地表温度合成方式对TVDI预测精度影响

Effect of LST Synthesis Methods on Prediction Accuracy of TVDI

针对目前干旱指数(temperature vegetation dryness index,TVDI)构建过程中地表温度(land surface temperature,LST)合成方式的不确定性问题,探讨了LST合成方式对TVDI预测精度的影响,提出了预测效果较好的LST合成方式。以巢湖流域为研究区,利用2013年6月的MODIS LST数据和归一化植被指数数据,构建TVDI预测模型,并结合降水数据对该模型预测结果进行定量验证。从不同时间尺度上(旬、月),探讨LST合成方式对于TVDI预测精度的影响。研究结果表明:(1)LST平均值合成构建的干旱指数TVDI与降水距指数(precipitation anomaly,PA)相关性在不同时间尺度上表现出现较大差异,上旬和中旬TVDI与PA均不存在相关性,下旬和全月TVDI与PA存在显著负相关性(p<0.01),相关系数分别为-0.31和-0.34;(2)LST最大值合成构建的干旱指数TVDI与PA在不同时间尺度上均存在显著负相关性(p<0.01或p<0.05),上旬、中旬、下旬及全月相关系数分别为-0.29、-0.25、-0.31、-0.41。综合分析发现,在月尺度上采用LST最大值合成方式构建TVDI指数对干旱预测效果更好。

In view of the uncertainty of the land surface temperature (LST) synthesis methods in the temperature vegetation dryness index (TVDI) construction process,this paper discusses the influence of LST synthesis methods on TVDI prediction accuracy,and proposes a better LST synthesis method.Taking Chaohu Lake Basin as the research area,the TVDI prediction model is constructed based on MODIS LST data and NDVI data from June 2013,and the model accuracy is quantitatively verified with precipitation data.The effect of LST synthesis methods on the accuracy of TVDI in the drought prediction is investigated on different time scales (10 days and the whole month).The results show that the correlation between TVDI constructed by LST mean value synthesis and precipitation anomaly (PA) is significantly different on different time scales.There is no correlation between TVDI and PA in the first 10 days and the middle 10 days,there is a significant negative correlation between TVDI and PA in late 10 days and the whole month ( p <0.01),and the correlation coefficients are -0.31 and -0.34,respectively.There is a significant negative correlation between TVDI constructed by LST maximum value synthesis and PA on different time scales ( p <0.01 or p <0.05),and the correlation coefficients of the first 10 days,the middle 10 days,the late 10 days and the whole months are -0.29,-0.25,-0.31 and -0.41,respectively.The comprehensive analysis shows that the TVDI constructed by LST maximum value synthesis on the monthly scale is better for drought prediction.