A statistical spatial downscaling algorithm of TRMM precipitation based on NDVI and DEM in the Qaidam Basin of China

• Published in: Remote sensing of environment Vol. 115; no. 12; pp. 3069 - 3079
• Main Authors: Jia, Shaofeng, Zhu, Wenbin, Lű, Aifeng, Yan, Tingting
• Format: Journal Article
• Language: English
• Published: New York, NY Elsevier Inc 15.12.2011; Elsevier
• Subjects: Algorithms; Animal, plant and microbial ecology; Applied geophysics; Basins; Biological and medical sciences; China; data collection; DEM; digital elevation models; Discrete element method; Downscaling; Earth sciences; Earth, ocean, space; ecology; environmental factors; Exact sciences and technology; Fundamental and applied biological sciences. Psychology; General aspects. Techniques; hydrology; Internal geophysics; linear models; meteorology; NDVI; normalized difference vegetation index; Precipitation; Precipitation (meteorological); Qaidam Basin; radar; rain; rain gauges; regression analysis; Remote sensing; satellites; Stations; Teledetection and vegetation maps; topography; TRMM; Vegetation; Qinghai China; Far East; Qaidam Basin; Asia; China; China, People's Rep., Qinghai Prov., Qaidam Basin; China, People's Rep; TRMM; Precipitation; Downscaling; Qaidam Basin; NDVI; DEM; algorithms; data; Vegetation index; Scale reduction; springs; tropical zone; Environmental factor; vegetation; global; Field; precipitation; ecology; Relation; spatial resolution; spatial variations; high resolution; Availability; satellites; detection; hydrology; meteorology; topography; Statistical character; remote sensing; technology; quality
• ISSN: 0034-4257; 1879-0704
• DOI: 10.1016/j.rse.2011.06.009

The availability of precipitation data with high spatial resolution is of fundamental importance in several applications such as hydrology, meteorology and ecology. At present, there are mainly two sources of precipitation estimates: raingauge stations and remote sensing technology. However, a large number of studies demonstrated that traditional point measurements based on raingauge stations cannot reflect the spatial variation of precipitation effectively, especially in ungauged basins. The technology of remote sensing has greatly improved the quality of precipitation observations and produced reasonably high resolution gridded precipitation fields. These products, derived from satellites, have been widely used in various parts of the world. However, when applied to local basins and regions, the spatial resolution of these products is too coarse. In this paper, we present a statistical downscaling algorithm based on the relationships between precipitation and other environmental factors in the Qaidam Basin such as topography and vegetation, which was developed for downscaling the spatial precipitation fields of these remote sensing products. This algorithm is demonstrated with the Tropical Rainfall Measuring Mission (TRMM) 3B43 dataset, the Digital Elevation Model (DEM) from the Shuttle Radar Topography Mission (SRTM) and SPOT VEGETATION. The statistical relationship among precipitation, DEM and Normalized Difference Vegetation Index (NDVI), which is a proxy for vegetation, is variable at different scales; therefore, a multiple linear regression model was established under four different scales (0.25°, 0.50°, 0.75° and 1.00°, respectively). By applying a downscaling methodology, TRMM 3B43 0.25° × 0.25° precipitation fields were downscaled to 1 × 1 km pixel precipitation for each year from 1999 to 2009. On the basis of three criteria, these four downscaled results were compared with each other and the regression model established at the resolution of 0.50° was selected as the final downscaling algorithm in this study. The final downscaled results were validated by applying the observations for a duration of 11 years obtained from six raingauge stations in the Qaidam Basin. These results indicated that the downscaled result effectively captured the trends in inter-annual variability and the magnitude of annual precipitation with the coefficient of determination r 2 ranging from 0.72 to 0.96 at six different raingauge stations. ► The relationship among precipitation, DEM, and NDVI can be used to downscale the precipitation estimates. ► The relationship among precipitation, NDVI, and DEM, is variable at different scales. ► The selection of the optimal scale is crucial to the final downscaled results. ► Model fitting is not a good indicator for the selection of the optimal scale.