基于NDVI优化选择的土壤水分数据同化

Soil moisture data assimilation based on NDVI optimization

时间序列上遥感观测数据的准确性会对同化结果有较大的影响。该文以宁夏回族自治区固原市为例,通过北方生产力生态模型模拟2008年5-7月逐日的土壤湿度,按照不同日期的归一化植被指数(normalized difference vegetation index,NDVI)阈值,分别利用MODIS资料计算出基于红光和近红外波段的垂直含水量指数、改进的垂直干旱指数和基于近红外波段和短波红外波段的短波红外垂直水分胁迫指数,和宁夏中南部的气象站实测土壤水分建立关系,并用不同遥感指数反演的土壤水分作为观测值进行同化。结果表明,在作物的不同生长时期,垂直含水量指数、改进的垂直干旱指数和短波红外垂直水分胁迫指数的反演效果不同,基于NDVI优化遥感反演结果,选择准确性更高的反演结果作为同化观测值,能提高同化土壤水分的精度。该研究表明在不同时间段内使用更为准确的遥感监测结果作为观测值进行同化可以提高同化的精度。

Soil moisture could be better estimated through assimilation at various observations into ecosystem models to use all sources of information well. However, different assimilation results could be got from different observations. And the results might have big differences. Three different remote sensed surface soil moisture derived from MODIS red、 near infrared and shortwave infrared bands was assimilated to initialize the soil moisture in BEPS（boreal ecosystem production simulator） in May to July in 2008, taking Guguan city in Ningxia as a case. Three different drought indexes PDI （perpendicular drought index）、SPSI （shortwave infrared perpendicular water stress index）and MPDI （modify perpendicular drought index）to derive the surface soil moisture were chose based on the time-series NDVI （normalized difference vegetation index） threshold. An Ensemble Kalman Filter was used to perform the data assimilation. The in-situ sites’ observations were used to verify the assimilation results which were got from three different remote sensed results. It was demonstrated that the method of remote sensed soil moisture assimilation could help to improve the results in BEPS model. And the assimilation using accurate remote sensed result as observation in different time series could help to improve the soil moisture results.