An improved statistical downscaling scheme of Tropical Rainfall Measuring Mission precipitation in the Heihe River basin, China

Abstract

Estimating an accurate spatial distribution of precipitation with high resolution is necessary for hydrological and ecological applications, especially in data‐scarce and terrain‐complicated river basins. Satellite‐based precipitation data have been widely used to measure the spatial patterns of precipitation, but an improvement in accuracy and resolution is needed. In this article, a new statistical downscaling method is proposed to generate improved monthly precipitation fields at a higher spatial resolution of 1 km in Heihe River basin (HRB), China. The presented methods employed the geographical weighted regression (GWR) method to explore the non‐stationarity between precipitation and its factors, and used the high‐accuracy surface modelling method (HASM) to compensate for the errors produced in the GWR downscaling process. The GWR model was first established under five different spatial scales, and the optimal relation between precipitation derived from the Tropical Rainfall Measuring Mission (TRMM) and its influencing factors was found for each month. The errors caused during the scale change were modified by performing HASM as a data merging framework, which considered both the local climate characteristics and meteorological observations. Results showed that the GWR downscaling method could not generate spatial patterns of precipitation similar to those of the original TRMM products. Although the performance of the GWR method after residual interpolations using Kriging, IDW, and tension Spline was improved, there existed significant variations in some regions, and the accuracy of those methods was still not satisfactory. In comparison with the other four models, GWR‐HASM showed better performance in reproducing the precipitation field at a high spatial resolution. Results indicate that the proposed downscaling method appears feasible for precipitation estimation in data‐scarce river basins.

Relationships between Spatial and Temporal Variations in Precipitation, Climatic Indices, and the Normalized Differential Vegetation Index in the Upper and Middle Reaches of the Heihe River Basin, Northwest China

“…The Heihe River Basin (HRB) originates in the Qilian Mountains, and discharges into Juyanhai Lake. It is located along the land and maritime Silk Roads ("One Belt, One Road") between 97.1 • E-102.0 • E and 37.7 • N-42.7 • N, with a total area of approximately 1.43 × 10 5 km 2 [45][46][47]. The second largest inland river located in northwest China, the Heihe River is divided into three reaches (upper (UHRB), middle (MHRB), and lower) by two hydrological stations (Yingluoxia and Zhengyixia).…”

Improving daily precipitation estimates for the Qinghai‐Tibetan plateau based on environmental similarity

“…AEFs have been introduced in many regression models to predict precipitation based on the assumption that significant correlation exists between precipitation and AEFs. The correlation can be readily established for monthly or yearly precipitation data (Chen et al, 2014;Qiang et al, 2016;Ma et al, 2017;Zhao et al, 2018), but is much less easily established for daily precipitation (Chen et al, 2018a). In addition, many regression models spatially interpolate residuals of estimates (Xu et al, 2015;Chen et al, 2018a), assuming significant spatial autocorrelation of residuals.…”

Downscaling satellite‐derived daily precipitation products with an integrated framework

“…The spatial downscaling of satellite precipitation product is based on the assumption that there is a significant relationship between precipitation and environmental factors. Given the widespread nature of vegetation response to precipitation, most of the existing downscaling methods use NDVI as an auxiliary variable or as one of auxiliary variables (Immerzeel et al, 2009;Jia et al, 2011;Duan and Bastiaanssen, 2013;Chen et al, 2014;Zhao et al, 2018). Despite the fact that these methods have been applied to a variety of land types, they only focus on satellite-derived annual and monthly precipitation products due to no significant relationship existing between daily precipitation and vegetation.…”