Abstract
This paper presents a geostatistical downscaling procedure to improve the spatial resolution of precipitation data. The kriging method with external drift has been applied to surface rain intensity (SRI) data obtained through the Operative Precipitation Estimation at Microwave Frequencies (OPEMW), which is an algorithm for rain rate retrieval based on Advanced Microwave Sounding Units (AMSU) and Microwave Humidity Sounder (MHS) observations. SRI data have been downscaled from coarse initial resolution of AMSU-B/MHS radiometers to the fine resolution of Spinning Enhanced Visible and InfraRed Imager (SEVIRI) flying on board the Meteosat Second Generation (MSG) satellite. Orographic variables, such as slope, aspect and elevation, are used as auxiliary data in kriging with external drift, together with observations from Meteosat Second Generation-Spinning Enhanced Visible and InfraRed Imager (MSG-SEVIRI) in the water vapor band (6.2 µm and 7.3 µm) and in thermal-infrared (10.8 µm and 8.7 µm). The validation is performed against measurements from a network of ground-based rain gauges in Southern Italy. It is shown that the approach provides higher accuracy with respect to ordinary kriging, given a choice of auxiliary variables that depends on precipitation type, here classified as convective or stratiform. Mean values of correlation (0.52), bias (0.91 mm/h) and root mean square error (2.38 mm/h) demonstrate an improvement by +13%, −37%, and −8%, respectively, for estimates derived by kriging with external drift with respect to the ordinary kriging.
Highlights
Rainfall is of primary importance in many scientific fields, such as meteorology, hydrology, agriculture, ecology and other environmental sciences [1]
A crucial step in the method of kriging with external drift lies in the selection of the auxiliary variables, which strongly depends on the parameter to be downscaled
The continuous statistics show the comparison between rain gauge observations against the original Operative Precipitation Estimation at Microwave Frequencies (OPEMW) surface rain intensity (SRI) data, the downscaled SRI data by ordinary kriging (OK) and the downscaled SRI data by kriging with external drift (KED), obtained using the trend producing the best results for each case
Summary
Rainfall is of primary importance in many scientific fields, such as meteorology, hydrology, agriculture, ecology and other environmental sciences [1]. A large number of downscaling methods has been developed and applied in the last few years Most of these methods are based on the correlation between rainfall and environmental information such as latitude, longitude, altitude, slope, aspect and other orographic characteristics. Generation Radar (NEXRAD) daily precipitation fields were downscaled from 16 km to 4 km by considering orographic effects on precipitation distribution [3] This method consists of three parts, namely the rain-pixel clustering, the multivariate regression and the random cascade. [5] explored the relation between TRMM rainfall estimates and NDVI at different spatial scales; the derived relation has been used to develop a downscaling method based on an exponential regression model. The above studies demonstrate that downscaled precipitation data better capture the spatial variability compared to the original datasets
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have