基于多种高分辨率卫星数据的TRMM降水数据降尺度研究——以内蒙古地区为例

利用内蒙古地区2001~2010年42个站点实测降水数据作为“真值”,采用LOO（Leave-One-Out）交叉验证、多元逐步回归等方法,构建TRMM（Tropical Rainfall Measuring Mission）降水数据与地形及气候等要素之间的多元回归关系,在此基础上,利用回归值+残差值的方法,获得空间分辨率为1 km×1 km的TRMM年降水数据,并对降尺度TRMM数据进行精度检验。研究表明：① TRMM数据可用于区域年降水量估计,且与实测年降水量呈显著线性关系;② 通过建立不同年份、不同空间分辨率TRMM数据与其它遥感数据的多元统计模型,研究发现在中尺度下TRMM与观测年降水数据拟合效果较好,且在空间分辨率为0.50°×0.50°时的拟合效果最好;③ 降尺度分析提高了TRMM数据对研究区降水时空特征的描述能力,确定性系数、标准误差和偏差均有明显改善,表明降尺度算法在将TRMM降水数据空间分辨率提高到1 km×1 km的同时,并能提高降水数据的精度。

Downscaling Analysis of TRMM Precipitation Based on Multiple High-resolution Satellite Data in the Inner Mongolia,China

Precipitation is a significant part of the hydrological cycle and so the investigation of changes in precipitation characteristics is normally the first step in investigating the impact of climate change on water availability. However, high-resolution precipitation datasets are seldom available, which, to some extend, limits our understanding of spatio-temporal patterns in precipitation regies and basin scale hydrology. Since regions around the globe with sufficient precipitation gauge networks that are up to the challenge of this type of research are few and far between, precipitation estimations from space borne sensors are often applied to supplement the information collected from existing low-resolution precipitation gauge networks in near real-time applications. Based on observed daily precipitation data as “real data” from 42 stations covering a period of 2001-2010 in the Inner Mongolia, China, TRMM (Tropical Rainfall Measuring Mission) precipitation data with spatial resolution of 0.25°×0.25° was downscaled based on statistical relations between NDVI, meteorological variables, and DEM using LOO (Leave-One-Out) cross evaluation method, spatial autocorrelation analysis methods. The lag time between NDVI and precipitation changes was also considered in this study. The results indicated that: 1) On annual scale, TRMM data can be used for estimation of annual precipitation amount in the Inner Mongolia Region and linear relations can be identified between annual TRMM and observed precipitation data; 2) The analysis of the effect of lagging on the vegetation response to precipitation indicates that a period of 10 days accurately is the lag time, showing the relatively rapid vegetation response to precipitation. Moran’s I indicates that the vegetation index is spatially correlated and identified spatial heterogeneity in the transitional zones between different types of land use and land cover and the unused land area; 3) Significant relations between TRMM and observed precipitation can be detected at moderate spatial scale with spatial resolution of 0.50°×0.50°; 4) Temperature is found to be an important factor influencing downscaling of TRMM precipitation data due to high sensitivity of NDVI changes to temperature variations in the Inner Mongolia Region. Inner Mongolia is topographically levelling and hence DEM is not the significant factor having impacts on downscaled TRMM data; 5) Downscaled TRMM can well reflect spatial patterns of annual precipitation changes. Less precipitation can be found in west Inner Mongolia and more precipitation in south and southeast Inner Mongolia. This study provides another possibility in evaluation of spatial patterns of precipitation changes, and hence provides and right precipitation dataset for conservation of grassland and also irrigation management in the highly eco-environmentally fragile region.