Vegetation Distribution Dynamics And Modeling Study In The Upper Heihe River Basin

Under the background of global environmental changes, water use and management is critical for the integrated management and sustainable development in the arid area. The present study examined the spatial and temporal variations of LAI for five ecosystem types and simulated ecosystem types by using CART static vegetation model within a watershed with a complex topography in the Upper Heihe River Basin,a major inland river in the arid and semi-arid western China. We integrated remote sensing-based GLASS (Global LAnd Surface Satellite) LAI products, interpolated climate data, watershed characteristics, and land management records for the period of 2001-2012. We determined the relationships among LAI,topography, air temperature and precipitation, and grazing history by five ecosystem types (Alpine Meadow, Sparse Alpine Vegetation, Shrub, Alpine Steppe and Coniferous Forest) using several advanced statistical methods. The fluctuations of observed LAI varied greatly across the watershed and were classified as 'Improved'(Z>1.28), 'Stabilized' (-1.28?Z?1.28), and 'Degraded' (Z<-1.28) areas.The main conclusions are as follows:(1) Growing seasonal temperature is high related to elevation in the Upper Heihe River Basin. High temperature in eastern and northern area with low elevation but low temperature in western area with high elevation. Growing seasonal precipitation in the central area lower than in other area. Closer to outlet,less precipitation. From 2001 to 2012,growing seasonal temperature is increasing significant. Growing seasonal precipitation in western area is significant increasing as well,but in eastern area is not significant changing even showed decreasing trend.(2) We show that long-term mean LAI distribution had an obvious vertical pattern as controlled by precipitation and temperature in a hilly watershed. Overall,watershed-wide mean LAI had an increasing trend overtime for all ecosystem types during 2001-2012, presumably as a result of global warming and a wetting climate.Low temperature limits vegetation growth,high temperature promotes water stress.The most suitable for the growth of vegetation at the altitude of 2700m-3900m, the growing seasonal temperature of 4-11 ?, and precipitation of 300-580mm.Coniferous Forest had the largest improvement, and the highest degree of Sparse Alpine Vegetation degradation. The growing seasonal LAI fluctuated greatly,especially in high altitude glaciers, Sparse Alpine Vegetation and Coniferous Forest growing area and Desert. High dispersion degree indicated degradation of snowline,variation of timberline and desert ecosystem restoration had great influence on the interannual variation of growing seasonal LAI under the background of climate warming.(3) The LAI increasing trend was significantly affected by warming in"Improved" area, but precipitation decreasing trend in "Degraded" area limited vegetation growth. The relationships between LAI and temperature and precipitation were obviously different at spatial scale. Extreme weathers such as cold spells and droughts could substantially affect inter-annual variability of LAI dynamics. We found that climate was not the only driver for temporal vegetation changes for all land cover types. Grazing partially contributed to the decline of LAI in "Degraded"areas and masked the positive climate warming effects in other areas. We concluded that temporal and spatial LAI dynamics were affected by both climate variations and human disturbances in the study basin. The vegetation simulation accuracy for grazing areas was not as high as for areas with less human activity.(4) Based on the status of 1:10 million vegetation distribution map, we screened and confirmed the threshold distribution of key factor like temperature, precipitation,topography and soil,corresponding to different vegetation types to obtain different types of vegetation classification criteria and construction and optimization of static vegetation model by using CART method. The optimal classification tree is obtained by CART model with 52 nodes. The test accuracy of the model is 63.57%, and the simulation accuracy is about 64.99%. Vegetation affected by grazing lightly like Sparse Alpine Vegetation, was simulated more accurate than grazing vegetation types like Shrub, Alpine Meadow and Steppe. Future monitoring studies should focus on the functional interactions among vegetation dynamics, climate variations, and human disturbances.