| As an important indicator of ecosystem,vegetation is critical to the functioning of earth's life-support system and services provided by ecosystems.Long-term changes of vegetation influence not only the provisions of ecosystems,but also the properties of earth's surface,the global/regional climate processes,and the circulations of energy and materials(water,carbon,etc.).Therefore,monitoring vegetation dynamics,analyzing the processes of those changes comprehensively,identifying and quantifying the driving factors are essential to understand the overall situation of regional ecosystem and help offices making decisions reasonably to promote regional ecosystem toward to a better condition.An in-depth understanding of any change needs to be based on continuous observations over a decade or even decades.Limitations of time and space of human's understands on earth's surface were broken through by the emergence of remote sensing which could repeat sampling on earth surface.Remote sensing makes it possibility for in-depth analysis of long-term vegetation changes and their mechanism in the global or large-scale areas.The accumulation of remote sensing datasets and ground observations,the extension of timeseries products,and the improvement of the spatial resolution and performances of several wide-field sensors have been provided solid databases for the studies of long-term vegetation changes.However,it also brings new opportunities and challenges to this traditional field.How to extract automatically the accurate and detailed information on vegetation long-term changes,including where,when and durations,how and why vegetation changes took place,and eventually predicting the future trends,are the key issues in this field.In view of those above,MODIS(Moderate Resolution Imaging Spectroradiometer)NDVI(Normalized Difference Vegetation Index)products at 250 m spatial resolution were selected to accurately monitor vegetation dynamics in the Hexi region from 2001 to 2017 by adopting two widely used long-term change detection methods.Then,a framework based on a temporal trajectory fitting algorithm was constructed to detect the detailed processes of vegetation gradual changes comprehensively.Finally,another framework for distinguishing the driving factors of climate varialbes and human activities was also developed in the study to quantify the relationships between vegetation changes and climate variations and to deeply understand the interventions of human activities on vegetation changes in Hexi region.The main contents and conclusions of this study are shown as follow:(1)Two most widely used trend analyses of vegetation long-term changes: linear regression analysis,the combination of Theil-Sen slope estimation and Mann-Kendall test were compared in this study to determine their applicability on regional scale.Our results indicated that results of the two methods were highly consistent in both spatial pattern and amount(slope value)of vegetation changes;newly reclaimed oases where the processes of vegetation changes during the observation period were non-linear,were the regions where larger differences in slope values were detected by the two methods.Simple linear regression was completely infeasible in those places,which may be the reasons for the quite differences in the slope values detected by the two methods,indicating that the combination of Theil-Sen slope estimation and Mann-Kendall test is more robust.In addition,most(about 60 %)of vegetation covered areas(NDVImax ? 0.2)in the Hexi region experienced significant changes during the period 2001–2017 and vegetation improvements were dominant;vegetation improvements mainly located in the Qilian mountains,the transition zones between oases and deserts,and the downstream areas of rivers;moreover,vegetation in the Qilian mountains improved slightly while vegetation in the edge of oases improved greatly;areas with a declining trend in NDVI timeseries mainly located in interiors of oases,such as Liangzhou,Minqin,Ganzhou,Gaotai and Guazhou oases.(2)A novel framework based on the temporal trajectory fitting algorithm was proposed to comprehensively understand the processes of vegetation gradual changes.It was a flexible method which generalized monotonous processes of vegetation gradual changes during the observation period into five patterns: linear pattern,exponential pattern,logarithmic pattern,logistic pattern and no-change pattern.All the first four patterns were further binned into two types,consisting of either positive or negative trend.Moreover,a logistic model with four parameters was adopted in the framework to simulate and distinguish all the three nonlinear patterns automatically.Unlike the widely used trend analyses which hypothesizes that ecosystem/vegetation always change linearly in a direction,the new framework could detect the non-linear change patterns or the concealed short-term trends in long-term timeseries,determining where,when,and how vegetation changes took place,and predicting vegetation changes in future according to the temporal trajectories of NDVI timeseries.The detected patterns and time of vegetation changes were helpful to explore the human activities driving vegetation dynamicss.Finally,results of the point and regional verifications in Shiyang River Basin indicated the accuracy and efficiency of the new framework.(3)Spatial pattern of vegetation long-term changes detected by the new framework was highly consistent with that of the two widely used trend analyses.However,more areas with significant changes in vegetation were detected by the new framework because NDVI timeseries were smoothed using a moving average algorithm which made it easy to pass the statistical significance test.In addition,our results showed that: about 60% of vegetation covered areas in the Hexi region experienced nonlinear or periodic change processes,in which the proportion of logical pattern was largest(30.28%),followed by linear pattern,the proportion of exponential pattern and logarithmic pattern were nearly same;increasing types were dominant in all patterns;logical decreasing pattern was dominant in all NDVI decreasing trends and its proportion was greater than the sum of the other three decreasing patterns.(4)Timing of vegetation changes varied among different regions.In general,vegetation conditions in the Heihe river basin started to change earlier(in the early stage of the study period)while that in the Shiyang river basin was the later,and the Shule river basin was the latest.Specifically,vegetation improvements in the upper areas of the Qilian mountains and the oases in the Heihe river basin started earlier(before 2005),while that at the east and west parts of the Qilian mountains started later(around 2010),and the timing of vegetation improvements at junction zones of the Shiyang river basin and Heihe river basin was the latest(around 2013).In oases of Hexi corridor,vegetation conditions in original oases began to improve at an earlier year(before 2005)while those were later in the marginal areas of oases(around 2007).The timing that NDVI timeseries began to decline were significantly different in different administrative units.Specifically,the timing that NDVIs began to decline in Wuwei Prefecture ranged from 2005 to 2009 while the timing that NDVIs began to decline in Jiuquan Prefecture at the years after 2010;NDVI began decline in Zhangye Prefecture took place at three periods: 2003–2005,2008–2010 and 2013.According to the patterns and the timings of vegetation changes detected by the new framework,the factors driving vegetation changes were identified by various ways,such as high-spatial images from Google Earth,field investigations and relevant reports.We concluded that urbanization,reduction of agricultural oasis,adjustment of agricultural planting structure(including constructions of greenhouses),mining were the main human activities for the decreasing trends in NDVI timeseries;expansion of agricultural oasis,improvements of agricultural productions and varies ecological restoration projects were the human factors for improvements in vegetation in the Hexi region.(5)In order to separate all factors driving vegetation dynamics,a second framework was proposed by adopting a decision tree algorithm,which combined the trend analysis(combination of Theil-Sen slope estimation and Mann-Kendall),Pearson correlation analysis between NDVI and climatic factors(temperature and precipitation),the residual trend approach.The framework quantified the individual and conjoint effects of either climate factor and human activities on vegetation dynamics.The geoprocessing tools in the software ArcMap were then adopted to the category of human activities.Finally,all driving factors in the Hexi region were separated,mapped and quantified.The results showed that significant changes in vegetation caused by human activities,precipitation,the interaction of precipitation and human activities,temperature,the interaction of temperature and human activities,the interaction of temperature and precipitation,and the interaction of the three factors accounted for 48.07 %,11 %,25.11 %,3.59 %,2.43 %,3.07 % and 6.73 % of the total changed areas respectively.Obviously,human activity was the most important factor affecting vegetation changes in the Hexi region,especially in the Shiyang river basin where interferences from human activities on ecosystem were strongest.Specifically,the percentage of vegetation improvements caused by ecological restoration policies,agricultural oasis expansion,and changes in agricultural practices were 7.26 %,8.79 %,and 30.33 %,respectively.Generally,ecological restoration policies contributed greatly to the improvements of vegetation.In addition,72.05 % of the decreasing trends in NDVI timeseries were caused by changes in agricultural practices(49.68 %),agricultural oasis shrinkage due to ecological restorations(5.43 %),and urbanization(16.94 %).Those should be treated as regional eco-social adjustments for developments and optimization of the ecological environment,rather than land degradations. |
PDF Full Text Request