Comprehensive error analysis of satellite precipitation estimates based on Fengyun-2 and GPM over Chinese mainland

Highlights

• A comprehensive comparison between Fengyun-2G and Fengyun-2F over Chinese mainland.

• Exploring the primary error sources of Fengyun products in Chinese mainland.

• Analyzing the dependence of Fengyun products on topographic complexity.

• Studying the capability of Fengyun products on different rainfall intensities.

Abstract

The spatial-temporal error characteristics of four mainstream satellite precipitation products developed by China and United States, respectively, including the Fengyun-based (FY-2F and FY-2G) and the GPM-based (IMERG-Late and IMERG-Final) over Chinese mainland were comprehensively analyzed from January 2018 to December 2019. In general, both IMERG-Final and FY-2G perform better at the hourly and daily scales. While the FY-2F product has the relatively worst performance with lowest correlation coefficient (CC) and highest root mean square error (RMSE) values. In particular, FY-2F has considerable total bias and missed precipitation errors in summer when compared to other three precipitation products. As for winter, the IMERG product suites exhibit significantly overestimation in north-central region of China, while an opposite underestimation occurred in the Fengyun products. Among the four precipitation estimates, IMERG-Final and FY-2G have the lowest normalized-RMSE (NRMSE) at the elevation range of 100–300 m and 300–500 m at daily scale, respectively. The performance at the hourly scale was found to be similar for IMERG-Final and IMERG-Late, but both of which are slightly superior to FY-2G across the elevation ranges. In terms of the detectability for different rain intensities, the IMERG products performed best at higher rain rates, while the Fengyun-based precipitation estimates are superior to IMERG at relatively lower rain ones. The assessment results reported here will provide some valuable feedbacks for algorithm developers of Fengyun products, and enable data users to further understand the error characteristics and potential deficiencies of Fengyun precipitation estimates.

Introduction

Precipitation is a vital element of the terrestrial water cycle. It is not only one of the crucial ways for human beings to obtain hydrological resources, but also the ultimate source of freshwater resources used by human beings (Allen and Ingram, 2002; Wu et al., 2013). Therefore, it is essential to measure, monitor, and analyze rainfall events accurately. Traditionally, precipitation measurement is mainly operated by rain gauges and ground-based weather radars (Li et al., 2013; Yong et al., 2016). Unfortunately, rain gauges and ground-based weather radars are distributed unevenly and sparsely, and restricted for the remote areas and mountain regions (Huffman et al., 2000; Shen et al., 2020). With the development of precipitation-related sensors and retrievals, satellite-based remote sensing has become a popular alternative way to provide spatial-temporal continuous precipitation on the global scale (Lu and Yong, 2018).

With the launch of the core satellite of Tropical Rainfall Measurement Mission (TRMM) in 1997, it laid the foundation of integrated multi-satellite retrieval technology for precipitation, thus entering the TRMM era (Liu et al., 2012; Pombo and de Oliveira, 2015; Hou et al., 2014; Prakash et al., 2018). At present, several quasi-global products include TRMM-based multi-satellite precipitation analysis product (TMPA; Huffman et al., 2007), Global Satellite Mapping of Precipitation (GSMaP; Kubota et al., 2007), Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks (PERSIANN; Hong et al., 2004; Sorooshian et al., 2000), Climate Prediction Center Morphing Method (CMORPH; Joyce et al., 2004), and Integrated Multi-satellite Retrievals for GPM (IMERG; Huffman et al., 2019a, Huffman et al., 2019b).

Satellite precipitation products (SPPs) have wider spatial coverages and gradually increasing data accuracy, which attract extensive attention in various fields, such as water resources management and flooding monitoring (Chen et al., 2019, Chen et al., 2020; Lu and Yong, 2020; Tian et al., 2010b; Tian et al., 2010a; Ushio et al., 2009; Xie et al., 2007; Yong et al., 2009). Shen et al. (2010) conducted a comprehensive analysis of the performance of six high-resolution SPPs (including PERSIANN, NRL, CMORPH, MWCOMB, TRMM-3B42, and TRMM-3B42RT). The results showed that the six SPPs well captured the overall spatial distribution and temporal change patterns, but underestimated the plum rains in eastern and central China. Using the Chinese daily Precipitation Analysis Product (CPAP) as a reference, Zhao et al. (2018) quantitatively evaluated three SPPs based on GSMaP V6 over eight regions in Chinese mainland. Zhang et al. (2019) assessed the performance of IMERG and GSMaP products during extreme precipitation storms in South China. Chen et al. (2020) conducted a comprehensive evaluation and comparison of six purely SPPs on a global and regional scale, and revealed the relationship between inversion accuracy and precipitation intensity, and analyzed the primary error sources of these purely SPPs over Chinese mainland. With the launch of the Fengyun series satellites, China has Fengyun-based precipitation estimates (FYSPs). For instance, Fengyun-2 series of geostationary meteorological satellites are the first generation of geostationary satellites developed by China. Currently, there are three on-orbit Fengyun-2 satellites providing meteorological observation services for East Asia. However, few investigations have been conducted to error analysis and evaluation from FYSPs (Chen et al., 2016; Lu et al., 2020; Tian et al., 2009; Xu et al., 2019).

The main differences between the FYSPs and other global near-real-time precipitation products are as follows: The FYSPs integrate ground reference in real-time. On the other hand, satellite precipitation near-real-time products (i.e., IMERG-Early, and GSMaP-NRT) have not been calibrated, nor have they integrated gauge data. For example, the IMERG-Final delay time for correction using ground stations around 3.5 months. Therefore, the FYSPs that release data in real-time in Chinese mainland have great application potential. With the increasing influence of Fengyun series satellites, it is necessary to evaluate the performance and availability of FYSPs (Lu et al., 2020; Tang et al., 2016a).

The purpose of this study is to quantitatively evaluate the Fengyun-2F (hereafter referred to FY-2F) and Fengyun-2G (hereafter referred to FY-2G) precipitation estimates and make a comparative analysis against two GPM-based precipitation products (i.e., IMERG-Late and IMERG-Final) over Chinese mainland at hourly and daily scales. Additionally, this paper comprehensively investigates the performance of FY-2F and FY-2G under different terrain conditions. Finally, the primary error sources of satellite precipitation estimates over Chinese mainland are identified and analyzed. The goal of this study are as follows: (1) analyzing the performance of the latest FYSPs in Chinese mainland; (2) exploring the dependence of FYSPs on topographic complexity; (3) studying the capability of FYSPs on different rainfall intensities. The article is organized by describing the datasets (precipitation datasets of both satellite and gauge) and methodology in Section 2. Section 3 presents result by comparing SPPs to gauges, for statistical indicators, error components, topographic complexity, and rainfall intensity. A discussion of results follows in Section 4 and conclusions are provided in Section 5.