遥感蒸散模型的时间重建方法研究

Research on Temporal Reconstruction of Evapotranspiration by Using Remote Sensing

本文提出的遥感蒸散模型的时间重建方法考虑了逐日蒸散可能的变化情况,在阻抗-彭曼单层模型的基础上,将图像去云处理与时间重建结合在一起,先使用SEBS能量平衡余项模型获得晴好日的地表阻抗信息,选择了叶面积指数LAI作为反映植被表面阻抗的参量,通过时间序列谐波分析(HANTS)获得逐日逐像元的LAI信息,将晴好日的阻抗信息扩展至待定日,并使用因子函数表达了极端温湿条件对植被叶面阻抗的限制作用。使用中科院禹城站2003年作物季的大型蒸渗仪数据对模型所得逐日蒸散结果进行了验证。在作物生长季,模型结果相对于实测结果表现出了良好的相关性(R2≈0.7),远优于作为对比的蒸发比不变法。受单层模型假设局限,方法在裸地及稀疏植被上的结果还存在着较大的误差。

Temporal reconstruction of evapotranspiration （ET）, which means deriving continuous, complete annual ET from fragmentary satellite measurement, is a problem full of uncertainty in remote sensing application. Traditionally, evaporative fraction （LE/H） is simplified as constant in a short period so that weekly or long term ET could be estimated from a single clear satellite image. In this way, variation of daily ET is often neglected and the amount of ET is hard to compare with those at the same time in another year. The objective of this research is to develop a new reconstruction algorithm to retrieve continuous and actual ET dataset and provide valuable temporal profile for agriculture and ecology application. This algorithm considers both the spatial and temporal discontinuity, and is a combination of SEBS （Surface Energy Balance System） and Penman- Monteith model： SEBS model is used to derived latent heat in clear days, and then surface resistance is inverted from PM equation; Leaf area index （LAI）, is interpolated and smoothed to daily term by using HANTS （Harmonic Analysis of Time Series） method. Then surface resistance of the cloudy days is related to those from neighboring clear days with a function of LAI, minimal air temperature and vapour pressure deficit. Daily ET estimation is compared to lysimeter measurements recorded in Yucheng agriculture site and shows a good correlation coefficient in crop growing season （R^2≈ 0.7）. Model result is not satisfactory on bare and sparse land because of the limitation of the one-layer assumption in PM equation, which requests that an independent component of soil evaporation should be added into the algorithm.