中文摘要：

基于GIMMS 3g、PAL、LTDR V3、FASIR及MODIS 5种不同的遥感影像数据,及洞庭湖流域30个气象站点的气温、降水和日照时数月值数据,采用逐像元一元线性回归模型,对比分析了在不同数据集背景下的植被变化趋势,并基于时间序列长度以及数据精度的考虑,选择GIMMS 3g作为研究洞庭湖流域植被覆盖变化的基础数据集,进而利用相关系数和多元线性回归分析模型探讨流域植被覆盖变化与降水、气温、日照时数的关系。结果表明：（1）过去29a间,流域GIMMS 3g NDVI在时空尺度上均以增加趋势为主,1998~2000年出现最大的降低,以植被覆盖较好的山区减少最快;（2）去除年际变化趋势和季节性影响的NDVI与同期降水量距平、累积3个月气温距平值及同期日照时数距平值相关性程度最高;（3）统计意义上,降水量、气温和日照时数解释洞庭湖流域植被月NDVI变化的37%,日照时数对该流域NDVI变化的影响最大,其次为降水和气温;（4）在时间和空间范畴,生长季NDVI可以作为反映洞庭湖流域森林覆盖率变化的指标。

英文摘要：

Trend and interannual variability of growing season vegetation cover were examined for the Dongting Lake Watershed over the period of 1982--2010, using GIMMS 3g together with other AVHRR NDVI products and MODIS vegetation index products. The four datasets of AVHRR showed a great similarity in spatial distribution of NDVI trends from 1982 to 1999, while comparison of MODIS EVI/ NDVI and GIMMS 3 g during 2000--2010 showed a significant difference in NDVI trends between MODIS NDVI and GIMMS 3g. This is likely owing to MODIS NDVI saturation in the study area. Spatial distribution of trends resulting from GIMMS 3 g and MODIS EVI had some consistency. Thus GIMMS 3 g was selected for detailed analysis because of its long temporal coverage and consistency with other products. Possible influencing factors （precipitation, temperature, sunshine hours and vegetation cover） leading to interannual variations of growing season NDVI were explored using multiple linear regression and correlation analysis. The results indicate that an increasing NDVI trend occurred in the study area during the past 29 years, only 9.5% pixels in the Dongting Lake Watershed showed some NDVI reduction. The eastern, western and southern parts of the region covered by mountain forests experienced a large decreased in NDVI during 1998--2000. De-trended NDVI monthly anomaly was significantly correlated with de-trended concurrent precipitation anomaly, concurrent sunshine duration and cumulative threemonth temperature anomaly. The three climate factors explained 37% temporal variability of de-trended growing season monthly NDVI of the study watershed, while these factors failed to explain the overall NDVI increase during 1982--2010. This suggests that the increase in NDVI is more likely associated with an increase in vegetation cover over the period likely owing to the improvement of social-economic condition in the study area. This is supported by good correlation between the NDVI trend and the government reported forest cover change, and consistent