Probabilistic Precipitation Estimation with a Satellite Product

Nir Krakauer,Ajay Jha,Tarendra Lakhankar,Jeeban Panthi,Soni Pradhanang

doi:10.3390/cli3020329

Abstract

Satellite-based precipitation products have been shown to represent precipitation well over Nepal at monthly resolution, compared to ground-based stations. Here, we extend our analysis to the daily and subdaily timescales, which are relevant for mapping the hazards caused by storms as well as drought. We compared the Tropical Rainfall Measuring Mission (TRMM) Multi-satellite Precipitation Analysis (TMPA) 3B42RT product with individual stations and with the gridded APHRODITE product to evaluate its ability to retrieve different precipitation intensities. We find that 3B42RT, which is freely available in near real time, has reasonable correspondence with ground-based precipitation products on a daily timescale; rank correlation coefficients approach 0.6, almost as high as the retrospectively calibrated TMPA 3B42 product. We also find that higher-quality ground and satellite precipitation observations improve the correspondence between the two on the daily timescale, suggesting opportunities for improvement in satellite-based monitoring technology. Correlation of 3B42RT and 3B42 with station observations is lower on subdaily timescales, although the mean diurnal cycle of precipitation is roughly correct. We develop a probabilistic precipitation monitoring methodology that uses previous observations (climatology) as well as 3B42RT as input to generate daily precipitation accumulation probability distributions at each 0.25° x 0.25° grid cell in Nepal and surrounding areas. We quantify the information gain associated with using 3B42RT in the probabilistic model instead of relying only on climatology and show that the quantitative precipitation estimates produced by this model are well calibrated compared to APHRODITE.

Highlights

Precipitation products based on remote sensing offer the potential for improving hazard response and water resource management in mountainous areas with inadequate near-real-time ground-based data [1]
Gridded daily precipitation product, which is based on station precipitation observations, and against automatic weather stations that provide precipitation time series at high time resolutions
3B42) and ground-based products (APHRODITE and the automatic weather stations (AWSs)). Both satellite products have broadly similar regional spatial patterns of mean precipitation as APHRODITE when evaluated for the overlap period of 2001–2007 (Figure 1). 3B42 shows the benefits of retrospective calibration against rain gauges in a sharper precipitation gradient across the Himalayas that compares better with

Summary

Introduction

Precipitation products based on remote sensing offer the potential for improving hazard response and water resource management in mountainous areas with inadequate near-real-time ground-based data [1]. We have previously compared the performance of several remote sensing based precipitation products over Nepal relative to station observations on the monthly timescale, finding that the Tropical Rainfall Measuring. We extend this work by (a) considering the reliability of satellite precipitation at daily and subdaily temporal resolutions, which are better suited than monthly resolution for most hydrological hazard assessment work; (b) comparing the performance of the near-real-time. TMPA 3B42RT product to the research TMPA 3B42 product (the 3B43 product previously assessed is the monthly-resolution version of 3B42); (c) estimating rainfall rate probabilities conditional on the satellite data [7], which enables, for example, the mapping of areas that experienced heavy rainfall with high probability.

Methods

Results

Conclusion