average NPP of the potential DBF was 528.9 gC·m^-2·a^-1 and the NPP of all the potential DBF in this region could reach 7.56×10⁶ tC·a^-1.Compared with results from similar researches,the modeled results here was reliable.