1. Introduction
Tropical cyclones (TCs) are one class of the natural disasters that cause substantial damage to coastal areas worldwide (Mendelsohn et al. 2012; Zhu and Quiring 2017). In the United States, TCs are the most costly weather disaster and they have caused ~997.3 billion U.S. dollars in damage since 1980 (NOAA/NCEI 2021). Heavy precipitation associated with TCs can lead to severe flooding and this causes substantial social and economic losses (Villarini et al. 2011; Zhu et al. 2015). For example, when Hurricane Harvey made landfall in Texas in 2017, it produced more than 1000 mm of precipitation in some areas (van Oldenborgh et al. 2017). Harvey caused historic flooding in Texas and resulted in ~125 billion U.S. dollars in damage (Blake and Zelinsky 2018). More recently, Hurricane Florence (2018) produced record-breaking precipitation (more than 700 mm) in North Carolina and caused ~17 billion U.S. dollars in damage (Stewart and Berg 2019). Li et al. (2019) demonstrated that a better understanding of precipitation patterns can improve the reliability and resilience of water management systems and it can help decision makers employ appropriate adaptation, mitigation, and avoidance strategies for extreme precipitation events. Therefore, a better understanding of the spatial and temporal patterns of tropical cyclone precipitation (TCP) is important.
Previous studies have used a variety of different data sources for TCP, including gauge observations (Kunkel et al. 2013; Zhu and Quiring 2017), weather radar (Rendon et al. 2013; Weissman and Bourassa 2011), and satellite remote sensing (Kidder et al. 2000; Lin and Qian 2019). Each of these data sources has strengths and weaknesses. For example, a strength of gauges is that they provide a direct measurement of precipitation. However, gauges tend to be relatively sparse, they have an uneven spatial distribution and they often lack high temporal resolution. This makes it challenging to use gauge observations to study TCP (Hegerl et al. 2015; Kidd et al. 2017). Ground-based radar estimations provide a relatively high spatial and temporal resolution (Guirguis and Avissar 2008). However, the accuracy of radar-based estimations of TCP are influenced by many factors, such as the brightband effect (Guirguis and Avissar 2008), terrain interference (Serafin and Wilson 2000), and determining the appropriate Z–R relationship (Klazura et al. 1999). Satellite remote sensing is another widely used approach for studying TCP. Satellite-based products provide spatially continuous precipitation data at regional and global scales. Many satellites have been launched since the 1980s, among which, the Tropical Rainfall Measuring Mission (Huffman et al. 2007) is commonly used in TCP-related studies (Fritz et al. 2016; Jiang and Zipser 2010; Jiang et al. 2019; Pei and Jiang 2018).
The TRMM Multisatellite Precipitation Analysis (TMPA) provides precipitation estimates between 50°N and 50°S with high spatial (0.25°) and temporal resolution (3 h) (Huffman et al. 2007). Multiple studies have shown that TMPA can provide reliable estimates of precipitation at monthly and annual time scales, but it has lower skill for precipitation events at a finer time scale (Huffman et al. 2007). TMPA has been previously used for regional and global TCP studies (Jiang et al. 2019; Jiang and Ramirez 2013; Khouakhi et al. 2017; Lonfat et al. 2004; Matyas 2014; Prat and Nelson 2013, 2016; Shepherd et al. 2007). Many regional studies have demonstrated that TMPA can accurately represent the spatial pattern of daily TCP (AghaKouchak et al. 2011; Chen et al. 2013b; Deo et al. 2017; Zhu and Quiring 2017). However, TMPA has a tendency to overestimate light rain and underestimate heavy rainfall in tropical cyclones (Deo et al. 2017; Zhu and Quiring 2017). Although version 7 of TMPA uses a new algorithm to account for bias that considers the reflectivity–precipitation relationship and influence of terrains (Huffman and Bolvin 2018a; Huffman et al. 2010; Huffman et al. 2007; Zulkafli et al. 2014), Deo et al. (2017) and Chen et al. (2013c) demonstrated that TMPA 3B42 v7 tends to perform poorly in estimating TCP over complex terrain. The amount of TCP underestimation increases with elevation. In addition, the accuracy of TMPA in estimating TCP can vary with the distance from the storm center, the intensity of the storm, the latitude of the storm, and other considerations, such as interactions with the Asian monsoon and when it occurs during the TC season (Chen et al. 2013a; Deo et al. 2017). TMPA tends to provide more skillful TCP estimates over the ocean than over the land because of its inability to capture the orographic enhancement of TCP after landfall (Chen et al. 2013b). The post-real-time version of TMPA has better agreement with gauges and ground-based radars than the real-time version, as demonstrated during several tropical-related heavy rainfall events in Louisiana (Habib et al. 2009). An evaluation of the ability of numerous satellite-based precipitation products to capture heavy rainfall during TCs demonstrated that all of them tend to underestimate precipitation (S. Chen et al. 2013). However, past research has shown that TMPA is the second-best source of TCP data, after ground-based radar, and so it provides a relatively reliable record of TCP in data-sparse regions (S. Chen et al. 2013).
Based on the success of TRMM, precipitation data collection continued through the joint National Aeronautics and Space Administration (NASA)–Japanese Aerospace Exploration Agency (JAXA) Global Precipitation Measurement (GPM) mission (Huffman et al. 2019; Huffman et al. 2020). The GPM Core Observatory launched in 2014. GPM features a number of improvements over TRMM, including the Core Observatory’s instruments (Hou et al. 2014), spatial coverage (Blumenfeld 2015), and formalized partnerships with other agencies to obtain their satellites’ precipitation-relevant sensor data (Tang et al. 2016). Some applications of the Integrated Multisatellite Retrievals for GPM (IMERG) for measuring TCP have been conducted. Rios Gaona et al. (2018) used IMERG v04 Final Run data to investigate the spatial structures of TCP based on 166 TCs worldwide. Thakur et al. (2018) quantitatively identified the asymmetry and distribution of TCP over Bay of Bengal. A large number of studies have compared TMPA and IMERG over the globe. For example, Liu (2016) compared the TMPA 3B43 v7 precipitation with IMERG v03D Final Run worldwide and found that differences between TMPA and IMERG vary with surface type (land or ocean) and precipitation rate. Liu (2016) used monthly products, which are reasonable for climatological comparisons, but they are less appropriate for characterizing extreme events like TCs. Libertino et al. (2016) assessed the timing of extreme rainfall events globally. They found that 10% of events are more accurately identified by IMERG Final Run 3IMERGHH than TMPA 3B42 v7 based on comparisons with a rain gauge network. However, their study only assessed the day of occurrence of intense rainfall, but not the rainfall amount. Guilloteau et al. (2017) assessed the precipitation rate estimated by TRMM Microwave Imager (TMI) and GPM Microwave Imager (GMI) at different spatial scales. TMI is the passive microwave instrument that is used in the TRMM Combined Instrument (TCI) calibration dataset for TMPA, while GMI is used in the GPM Combined Instrument (GCI) for IMERG. They concluded that GMI can accurately reproduce precipitation at finer scales than TMI over the ocean, while over the land, they have similar performance. The magnitude of finescale noise is reduced in the 2017 version of the Goddard Profiling Algorithm (GPROF2017; Passive Microwave Algorithm Team 2018) employed by IMERG for passive microwave retrievals, but the spatial variability remains.
Global validation studies are limited by inconsistencies in the number of rain gauges over time and space (Stevenson and Schumacher 2014). Therefore, many studies have also compared TMPA and IMERG products at regional scales. Xu et al. (2017) compared IMERG and TMPA 3B42 v7 data with a high-density rain gauge network over the Tibetan Plateau and concluded that IMERG tends to better capture light rains in this high-elevation region. He et al. (2017) and Wang et al. (2017) evaluated IMERG Final Run level 3 and TMPA 3B42 v7 estimated precipitation at Mekong River basin and stated that IMERG is better at capturing these rainfall events. However, He et al. (2017) and Wang et al. (2017) disagreed on the performance of IMERG in capturing heavy rainfall. Peng et al. (2020) compared the precipitation detection ability of four satellite products, including IMERG v06B and TMPA 3b43, over an agricultural region in China and concluded that IMERG performed better for light rain and snowfall detection. They also mentioned that the estimation accuracy of satellite products may vary in different regions. Other evaluation and comparison studies have been conducted in the United States (Gebregiorgis et al. 2018), China (Fang et al. 2019; Tang et al. 2016), Malaysia (Tan and Santo 2018), and eastern Asia (Kim et al. 2017).
Evaluation of the performance of satellite-derived precipitation have also been undertaken for hydrological applications. Yuan et al. (2017) assessed the performance of IMERG Final Run v05 and TMPA 3B42 v7 in streamflow simulations in a data-sparse mountainous watershed. They found significant underestimation in total runoff and high flows. Jiang et al. (2018) evaluated the hydrological impacts of IMERG Final Run v05 and TMPA 3B42 v7 in a midlatitude basin and found that a hydrological model driven by IMERG had a higher correlation with observations and higher Nash–Sutcliffe efficiency than TMPA. Chen et al. (2020) evaluated two IMERG Final Run products (V06Uncal and V06Cal) and investigated the hydrological responses under Hurricane Harvey. They found that IMERG tended to overestimate the low-to-moderate precipitation intensity but underestimate the highest precipitation intensities.
In summary, although there are many studies comparing TMPA and IMERG, none of them have focused solely on North Atlantic TCP. This knowledge gap needs to be filled because 1) TC-related heavy rainfall events have significant socioeconomic impacts in coastal regions; 2) unlike synoptic rainfall events, TCP is highly spatially heterogeneous, causing both heavy and light rainfall; and 3) TCP influences both ocean and land areas and therefore satellites provide one of the best ways to estimate precipitation. This study will address this knowledge gap by answering two research questions: 1) How much more accurate is IMERG-F than TMPA in capturing TCP? 2) How close is IMERG-F-based TCP to true precipitation?
To achieve these objectives, we will utilize a novel gauge-based reference dataset that has specifically been designed for TCP to evaluate IMERG-F and TMPA. This study will focus on North Atlantic TCs that have influenced the contiguous United States and Mexico between 2014 and 2018. The data sources and evaluation metrics used in this study are described in section 2. This is followed by a comprehensive evaluation of TMPA and IMERG-F in section 3. The discussion and limitations are summarized in section 4, and we conclude with a presentation of key findings in section 5.
2. Data and methods
a. Study area
North Atlantic tropical cyclones have a significant influence on seasonal precipitation patterns along the East Coast of the United States, the Gulf of Mexico, and most of Mexico (Dominguez and Magaña 2018; Larson et al. 2005). Therefore, this study will examine the accuracy of satellite-derived TCP in these regions of the United States and Mexico between 2014 and 2018 (as shown in Fig. 1). In the United States, the study area extends from the East Coast westward to the Great Plains. The study area also covers the entire southern and eastern portions of Mexico. The elevation in the study area ranges from 0 to 5630 m (southern Mexico). There was a total of 35 tropical storms that caused TCP in the study area between 2014 and 2018 (Fig. 2), and all but one of these storms occurred during the hurricane season (May–November).
b. TRMM 3B42V7
This study uses the TRMM 3B42V7 product (Huffman et al. 2007). It is based on version 7 of TMPA that combines multiple independent satellite precipitation estimates and uses calibrated infrared (IR) estimates to fill the passive microwave coverage gaps in time and space (Huffman and Bolvin 2018b). TMPA has a 3-h temporal resolution and a 0.25° spatial resolution, and it is corrected by the monthly gauge data. Previous studies have reported that TMPA is generally more accurate over the ocean than over land (Ebert et al. 2007; Sapiano and Arkin 2009) because the algorithm uses both low- and high-frequency signals over the ocean and only the high-frequency signal over land. The time period that we focused on in this study is from 2014 to 2018, which is the transitional period from TRMM to GPM. Although the TRMM satellite descended to an altitude that precluded useful TRMM Precipitation Radar data in October 2014 and it was completely shut down in April 2015, the TMPA product continued to be generated till December 2019. As noted in Huffman (2020), NASA kept generating TMPA using other data sources until the GPM products can supersede it completely. Huffman (2020) also stated that “the actual demise of TRMM is not the substantive issue for the TMPA.” Therefore, we believe that during the transitional period (2014–18), when TRMM had been shut down and GPM had not yet fully calibrated and understood, TMPA provides high-quality data generated using a consistent algorithm. The GPM era started in 2014 with the launch of the GPM Core Observatory. The transitional period (2014–18) provides coverage from both TMPA and IMERG and it is the best option for evaluating the improvement of IMERG over TMPA. This is also why most previous comparisons of TMPA and IMERG were performed using this time period. The precipitation rate for each TC from 2014 to 2018 was derived from the NASA Goddard Earth Sciences (GES) Data and Information Services Center (DISC).
c. GPM IMERG V06
In this study, we used the recently released GPM IMERG V06 data (Final Run, hereafter referred to as IMERG-F), which has a 0.1° spatial resolution and a 30-min temporal resolution. IMERG-F merges, interpolates, and cross calibrates different satellite-derived precipitation estimates, including microwave precipitation estimates, microwave-calibrated infrared (IR) satellite estimates and precipitation gauge analyses. IMERG-F is also corrected by monthly gauge data. The precipitation rate for each TC from 2014 to 2018 was obtained from NASA GES DISC. In this study, we used the conservative interpolation method (Jones 1999) to upscale to 0.25° to match the spatial resolution of TMPA. This approach is recommended by the National Center for Atmospheric Research (Jones 1999; NCAR 2014). The original IMERG-F data (0.1°) is also included in spatial-average analyses.
d. Technical differences between TMPA and IMERG
Based on Huffman (2020), the following technical improvements have been made from TMPA to IMERG. TMPA calibrates the TMI (before 2015), Advanced Microwave Scanning Radiometer for Earth Observing Systems (AMSR-E), Special Sensor Microwave Imager (SSMI), Special Sensor Microwave Imager/Sounder (SSMIS), Advanced Microwave Sounding Unit (AMSU), and Microwave Humidity Sounder (MHS) to the TCI. Microwave-adjusted merged geo-infrared (IR) is used to fill spatial and temporal gaps. The 2010 version of the Goddard Profiling Algorithm (GPROF2010) is adopted to process input microwave rain rates. The final TMPA precipitation rate has a 3-hourly temporal resolution and a 0.25° spatial resolution. Besides the passive microwave sensors used by TMPA, IMERG incorporates more microwave sensors, including GMI and the Advanced Temperature and Moisture Sounder (ATMS). GPROF2017 is used as the input microwave algorithm, except for SAPHIR (Sondeur Atmosphérique du Profil d'Humidité Intertropical par Radiométrie), which is calculated by the Precipitation Retrieval and Profiling Scheme (PRPS). IMERG also uses the Climate Prediction Center (CPC) morphing Kalman filter (CMORPH-KF) quasi-Lagrangian time interpolation procedure and the Precipitation Estimation from Remotely Sensed Information Using Artificial Neural Networks–Cloud Classification System (PERSIANN-CCS) as infrared calibration data. The IMERG-F product has a half-hour temporal resolution and a 0.1° spatial resolution.
e. Gauge observations
We used daily gauge observation to evaluate TMPA and IMERG-F. A total of 11 850 gauges from the United States and 1943 gauges from Mexico were used. Figure 1 shows the spatial distribution of the gauges used in this study. The rain gauge data in the United States were obtained from the Daily Global Historical Climatology Network (GHCN-D) (Menne et al. 2012). The rain gauge data in Mexico were obtained from the National Weather Service of Mexico (Servicio Meteorológico Nacional of the National Water Commission). The daily observations of precipitation from gauges are partially independent of TMPA and IMERG-F data because both satellite products are only corrected by the monthly gauge data from Global Precipitation Climatology Centre (GPCC), not the daily.
f. Extracting TCP climatology from gauges and satellite products
A rain-gauge-based TCP climatology was constructed using the moving boundary technique (MBT). Similar to the approach developed by Zhu and Quiring (2017), we used a daily moving boundary to define which rain gauges received TC-related precipitation. A daily moving boundary was constructed by connecting circles with 500-km radius centered on the storm center locations reported by the International Best Track Archive for Climate Stewardship (IBTrACS; Knapp et al. 2018). Normally there are four connected circles for each day because the IBTrACS has a 6-h observation interval. We first merged the daily rain gauge based precipitation from GHCN-D and the National Weather Service of Mexico so that we would have the greatest possible number of gauge observations in our study area. Since there is no unified document about the observation time for those daily gauges, we assumed that daily precipitation data are collected at 0700 eastern time, which is the typical observing practice in North America (DeGaetano 2000). This time was used to identify the location of the TCs from IBTrACS and to define the TC boundary. This observation time was also used to produce daily TCP estimates from the TMPA and the IMERG-F rain rates. Following the methods of Zhu and Quiring (2017), we simulated the wind field for each TC and applied the appropriate bias adjustments for wind-induced gauge undercatch to each rain gauge. Inverse distance weighting (IDW) spatial interpolation was applied to the gauges to create daily gauge-based precipitation at 0.25° spatial resolution. IDW calculates the value at the target location based on a weighted average of its neighboring gauges. The weights are inversely proportional to the distance between each neighbor (Shepard 1968). We used a search radius of 30 km and a power of −2 as the two parameters for the IDW. This setup is the same as Zhu and Quiring (2017) and it has been optimized and validated for TC precipitation. Based on IDW, we calculated the gauge-based TCP at center of each satellite pixel and assumed that TCP at pixel center represents the areal averaged TCP in that pixel. For each of the 0.25° grid point, there are an average of three rain gauges that are used in the IDW spatial average.
g. Evaluation metrics
Five evaluation metrics are used in this study: bias (E; Golub and Van Loan 1996), relative bias (RE; Golub and Van Loan 1996), mean absolute error (MAE; Willmott and Matsuura 2005), correlation coefficient (r; Taylor 1997), and Nash–Sutcliffe efficiency (NSE; Nash and Sutcliffe 1970), and the equations for calculating them are provided in Table 1. These metrics are chosen because they can describe the similarity of two products from different angles. Bias shows the magnitude of the difference between gauge observation and satellite estimation. Relative bias indicates the bias of satellite estimates relative to the gauge observation. Mean absolute error describes the averaged magnitude of the error. Correlation coefficient represents the degree of agreement between gauge observation and satellite estimation. The Nash–Sutcliffe coefficient of efficiency is a commonly used statistic metric to evaluate model performance (Legates and McCabe 1999). In addition, the t test (Kalpić et al. 2011) and the Kolmogorov–Smirnov (KS; Lopes 2011) test are used for evaluating statistical significance at 95% confidence level (p < 0.05).
Evaluation metrics used to compare gauge measurements (G) and satellite precipitation products (S) in this study.
3. Results
a. Evaluation of annual TCP
Annual TCP in each grid cell is accumulated from the TCP that occurred from all TCs in each year. We calculated the mean annual TCP from 2014 to 2018 (hereafter referred to as mean annual TCP) by averaging the annual TCP in each grid. As shown in Table 2, the spatially averaged mean annual TCP based on the gauge observation is 51.88 mm. Both TMPA and IMERG-F underestimate the spatially averaged mean annual TCP. However, interpolated IMERG-F has significantly smaller MAE (13.78 mm), bias (−0.45 mm), and relative bias (−0.85%) than TMPA. There is not a statistically significant difference between the original and interpolated IMERG-F at the annual scale.
Comparison of mean annual TCP from rain gauges, TMPA, and IMERG-F. The “n/a” denotes that MAE cannot be calculated because the number of grid cells in 0.1° IMERG-F and 0.25° gauge data are not comparable. Bold values denote statistically significant improvements from TMPA to IMERG-F at the 95% confidence level.
Figure 3 shows the spatial patterns of mean annual TCP based on IMERG-F, TMPA and the gauges. More TCP occurs over the southeast and Gulf Coast of the United States. Specifically, gauge-based mean annual TCP indicates that along the Gulf Coast, Texas and Louisiana had the greatest mean annual TCP, while along the Atlantic coast, South Carolina and North Carolina had the greatest mean annual TCP between 2014 and 2018. Both TMPA and IMERG-F can accurately capture the general spatial patterns in mean annual TCP. However, significant differences are evident in some regions in Figs. 3d and 3e. TMPA significantly overestimated TCP in north Georgia, South Carolina, Virginia, and southern Mexico. It also significantly underestimated TCP in eastern North Carolina. Although the spatial patterns of over and underestimation are similar for IMERG-F, the area with statistically significant differences is smaller.
The bias distributions of TMPA and IMERG-F mean annual TCP are presented in Fig. 4. In terms of the shape of the histogram, TMPA and IMERG-F have similar bias distributions (statistically insignificant based on KS test). However, substantially more locations in IMERG-F have relatively small biases (±6 mm yr−1) than TMPA. TMPA has more locations with negative and large positive (>50 mm yr−1) biases than IMERG-F, while IMERG-F has more locations with biases ranging from 33 to 51 mm yr−1 than TMPA.
To compare the performance of IMERG-F and TMPA in regions that received similar TCP, we categorized each location into one of three categories based on the mean annual TCP: light TCP category has a mean annual TCP < 100 mm yr−1; moderate TCP category has a mean annual TCP between 100 and 300 mm yr−1; heavy TCP category has a mean annual TCP > 300 mm yr−1. Figure 5 shows the two-dimensional density plots of TMPA and IMERG-F versus gauge observations and the correlation coefficients are provided in Table 3. The density plots are shown as percentage, which is normalized by the total number grids in each category. In all three categories, mean annual TCP from both TMPA and IMERG-F are significantly correlated with gauge-based mean annual TCP. The moderate TCP category has the strongest correlation between the satellite and gauge observations, followed by the heavy TCP and light TCP categories. The light TCP category shows no systematic biases for both TMPA and IMERG-F. However, IMERG-F has a significantly (p < 0.05) higher correlation (r = 0.69) than TMPA (r = 0.61) in the light TCP category. In the moderate TCP category, TMPA shows denser dots below 1:1 line than IMERG-F, indicating TMPA has a larger underestimation bias than IMERG-F. This is consistent with the bias distribution shown in Fig. 4. However, the differences in correlation between TMPA and IMERG-F are not statistically significant. In the heavy TCP category, both TMPA and IMERG-F underestimate the mean annual TCP. IMERG-F has a significantly (p < 0.05) stronger correlations (r = 0.77) with gauge observations than TMPA (r = 0.71).
Correlation coefficient between TMPA, IMERG-F, and gauge-observed mean annual TCP. Correlation is calculated based on all locations which have a mean annual TCP that falls into the light (mean annual TCP < 100 mm yr−1), moderate (mean annual TCP between 100 and 300 mm yr−1), or heavy (mean annual TCP > 300 mm yr−1) categories. All of the correlation coefficients are statistically significant at the 95% confidence level. Bold values denote statistically higher correlation than the other product at the 95% confidence level.
b. Evaluation of TCP events
This section presents an event-based analysis of TCP. Figure 6 compares the TMPA and IMERG-F-based estimates of TCP versus the gauge-observed TCP for all the TC event from 2014 to 2018. Each TC is considered as a separate event. The precipitation for each event is calculated by spatially averaging the total precipitation in all locations affected by the TC. Both TMPA and IMERG-F are significantly correlated with gauge observations. The original IMERG-F has the highest correlation coefficient with gauge TCP (r = 0.96). It is slightly higher than interpolated IMERG-F (r = 0.95) and TMPA (r = 0.90). TMPA substantially underestimated TCP during Hurricane Harvey (2017). Both the original and interpolated IMERG-F performed significantly better than TMPA during Harvey, but it is still substantially less than the gauge observed TCP. A detailed comparison of TCP during TC events that produce extreme rainfall is provided in section 3d.
Three statistics (MAE, RE, and NSE) were used to evaluate the accuracy of TMPA and IMERG-F-based estimations of TCP for each TC event (Table 4). Based on MAE, IMERG-F significantly (p < 0.05) outperforms TMPA in 17 of 35 storms, while TMPA performs better than IMERG-F in only 3 storms. RE and NSE show that IMERG-F performs significantly better than TMPA in 14 and 27 TC events, respectively. Figure 7 illustrates the relationship between TC severity (based on TCP depth) and TMPA/IMERG-F performance. Since the scatterplots of MAE and RE are consistent with each other, we only show the results for MAE and NSE in Fig. 7. Based on MAE, the differences between TMPA and IMERG-F are more pronounced in TC events with more TCP. Based on NSE, IMERG-F performs better than TMPA in TC events with relative light TCP. This indicates that the improvements that have been made to the IMERG-F algorithm are helpful in both lighter and heavier TCP events.
Comparison of TMPA and IMERG-F to gauge observations in each TC event that affected the contiguous United States and Mexico between 2014 and 2018 based on mean absolute error (MAE), relative error (RE), and Nash–Sutcliffe efficiency (NSE). “N/A” denotes 0 TCP over land. Bold values denote statistically significant differences (p < 0.05) in MAE and NSE between IMERG-F and TMPA.
We also examined how MAE varies by TC category (Fig. 8). TC category was determined by the maximum Saffir–Simpson category during the life of each storm. The biases for both TMPA and IMERG-F are substantially higher for peak-category 4 hurricanes than the other categories. Generally, it does not appear that there is a systematic relationship between tropical cyclone intensity and the accuracy of TMPA or IMERG-F-based estimates of TCP. There is not a statistically significant (p < 0.05) difference in MAE between IMERG-F and TMPA in the tropical storm/depression and peak-category 1–3 hurricanes, while in both peak-category 4 and 5 hurricanes, the performance of IMERG-F is statistically significantly better than TMPA based on MAE. The differences between the original and interpolated IMERG-F are also not significant in the tropical storm/depression and peak-category 1–3 hurricanes. However, in peak-category 4 and 5 hurricanes, these differences increase substantially as the tropical cyclone intensity increases.
c. Evaluation of daily TCP
We also evaluated TCP at the daily time scale by comparing the distributions of daily TCP from TMPA and IMERG-F with those from gauge observations (as shown in Fig. 9). In general, the shapes of three distributions are similar to each other based on the KS test at the 95% confidence level. This means that both TMPA and IMERG-F can provide estimates of daily TCP that are comparable to the gauge observations. When examining the upper tail of the TCP distribution, it is evident that there is less agreement between the satellites and gauge observations when daily TCP > ~150 mm. Based on the KS test for the upper tail specifically, the difference between the satellite estimates and gauge observations is significant at the 95% confidence level. Although both TMPA and IMERG-F underestimate the heavy TCP, IMERG-F matches the observations more closely than TMPA. When focusing on the lower end of the TCP distribution (Fig. 9, inset), the TMPA-estimated TCP overestimates the frequency when daily TCP is smaller than 5 mm. When daily TCP ranges from 6 to 10 mm, the TMPA and IMERG-F-based TCP are similar to each other.
d. Case study of extreme TCP events
Previous analyses have demonstrated that both TMPA and IMERG-F tend to systematically underestimate TCP during the TC events with extremely heavy rainfall (Chen et al. 2013b; Rios Gaona et al. 2018). Here we evaluate the spatial patterns of TCP in two recent TCs that produced record rainfall in the United States: Hurricane Harvey (2017) and Hurricane Florence (2018). Figure 10 shows the spatial pattern of TCP during Hurricane Harvey (2017). Gauge observations indicate that the heaviest TCP occurred in the coastal areas in Texas and the maximum TCP exceeded 850 mm. Both TMPA and IMERG-F show similar spatial patterns, but the maximum TCP was only 650 to 750 mm. Spatially, the gauge observations show that Harvey caused TCP from northern Mexico to Kentucky. TMPA shows a smaller TCP-affected area (1016 grid cells) than both IMERG-F (1201 grid cells) and the gauge observations (1740 grid cells). The spatial differences in TCP for Hurricane Harvey are shown in Fig. 11. IMERG-F has smaller biases than TMPA over the majority of the locations (orange pixels in Fig. 11c) that received TCP from Hurricane Harvey.
Gauge observations show that the maximum TCP from Hurricane Florence (2018) was >410 mm in North Carolina. Similar to Hurricane Harvey, both TMPA and IMERG-F estimate a lighter maximum TCP (260–310 mm) for Hurricane Florence. The bias map shows that both IMERG-F and TMPA systematically underestimated TCP in North Carolina and South Carolina and that these differences were around 120 mm in many locations. There are also some locations farther away from where Hurricane Florence made landfall where both TMPA and IMERG-F overestimated TCP. For example, there were locations in Ohio were both TMPA and IMERG-F overestimated TCP by ~40 mm (Figs. 11d,e).
We also calculated the TCP volume by multiplying TCP in depth by affected area. In Hurricane Harvey and Hurricane Florence, the total underestimations by TMPA are 49 091 096 952 m3 and 13 689 917 194 m3, respectively, which are higher than that of IMERG-F (Harvey: 20 865 168 332 m3 and Florence: 7 021 058 765 m3). The underestimations of TCP volume by both TMPA and IMERG-F suggest that the TCP-caused floods may be underestimated if they are based on these satellite measurements.
4. Discussion and limitations
Although TMPA has been one of the most widely used satellite precipitation products, our evaluation shows that TMPA underestimates TCP in the locations that receive more rainfall and overestimates TCP in the locations that receive less rainfall from TC events. This finding agrees with previous research. AghaKouchak et al. (2011) found that TMPA products tend to miss a significant volume of extreme rainfall in the United States. Chen et al. (2013a) compared the TMPA-based TCP with the Comprehensive Pacific Rainfall Database in Australia and found that TMPA performs poorly at coastal and island sites with higher elevations. Our results show that IMERG-F performed better than TMPA at capturing TCP both for times and locations that received extremely heavy rainfall and for those that received lower amounts of rainfall. One of the reasons that IMERG-F outperforms TMPA may be due to the improved morphing scheme. The morphing scheme replaced the IR precipitation scheme used to fill in the gaps in microwave precipitation in TMPA with the CMORPH (Joyce et al. 2004) and PERSIANN (Hong et al. 2004) schemes. This improvement not only provides a finer time resolution so that system evolution is more accurately captured, compared to the 3-h interval in TMPA, but also minimizes the low-quality IR contribution, which is only kept for filling long microwave gaps.
Spatial resolution is another fundamental improvement of IMERG over TMPA. Previous studies have demonstrated that the coarse spatial resolution of TMPA (0.25°) causes heavy precipitation to be underestimated over land (Tang et al. 2016) because heavy rainfall usually occurs at finer spatial and temporal scales (Fang et al. 2019; Touma et al. 2018). IMERG improved the spatial resolution to 0.1°. This leads to an improved ability to capture the spatial heterogeneity of heavy precipitation. To demonstrate this, we show the spatial pattern of TCP in Hurricane Harvey based on the original IMERG-F data (0.1° spatial resolution) in Fig. 12a and compare it with the gauge observations, TMPA and regridded IMERG-F in Figs. 10a–c. The spatial heterogeneity of TCP is better captured by the original 0.1° IMERG-F data. The maximum TCP captured by the original IMERG-F data (798 mm) is larger than those based on TMPA and regridded IMERG-F (<750 mm), which is closer to the gauge observations. The original IMERG-F data also performs better than TMPA and regridded IMERG-F in terms of capturing light rainbands. The light rainband over the junction of Louisiana, Arkansas, and Mississippi, which is missed by both TMPA and regridded IMERG-F, is well captured by the original IMERG-F. Figure 12b shows the cumulative distribution function (CDF) of TCP during Hurricane Harvey from all the grid cells. The KS statistic between the CDF of the original IMERG-F and the gauge observation is significantly (p < 0.05) smaller than that of TMPA and regridded IMERG-F.
Other theoretical reasons, such as extended microwave channels (up to 183 GHz) and more accurate reference data, may also improve the performance of IMERG on capturing TCP. For example, the newly added 89-GHz channel has been approved to better capture the ice scattering signals related to stratiform precipitation at inner core of TC (Chen et al. 2019).
In this study, we assumed that gauge observations are the ground truth and we used them to evaluate TMPA and IMERG-F-based TCP. Some previous studies have mentioned the disadvantages of using gauge observations. For instance, the data quality and spatial coverage vary with time and location (Kidd and Huffman 2011; Sun et al. 2018). In this study, we used quality-controlled the gauge observations and the gauge density is high (Zhu and Quiring 2017). Most of satellite pixels (5551 of 7341) have at least one gauge inside. Our analysis shows that there is no substantial difference between using occupied grids and all grids to generate TCP climatology (see appendix). However, there are still some locations with relatively few gauges and lower data quality. Therefore, it is likely that some of the differences between the satellite products and gauge observations are due to these issues. Another challenge for evaluating satellite estimation using observations from rain gauges is timing mismatch. For example, TMPA’s 3-h temporal interval can cause at most 1.5-h offset from the time of gauge observation. These issues need to be aware in the future studies. In addition, our rain-gauge-estimated daily TCP is partially independent of TMPA and IMERG-F since they both use monthly GPCC gauge-based product for corrections. Our analysis offers insights into the differences between rain gauges and satellite observations at the daily scale, with a focus on TCs. Moreover, since both TMPA and IMERG-F are corrected with reference to monthly climatology, which is usually lower than actual TCP, the undercatch correction can lead to the underestimation of satellite products on capturing TCP.
Kidd and Huffman (2011) noted that different interpolation methods can cause biases. IDW interpolates gauge observations to the center of the grid cell, while satellite data represents the areal averaged precipitation rate. In this study, we assumed that the TCP at the center of grid cell can well represent the mean TCP in that grid. However, we have to acknowledge that if the gauges are unevenly distributed or the density of gauges is too low, comparing IDW-interpolated data with satellite estimation may not be ideal.
In this study, our methodology is designed specifically for extracting TCP in the regions that are strongly affected by a TC. There is greater uncertainty in the estimates of TCP in areas that are located farther from the storm center because of the challenges of determining whether the precipitation can be attributed to the TC. This highlights the necessity of continuing to improve the methods of interpolating and assimilating gauge-based precipitation data.
Last but not least, since IMERG has only incorporated GPM observations since 2014, our study is limited to 5 years. As more data become available, this evaluation can be extended to provide more information about long-term trends and variability and to improve our understanding of the performance of IMERG.
5. Summary and conclusions
This study provides a comprehensive evaluation of the accuracy of TMPA and IMERG-F during TC events. Both products were evaluated by comparing with gauge observations over the United States and Mexico from 2014 to 2018. IMERG-F incorporates more advanced sensors and uses improved algorithms to estimate precipitation. This study has two novel aspects: 1) evaluating TMPA- and IMERG-F-based tropical cyclone precipitation, and 2) using gauge-based reference data that have undergone wind loss adjustment and special processing to better represent true TCP. The main findings are as follows:
Both TMPA and IMERG-F were generally able to accurately capture mean annual TCP over the United States and Mexico. TMPA slightly underestimates mean annual TCP, while IMERG-F shows no significant differences with gauge observations. Spatially, both TMPA and IMERG-F can accurately reproduce variations in mean annual TCP. However, both TMPA and IMERG-F significantly underestimated precipitation in the regions that receive the most TCP. Even so, the performance of IMERG-F was significantly better than TMPA in these regions because it had a higher correlation coefficient and fewer locations with statistically significant differences with the gauge observations than TMPA. There is no significant difference between the original and the interpolated IMERG-F at annual scale.
Focusing on the event-based analysis of TCP, IMERG-F improved the performance significantly in 17, 14, and 27 (out of 31) TC events over TMPA based on MAE, RE, and NSE, respectively. These improvements are more pronounced in the events with higher and lower amounts of TCP. The original IMERG-F shows the highest correlation with gauge observation at event scale, followed by the interpolated IMERG-F and TMPA.
The case study analysis of the two extreme TCP events, Hurricane Harvey (2017) and Hurricane Florence (2018), revealed that both IMERG-F and TMPA systematically underestimate the maximum TCP by ~150 mm. However, IMERG-F outperforms TMPA in most locations.
At the daily time scale, the underestimations by IMERG-F and TMPA increase substantially when the TCP > 150 mm. The differences between IMERG-F and the gauges are consistently smaller than those between TMPA and the gauges.
In conclusion, we have demonstrated that IMERG-F provides more accurate estimates of TCP as compared to TMPA on all time scales. These improvements are especially noticeable in the regions that receive the greatest and least TCP. However, neither IMERG-F nor TMPA is able to accurately estimate heavy TCP. Future applications based on either of these two products need to be treated with caution, if the focus is on heavy TCP.
Acknowledgments
TMPA (doi:10.5067/TRMM/TMPA/DAY/7) and IMERG-F (doi:10.5067/GPM/IMERG/3B-HH/06) data are available from NASA GES DISC (https://disc.gsfc.nasa.gov/). Gauge observations are derived from the Daily Global Historical Climatology Network (GHCN-D, https://www.ncdc.noaa.gov/ghcnd-data-access) and the National Weather Service of Mexico (by request). Tropical cyclone tracks are obtained from International Best Track Archive for Climate Stewardship (IBTrACS, https://www.ncdc.noaa.gov/ibtracs/).
APPENDIX
Using all Grid Cells or the Grid Cells with More than One Gauge
We include all grid cells with at least one gauge in them and cells with no gauges in them. There are a total of 7341 grid cells that have TCP from 2014 to 2018: 5551 contain at least one rain gauge and 1790 have no rain gauges. Figure A1 demonstrates the spatial distribution of the grid cells with rain gauges (blue) and those without (yellow).
Then we matched each empty grid with one nearest grid (neighbor grid) filled with gauge. The distances range from 18.91 to 148.14 km, 76.3% of those neighbor grids are within 27.80 km (0.25°) of the original empty grids. Then we picked all daily TCP associated with the empty grids and the neighbor grids and compare their distributions. There are 17 729 observations of daily TCP for both of the empty grids and neighbor grids. Their histograms are compared as Fig. A2. The difference between the two distributions is not substantial. The distributions generally agree well for TCP > 5 mm. Given that these two samples are not from the same location, differences shown here might also be due to factors other than the undersampling issue in spatial interpolation. Therefore, we do not think removing those empty grid cells will have a significant impact on our results.
REFERENCES
AghaKouchak, A., A. Behrangi, S. Sorooshian, K. Hsu, and E. Amitai, 2011: Evaluation of satellite-retrieved extreme precipitation rates across the central United States. J. Geophys. Res., 116, D02115, https://doi.org/10.1029/2010JD014741.
Blake, E. S., and D. A. Zelinsky, 2018: Tropical cyclone report: Hurricane Harvey (AL092017), 17 August–1 September 2017. NHC Tech. Rep., 77 pp., www.nhc.noaa.gov/data/tcr/AL092017_Harvey.pdf.
Blumenfeld, J., 2015: From TRMM to GPM: The evolution of NASA precipitation data. NASA, https://earthdata.nasa.gov/learn/articles/trmm-to-gpm.
Chen, F., Y. Fu, and Y. Yang, 2019: Regional variability of precipitation in tropical cyclones over the western North Pacific revealed by the GPM Dual-Frequency Precipitation Radar and Microwave Imager. J. Geophys. Res. Atmos., 124, 11 281–11 296, https://doi.org/10.1029/2019JD031075.
Chen, M., and Coauthors, 2020: Can remote sensing technologies capture the extreme precipitation event and its cascading hydrological response? A case study of Hurricane Harvey using EF5 modeling framework. Remote Sens., 12, 445, https://doi.org/10.3390/rs12030445.
Chen, S., and Coauthors, 2013: Performance evaluation of radar and satellite rainfalls for Typhoon Morakot over Taiwan: Are remote-sensing products ready for gauge denial scenario of extreme events? J. Hydrol., 506, 4–13, https://doi.org/10.1016/j.jhydrol.2012.12.026.
Chen, Y., E. E. Ebert, K. J. E. Walsh, and N. E. Davidson, 2013a: Evaluation of TMPA 3B42 daily precipitation estimates of tropical cyclone rainfall over Australia. J. Geophy. Res. Atmos., 118, 11 966–11 978, https://doi.org/10.1002/2013JD020319.
Chen, Y., E. E. Ebert, K. J. E. Walsh, and N. E. Davidson, 2013b: Evaluation of TRMM 3B42 precipitation estimates of tropical cyclone rainfall using PACRAIN data. J. Geophys. Res. Atmos., 118, 2184–2196, https://doi.org/10.1002/jgrd.50250.
DeGaetano, A. T., 2000: A serially complete simulated observation time metadata file for U.S. Daily Historical Climatology Network stations. Bull. Amer. Meteor. Soc., 81, 49–68, https://doi.org/10.1175/1520-0477(2000)081<0049:ASCSOT>2.3.CO;2.
Deo, A., K. J. E. Walsh, and A. Peltier, 2017: Evaluation of TMPA 3B42 precipitation estimates during the passage of tropical cyclones over New Caledonia. Theor. Appl. Climatol., 129, 711–727, https://doi.org/10.1007/s00704-016-1803-0.
Dominguez, C., and V. Magaña, 2018: The role of tropical cyclones in precipitation over the tropical and subtropical North America. Front. Earth Sci., 6, 19, https://doi.org/10.3389/feart.2018.00019.
Ebert, E. E., J. E. Janowiak, and C. Kidd, 2007: Comparison of near-real-time precipitation estimates from satellite observations and numerical models. Bull. Amer. Meteor. Soc., 88, 47–64, https://doi.org/10.1175/BAMS-88-1-47.
Fang, J., W. Yang, Y. Luan, J. Du, A. Lin, and L. Zhao, 2019: Evaluation of the TRMM 3B42 and GPM IMERG products for extreme precipitation analysis over China. Atmos. Res., 223, 24–38, https://doi.org/10.1016/j.atmosres.2019.03.001.
Fritz, C., Z. Wang, S. W. Nesbitt, and T. J. Dunkerton, 2016: Vertical structure and contribution of different types of precipitation during Atlantic tropical cyclone formation as revealed by TRMM PR. Geophys. Res. Lett., 43, 894–901, https://doi.org/10.1002/2015GL067122.
Gebregiorgis, A. S., and Coauthors, 2018: To what extent is the day 1 GPM IMERG satellite precipitation estimate improved as compared to TRMM TMPA-RT? J. Geophys. Res. Atmos., 123, 1694–1707, https://doi.org/10.1002/2017JD027606.
Golub, G. H., and C. F. Van Loan, 1996: Matrix Computations. 3d ed. The Johns Hopkins University Press, 694 pp.
Guilloteau, C., E. Foufoula-Georgiou, and C. D. Kummerow, 2017: Global multiscale evaluation of satellite passive microwave retrieval of precipitation during the TRMM and GPM eras: Effective resolution and regional diagnostics for future algorithm development. J. Hydrometeor., 18, 3051–3070, https://doi.org/10.1175/JHM-D-17-0087.1.
Guirguis, K. J., and R. Avissar, 2008: A precipitation climatology and dataset intercomparison for the western United States. J. Hydrometeor., 9, 825–841, https://doi.org/10.1175/2008JHM832.1.
Habib, E., A. Henschke, and R. F. Adler, 2009: Evaluation of TMPA satellite-based research and real-time rainfall estimates during six tropical-related heavy rainfall events over Louisiana, USA. Atmos. Res., 94, 373–388, https://doi.org/10.1016/j.atmosres.2009.06.015.
He, Z., L. Yang, F. Tian, G. Ni, A. Hou, and H. Lu, 2017: Intercomparisons of rainfall estimates from TRMM and GPM multisatellite products over the upper Mekong River basin. J. Hydrometeor., 18, 413–430, https://doi.org/10.1175/JHM-D-16-0198.1.
Hegerl, G. C., and Coauthors, 2015: Challenges in quantifying changes in the global water cycle. Bull. Amer. Meteor. Soc., 96, 1097–1115, https://doi.org/10.1175/BAMS-D-13-00212.1.
Hong, Y., K.-L. Hsu, S. Sorooshian, and X. Gao, 2004: Precipitation estimation from remotely sensed imagery using an artificial neural network cloud classification system. J. Appl. Meteor., 43, 1834–1853, https://doi.org/10.1175/JAM2173.1.
Hou, A. Y., and Coauthors, 2014: The Global Precipitation Measurement mission. Bull. Amer. Meteor. Soc., 95, 701–722, https://doi.org/10.1175/BAMS-D-13-00164.1.
Huffman, G. J., 2020: The transition in multi-satellite products from TRMM to GPM (TMPA to IMERG). NASA Doc., 5 pp., https://gpm.nasa.gov/resources/documents/transition-multi-satellite-products-trmm-gpm-tmpa-imerg.
Huffman, G. J., and D. T. Bolvin, 2018a: TRMM and other data precipitation data set documentation. NASA Doc., 46 pp., https://pmm.nasa.gov/sites/default/files/document_files/3B42_3B43_doc_V7_180426.pdf.
Huffman, G. J., and D. T. Bolvin, 2018b: Real-time TRMM multi-satellite precipitation analysis data set documentation. NASA Doc., 51 pp., https://gpm.nasa.gov/sites/default/files/document_files/3B4XRT_doc_V7_180426.pdf.
Huffman, G. J., and Coauthors, 2007: The TRMM Multisatellite Precipitation Analysis (TMPA): Quasi-global, multiyear, combined-sensor precipitation estimates at fine scales. J. Hydrometeor., 8, 38–55, https://doi.org/10.1175/JHM560.1.
Huffman, G. J., R. F. Adler, D. T. Bolvin, and E. J. Nelkin, 2010: The TRMM MultiSatellite Precipitation Analysis (TMPA). Satellite Rainfall Applications for Surface Hydrology, M. Gebremichael and F. Hossain, Eds., Springer, 3–22.
Huffman, G. J., D. T. Bolvin, E. J. Nelkin, and J. Tan, 2019: Integrated Multi-satellitE Retrievals for GPM (IMERG) technical documentation. NASA Tech. Doc., 77 pp., https://pmm.nasa.gov/sites/default/files/document_files/IMERG_doc_190909.pdf.
Huffman, G. J., and Coauthors, 2020: Integrated Multi-satellite Retrievals for the Global Precipitation Measurement (GPM) Mission (IMERG). Advances in Global Change Research, V. Levizzani et al., Eds., Springer, 343–353.
Jiang, S., and Coauthors, 2018: Statistical and hydrological evaluation of the latest Integrated Multi-satellitE Retrievals for GPM (IMERG) over a midlatitude humid basin in South China. Atmos. Res., 214, 418–429, https://doi.org/10.1016/j.atmosres.2018.08.021.
Jiang, H., and E. J. Zipser, 2010: Contribution of tropical cyclones to the global precipitation from eight seasons of TRMM data: Regional, seasonal, and interannual variations. J. Climate, 23, 1526–1543, https://doi.org/10.1175/2009JCLI3303.1.
Jiang, H., and E. M. Ramirez, 2013: Necessary conditions for tropical cyclone rapid intensification as derived from 11 years of TRMM data. J. Climate, 26, 6459–6470, https://doi.org/10.1175/JCLI-D-12-00432.1.
Jiang, H., C. Tao, and Y. Pei, 2019: Estimation of tropical cyclone intensity in the North Atlantic and northeastern Pacific basins using TRMM satellite passive microwave observations. J. Appl. Meteor. Climatol., 58, 185–197, https://doi.org/10.1175/JAMC-D-18-0094.1.
Jones, P. W., 1999: First- and second-order conservative remapping schemes for grids in spherical coordinates. Mon. Wea. Rev., 127, 2204–2210, https://doi.org/10.1175/1520-0493(1999)127<2204:FASOCR>2.0.CO;2.
Joyce, R. J., J. E. Janowiak, P. A. Arkin, and P. Xie, 2004: CMORPH: A method that produces global precipitation estimates from passive microwave and infrared data at high spatial and temporal resolution. J. Hydrometeor., 5, 487–503, https://doi.org/10.1175/1525-7541(2004)005<0487:CAMTPG>2.0.CO;2.
Kalpić, D., N. Hlupić, and M. Lovrić, 2011: Student’s t-tests. International Encyclopedia of Statistical Science, M. Lovric, Ed., Springer, 1559–1563.
Khouakhi, A., G. Villarini, and G. A. Vecchi, 2017: Contribution of tropical cyclones to rainfall at the global scale. J. Climate, 30, 359–372, https://doi.org/10.1175/JCLI-D-16-0298.1.
Kidd, C., and G. Huffman, 2011: Global Precipitation Measurement. Meteor. Appl., 18, 334–353, https://doi.org/10.1002/met.284.
Kidd, C., A. Becker, G. J. Huffman, C. L. Muller, P. Joe, G. Skofronick-Jackson, and D. B. Kirschbaum, 2017: So, how much of the Earth’s surface is covered by rain gauges? Bull. Amer. Meteor. Soc., 98, 69–78, https://doi.org/10.1175/BAMS-D-14-00283.1.
Kidder, S. Q., and Coauthors, 2000: Satellite analysis of tropical cyclones using the Advanced Microwave Sounding Unit (AMSU). Bull. Amer. Meteor. Soc., 81, 1241–1260, https://doi.org/10.1175/1520-0477(2000)081<1241:SAOTCU>2.3.CO;2.
Kim, K., J. Park, J. Baik, and M. Choi, 2017: Evaluation of topographical and seasonal feature using GPM IMERG and TRMM 3B42 over Far-East Asia. Atmos. Res., 187, 95–105, https://doi.org/10.1016/j.atmosres.2016.12.007.
Klazura, G. E., J. M. Thomale, D. S. Kelly, and P. Jendrowski, 1999: A comparison of NEXRAD WSR-88D radar estimates of rain accumulation with gauge measurements for high- and low-reflectivity horizontal gradient precipitation events. J. Atmos. Oceanic Technol., 16, 1842–1850, https://doi.org/10.1175/1520-0426(1999)016<1842:ACONWR>2.0.CO;2.
Knapp, K. R., H. J. Diamond, J. P. Kossin, M. C. Kruk, and C. J. Schreck, 2018: International Best Track Archive for Climate Stewardship (IBTrACS) Project, Version 4. NOAA/National Centers for Environmental Information, accessed 20 April 2019, https://doi.org/10.25921/82ty-9e16.
Kunkel, K. E., and Coauthors, 2013: Monitoring and understanding trends in extreme storms: State of knowledge. Bull. Amer. Meteor. Soc., 94, 499–514, https://doi.org/10.1175/BAMS-D-11-00262.1.
Larson, J., Y. Zhou, and R. W. Higgins, 2005: Characteristics of landfalling tropical cyclones in the United States and Mexico: Climatology and interannual variability. J. Climate, 18, 1247–1262, https://doi.org/10.1175/JCLI3317.1.
Legates, D. R., and G. J. McCabe Jr., 1999: Evaluating the use of “goodness-of-fit” measures in hydrologic and hydroclimatic model validation. Water Resour. Res., 35, 233–241, https://doi.org/10.1029/1998WR900018.
Li, Z., X. Li, Y. Wang, and S. M. Quiring, 2019: Impact of climate change on precipitation patterns in Houston, Texas, USA. Anthropocene, 25, 100193, https://doi.org/10.1016/j.ancene.2019.100193.
Libertino, A., A. Sharma, V. Lakshmi, and P. Claps, 2016: A global assessment of the timing of extreme rainfall from TRMM and GPM for improving hydrologic design. Environ. Res. Lett., 11, 054003, https://doi.org/10.1088/1748-9326/11/5/054003.
Lin, J., and T. Qian, 2019: Rapid intensification of tropical cyclones observed by AMSU satellites. Geophys. Res. Lett., 46, 7054–7062, https://doi.org/10.1029/2019GL083488.
Liu, Z., 2016: Comparison of Integrated Multisatellite Retrievals for GPM (IMERG) and TRMM Multisatellite Precipitation Analysis (TMPA) monthly precipitation products: Initial results. J. Hydrometeor., 17, 777–790, https://doi.org/10.1175/JHM-D-15-0068.1.
Lonfat, M., F. D. Marks, and S. Y. S. Chen, 2004: Precipitation distribution in tropical cyclones using the Tropical Rainfall Measuring Mission (TRMM) microwave imager: A global perspective. Mon. Wea. Rev., 132, 1645–1660, https://doi.org/10.1175/1520-0493(2004)132<1645:PDITCU>2.0.CO;2.
Lopes, R. H. C., 2011: Kolmogorov-Smirnov test. International Encyclopedia of Statistical Science, M. Lovric, Ed., Springer, 718–720.
Matyas, C. J., 2014: Conditions associated with large rain-field areas for tropical cyclones landfalling over Florida. Phys. Geogr., 35, 93–106, https://doi.org/10.1080/02723646.2014.893476.
Mendelsohn, R., K. Emanuel, S. Chonabayashi, and L. Bakkensen, 2012: The impact of climate change on global tropical cyclone damage. Nat. Climate Change, 2, 205–209, https://doi.org/10.1038/nclimate1357.
Menne, M. J., and Coauthors, 2012: Global Historical Climatology Network - Daily (GHCN-Daily), version 3 [Version 3.12]. NOAA National Climatic Data Center, accessed 12 February 2019, https://doi.org/10.7289/V5D21VHZ.
Nash, J. E., and J. V. Sutcliffe, 1970: River flow forecasting through conceptual models part I—A discussion of principles. J. Hydrol., 10, 282–290, https://doi.org/10.1016/0022-1694(70)90255-6.
NCAR, 2014: The Climate Data Guide: Regridding overview. https://climatedataguide.ucar.edu/climate-data-tools-and-analysis/regridding-overview.
NOAA/NCEI, 2021: U.S. billion-dollar weather and climate disasters. NOAA/NCEI, www.ncdc.noaa.gov/billions/.
Passive Microwave Algorithm Team, 2018: Global Precipitation Measurement (GPM) Mission: GPROF2017 Version 1 and Version 2 (used in GPM V5 processing). Algorithm Theoretical Basis Doc., 65 pp., http://rain.atmos.colostate.edu/ATBD/ATBD_GPM_V5B_April15_2018.pdf.
Pei, Y., and H. Jiang, 2018: Quantification of precipitation asymmetries of tropical cyclones using 16-year TRMM observations. J. Geophys. Res. Atmos., 123, 8091–8114, https://doi.org/10.1029/2018JD028545.
Peng, F., S. Zhao, C. Chen, D. Cong, Y. Wang, and H. Ouyang, 2020: Evaluation and comparison of the precipitation detection ability of multiple satellite products in a typical agriculture area of China. Atmos. Res., 236, 104814, https://doi.org/10.1016/j.atmosres.2019.104814.
Prat, O. P., and B. R. Nelson, 2016: On the link between tropical cyclones and daily rainfall extremes derived from global satellite observations. J. Climate, 29, 6127–6135, https://doi.org/10.1175/JCLI-D-16-0289.1.
Prat, O. P., and B. R. Nelson, 2013: Mapping the world’s tropical cyclone rainfall contribution over land using the TRMM Multi-satellite Precipitation Analysis. Water Resour. Res., 49, 7236–7254, https://doi.org/10.1002/wrcr.20527.
Rendon, S. H., B. E. Vieux, and C. S. Pathak, 2013: Continuous forecasting and evaluation of derived Z-R relationships in a sparse rain gauge network using NEXRAD. J. Hydrol. Eng., 18, 175–182, https://doi.org/10.1061/(ASCE)HE.1943-5584.0000579.
Rios Gaona, M. F., G. Villarini, W. Zhang, and G. A. Vecchi, 2018: The added value of IMERG in characterizing rainfall in tropical cyclones. Atmos. Res., 209, 95–102, https://doi.org/10.1016/j.atmosres.2018.03.008.
Sapiano, M. R. P., and P. A. Arkin, 2009: An intercomparison and validation of high-resolution satellite precipitation estimates with 3-hourly gauge data. J. Hydrometeor., 10, 149–166, https://doi.org/10.1175/2008JHM1052.1.
Serafin, R. J., and J. W. Wilson, 2000: Operational weather radar in the United States: Progress and opportunity. Bull. Amer. Meteor. Soc., 81, 501–518, https://doi.org/10.1175/1520-0477(2000)081<0501:OWRITU>2.3.CO;2.
Shepard, D., 1968: A two-dimensional interpolation function for irregularly-spaced data. ACM '68: Proceedings of the 1968 23rd ACM National Conference, ACM, 517–524, https://doi.org/10.1145/800186.810616.
Shepherd, J. M., A. Grundstein, and T. L. Mote, 2007: Quantifying the contribution of tropical cyclones to extreme rainfall along the coastal southeastern United States. Geophys. Res. Lett., 34, L23810, https://doi.org/10.1029/2007GL031694.
Stevenson, S. N., and R. S. Schumacher, 2014: A 10-year survey of extreme rainfall events in the central and eastern United States using gridded multisensor precipitation analyses. Mon. Wea. Rev., 142, 3147–3162, https://doi.org/10.1175/MWR-D-13-00345.1.
Stewart, S. R., and R. Berg, 2019: Tropical cyclone report: Hurricane Florence (AL062018). NHC Tech. Rep., 98 pp., www.nhc.noaa.gov/data/tcr/AL062018_Florence.pdf.
Sun, Q., C. Miao, Q. Duan, H. Ashouri, S. Sorooshian, and K.-L. Hsu, 2018: A review of global precipitation data sets: Data sources, estimation, and intercomparisons. Rev. Geophys., 56, 79–107, https://doi.org/10.1002/2017RG000574.
Tan, M. L., and H. Santo, 2018: Comparison of GPM IMERG, TMPA 3B42 and PERSIANN-CDR satellite precipitation products over Malaysia. Atmos. Res., 202, 63–76, https://doi.org/10.1016/j.atmosres.2017.11.006.
Tang, G., Y. Ma, D. Long, L. Zhong, and Y. Hong, 2016: Evaluation of GPM Day-1 IMERG and TMPA version-7 legacy products over Mainland China at multiple spatiotemporal scales. J. Hydrol., 533, 152–167, https://doi.org/10.1016/j.jhydrol.2015.12.008.
Taylor, J. R., 1997: An Introduction to Error Analysis: The Study of Uncertainties in Physical Measurements. 2nd ed. University Science Books, 327 pp.
Thakur, M. K., T. V. Lakshmi Kumar, S. Dwivedi, and M. S. Narayanan, 2018: On the rainfall asymmetry and distribution in tropical cyclones over Bay of Bengal using TMPA and GPM rainfall products. Nat. Hazards, 94, 819–832, https://doi.org/10.1007/s11069-018-3426-5.
Touma, D., A. M. Michalak, D. L. Swain, and N. S. Diffenbaugh, 2018: Characterizing the spatial scales of extreme daily precipitation in the United States. J. Climate, 31, 8023–8037, https://doi.org/10.1175/JCLI-D-18-0019.1.
van Oldenborgh, G. J., and Coauthors, 2017: Attribution of extreme rainfall from Hurricane Harvey, August 2017. Environ. Res. Lett., 12, 124009, https://doi.org/10.1088/1748-9326/aa9ef2.
Villarini, G., J. A. Smith, M. L. Baeck, T. Marchok, and G. A. Vecchi, 2011: Characterization of rainfall distribution and flooding associated with U.S. landfalling tropical cyclones: Analyses of Hurricanes Frances, Ivan, and Jeanne (2004). J. Geophys. Res., 116, D23116, https://doi.org/10.1029/2011JD016175.
Wang, W., H. Lu, T. Zhao, L. Jiang, and J. Shi, 2017: Evaluation and comparison of daily rainfall from latest GPM and TRMM products over the Mekong River basin. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens., 10, 2540–2549, https://doi.org/10.1109/JSTARS.2017.2672786.
Weissman, D. E., and M. A. Bourassa, 2011: The influence of rainfall on scatterometer backscatter within tropical cyclone environments—Implications on parameterization of sea-surface stress. IEEE Trans. Geosci. Remote Sens., 49, 4805–4814, https://doi.org/10.1109/TGRS.2011.2170842.
Willmott, C. J., and K. Matsuura, 2005: Advantages of the mean absolute error (MAE) over the root mean square error (RMSE) in assessing average model performance. Climate Res., 30, 79–82, https://doi.org/10.3354/cr030079.
Xu, R., F. Tian, L. Yang, H. Hu, H. Lu, and A. Hou, 2017: Ground validation of GPM IMERG and TRMM 3B42V7 rainfall products over southern Tibetan Plateau based on a high-density rain gauge network. J. Geophys. Res. Atmos., 122, 910–924, https://doi.org/10.1002/2016JD025418.
Yuan, F., and Coauthors, 2017: Assessment of GPM and TRMM multi-satellite precipitation products in streamflow simulations in a data-sparse mountainous watershed in Myanmar. Remote Sens., 9, 302, https://doi.org/10.3390/rs9030302.
Zhu, L., and S. M. Quiring, 2017: An extraction method for long-term tropical cyclone precipitation from daily rain gauges. J. Hydrometeor., 18, 2559–2576, https://doi.org/10.1175/JHM-D-16-0291.1.
Zhu, L., S. M. Quiring I. Guneralp, and W. G. Peacock, 2015: Variations in tropical cyclone-related discharge in four watersheds near Houston, Texas. Climate Risk Manage., 7, 1–10, https://doi.org/10.1016/j.crm.2015.01.002.
Zulkafli, Z., W. Buytaert, C. Onof, B. Manz, E. Tarnavsky, W. Lavado, and J.-L. Guyot, 2014: A comparative performance analysis of TRMM 3B42 (TMPA) versions 6 and 7 for hydrological applications over Andean–Amazon river basins. J. Hydrometeor., 15, 581–592, https://doi.org/10.1175/JHM-D-13-094.1.