Mapping the world's tropical cyclone rainfall contribution over land using the TRMM Multi-satellite Precipitation Analysis

Authors

  • Olivier P. Prat,

    Corresponding author
    1. Cooperative Institute for Climate and Satellites, North Carolina State University and NOAA/National Climatic Data Center, Asheville, North Carolina, USA
    • Corresponding author: Dr. O. P. Prat, Cooperative Institute for Climate and Satellites-NC (CICS-NC), North Carolina State University and NOAA/National Climatic Data Center, 151 Patton Ave., Asheville, NC 28801, USA. (olivier.prat@noaa.gov)

    Search for more papers by this author
  • Brian R. Nelson

    1. Remote Sensing Applications Division, NOAA/NESDIS/NCDC, Asheville, North Carolina, USA
    Search for more papers by this author

Abstract

[1] A study was performed to characterize over land precipitation associated with tropical cyclones (TCs) for basins around the world based upon the International Best Track Archive for Climate Stewardship (IBTrACS). From 1998 to 2009, data from the Tropical Rainfall Measuring Mission (TRMM) Multi-satellite Precipitation Analysis (TMPA) product 3B42, showed that TCs accounted for 5.5%, 7.5%, 6%, 9.5%, and 8.9% of the annual precipitation for impacted over land areas of the Americas, East Asia, South and West Asia, Oceania, and East Africa respectively, and that TC contribution decreased significantly within the first 150 km from the coast. Locally, TCs contributed on average to more than 25% and up to 61% of the annual precipitation budget over very different climatic areas with arid or tropical characteristics. East Asia represented the higher and most constant TC rain (118 mm yr−1±19%) normalized over the area impacted, while East Africa presented the highest variability (108 mm yr−1±60%), and the Americas displayed the lowest average TC rain (65 mm yr−1±24%) despite a higher TC activity. Furthermore, the maximum monthly TC contribution (8–11%) was found later in the TC season and depended on the peak of TC activity, TC rainfall, and the domain transition between dry and wet regimes if any. Finally, because of their importance in terms of rainfall amount, the contribution of TCs was provided for a selection of 50 urban areas experiencing cyclonic activity. Results showed that for particularly intense years, urban areas prone to cyclonic activity received more than half of their annual rainfall from TCs.

1. Introduction

[2] Tropical cyclones (TCs) constitute one of the major natural disasters around the world. A total of 894 TCs occurring from 1967 to 1991 were collectively responsible for nearly 900,000 deaths. For this reason, TCs ranked after drought as the second leading cause of loss of human life due to natural disasters [De et al., 2004]. Approximately 119 million people are exposed to TC hazards annually [United Nation Development Program (UNDP), 2004]. This large number is explained by the fact that about half the world's population live within 200 km of a coastline and the most cyclone-prone coastal regions are highly populated. For instance the majority of people who live in the Caribbean basin and US Gulf Coast live in the vicinity of the seashore. In China, about 60% of the population live in coastal areas and this proportion rises to more than 80% for Japan where no one is living more than 120 km from a coast [Hinrichsen, 1999]. While studies suggested that higher sea surface temperatures could result in more destructive TCs [Webster et al., 2005; Emanuel, 2005a; Dailey et al., 2009; Knutson et al., 2010; Villarini and Vecchi, 2012a, 2013], model projections provided conflicting trends for TC frequency and different long-term predictions with respect to the region considered [Knutson et al., 2010; Mendelsohn et al., 2012]. Once the subject of intense debate within the scientific community [Emanuel, 2005a, 2005b; Landsea, 2005; Pielke, 2005], recent studies have confirmed an upward trend in TC activity for the Atlantic basin while no similar trends were found for the other TC basins throughout the world [Kossin et al., 2007; Villarini and Vecchi, 2012b, 2013]. However, an anticipated rise in coastal population within the next decades [Creel, 2003] would undoubtedly increase TC-related damages [Pielke et al., 2008; Mendelsohn et al., 2012; Peduzzi et al., 2012]. In addition to important negative infrastructural, societal and economical impacts, TCs bring a large amount of rainfall and constitute an important contribution to water resources. In 2007, the intense TC Gamede established records for 72 h and 4 day rainfall accumulation of 3929 mm and 4869 mm respectively on the island of La Réunion [Quetelard et al., 2009]. A large bulk of the studies investigating TC rainfall contribution and their hydrological impacts have focused mainly on North America [Cry, 1967; Rodgers et al., 2001; Konrad et al., 2002; Larson et al., 2005; Atallah et al., 2007; Knight and Davis, 2007; Nogueira and Kleim, 2011; Prat and Nelson, 2013; Kam et al., 2013], East Asia [Ren et al., 2006; Wang and Chen, 2008; Yu et al., 2009; Lee et al., 2010] and Australia [Milton, 1980; Dare et al., 2012]. Other studies have investigated the link between TCs and extreme rainfall [Kim et al., 2006; Wu et al., 2007; Lau et al., 2008; Knight and Davis, 2009; Konrad and Perry, 2010; Barlow, 2011; Kunkel et al., 2011; Kuo et al., 2011] or natural hazards occurring in the aftermath of the passage of a TC such as floods, storm surge, debris flows, and landslides [Rappaport, 2000; Wooten et al., 2008; Dube et al., 2009; Eliot and Pattiaratchi, 2010; Villarini and Smith, 2010; Chang et al., 2011; Lin et al., 2011]. Global studies (covering e.g., the Caribbean's islands and the Antilles, the northern and southern Indian Ocean) that include all TC basins are seldom found.

[3] Due to global coverage, satellite rainfall products have proved useful in describing the structure of TCs and deriving the associated precipitation amount [Rodgers et al., 2001, Lonfat et al., 2004]. By providing consistent spatial coverage with subdaily temporal resolution, satellite rainfall estimates allow the opportunity to quantify the global TC contribution over large continental areas and basins around the world. Using over a decade of satellite rainfall estimates from the Tropical Rainfall Measuring Mission (TRMM), recent studies have investigated the global rainfall contribution for TC basins around the world [Jiang and Zipser, 2010; Jiang et al., 2011]. Similarly, the recent development of hurricane databases for all TC regions [Kossin et al., 2007; Knapp et al., 2010] facilitates the characterization of TC activity and provides a consistent data set to assess TC-related rainfall for the different basins.

[4] The aim of this work is to investigate the contribution of TC rainfall for the different TC basins around the world with a particular focus over land and coastal areas. For that purpose, precipitation data from the TRMM Multi-satellite Precipitation Analysis (TMPA 3B42) is used. This data product provides 3 hourly rainfall estimates at a spatial resolution of 0.25°×0.25° [Huffman et al., 2007]. While more general studies [Jiang and Zipser, 2010; Jiang et al., 2011], focused on global TC contribution to total rainfall regardless of land/ocean and regional considerations, this study provides the global contribution of TCs to the annual precipitation budget over land for the period 1998–2009.

[5] The paper is organized as follows: Section 2 describes the TMPA 3B42 product, TC database IBTrACS, and methodology. In section 3, precipitation associated with different TC basins are presented according to average contribution to the total precipitation budget and spatial distribution of TC rainfall for the period 1998–2009. In section 4, the interannual variability of TC-related precipitation is discussed and characterized by estimates of the monthly TC contribution. In section 5, results for selected urban centers around the world are presented (Appendix A) and the link between TC rainfall and water resources is investigated for areas of particular interest. The final section will summarize the key findings of the study.

2. Data and Methodology

2.1. TMPA 3B42 Precipitation Data Set

[6] Precipitation data from the TMPA 3B42 version 7 (hereafter TMPA) data product is used in this study. The product is a combination of different remotely sensed microwave [TRMM Microwave Imager (TMI), Special Sensor Microwave Imager (SSM/I), Advanced Microwave Scanning Radiometer (AMSR), Advanced Microwave Sounding Unit (AMSU)] and calibrated IR estimates of rain gauge corrected monthly accumulated precipitation (see Huffman et al. [2007] for details concerning the algorithm TMPA 3B42). The TMPA data set provides 3 hourly/0.25 degree precipitation estimates for the domain 50°S–50°N. From the TMPA data set, we derived annual and monthly tropical cyclone-related rainfall over the different TC basins for 12 TC seasons between 1998 and 2009. One of the caveats of the TMPA is the fact that it incorporates different satellite precipitation products for which retrieval algorithms were modified over the years. The incorporation of precipitation estimates from AMSU-B in 2001 generated low biases due to modification in the AMSU-B algorithm in 2003 [Huffman et al., 2007]. The TMPA precipitation estimates were greatly improved following a modification of the algorithm in 2007 [Huffman et al., 2007]. When compared to similar satellite precipitation data sets used for assessing TC rainfall, TMPA was found to perform relatively well in terms of accumulated rainfall distribution due to a monthly rain gauge adjustment [Yu et al., 2009]. However, for TMPA V6 considerable negative biases remained over land for instantaneous rain rates associated with TC rainfall [Yu et al., 2009; Villarini et al., 2011]. The TMPA V7 used here represents substantial improvement when compared to TMPA V6. In addition to incorporating additional satellite products along with the reprocessed versions of the merged algorithms (AMSU, TMI, AMSR, SSM/I), the most important upgrade consists of the use of a single, uniformly processed surface precipitation gauge analysis from the Global Precipitation Climatology Centre (GPCC) [Huffman and Bolvin, 2013]. The use of the GPCC rain gauge analysis explains most of the differences observed between V6 and V7 over land and over coastal areas. A brief comparison of the results obtained with the previous version (V6) is provided in Appendix B: Comparison of Results Obtained With TMPA V7 and V6.

2.2. Tropical Cyclones Track Information

[7] The TC tracks used in this study were taken from the International Best Track Archive for Climate Stewardship (WMO subset from v3r03 of IBTrACS: http://www.ncdc.noaa.gov/oa/ibtracs/), which gathers historical records of TC characteristics for all hurricane basins [Knapp et al., 2010]. IBTrACS is the most complete global historical data set of TC activity and is supported by data from the Regional Specialized Meteorological Centers (RSMCs) around the world. Along with the nature of the storm (extratropical, tropical, and subtropical), IBTrACS provides the storm center location, intensity, and other important characteristics of each tropical storm every 6 h [Knapp et al., 2010].

[8] The methodology used to determine TC-related rainfall follows Prat and Nelson [2013] where TC-related rainfall was identified as TMPA precipitation pixels found within a 500 km radius from the TC center. The 500 km radius criterion is similar to what has been used in comparable studies [Larson et al., 2005; Lau et al., 2008; Jiang and Zipser, 2010; Lee et al., 2010; Nogueira and Kleim, 2010; Schreck et al., 2011; Dare et al., 2012, among others] and encompasses the extent of the TC primary wind circulation domain (i.e., 80–400 km radius from TC center) and the curved TC cloud shield (i.e., 550–600 km radius) [Cry, 1967; Englehart and Douglas, 2001]. The main limitation in using a “one size fits all” 500 km radius criterion for each storm rests in the fact that precipitation from existing troughs/fronts might be included in the TC rainfall totals. Yet only subtle differences in precipitation were found for radii greater than 500 km [Larson et al., 2005; Prat and Nelson, 2013]. This methodology accounts for landfalling TCs [Atallah et al., 2007; Park et al., 2011] and tropical storms traveling along the coast and/or staying offshore. Finally, a linear interpolation of the TC center location and characteristics was performed every 3 h for consistency with the 3 hourly temporal resolution of TMPA. The storms are tracked independently so that the rain associated with each tropical storm is accounted for separately. The procedure follows the evolution of each storm using a 3 h time step. Each 3 h TMPA rainfall pixel for which the distance to the TC center location at the same time is less than 500 km is accounted as TC rainfall. The seasonal, annual, or monthly rain attributable to TCs is the sum of all the 3 h TMPA rainfall pixels.

2.3. Tropical Cyclone Activity From 1998 to 2009

[9] Figure 1 displays the 1205 TCs around the world for the period 1998–2009. The TC basins are easily identifiable on the map: (I) the North Atlantic Ocean (NAO), (II) the Northeast Pacific Ocean (NEPO), (III) the Northwestern Pacific Ocean (NWPO), (IV) the North Indian Ocean (NIO), (V) the South-West Indian Ocean (SWIO), (VI) the Australian Region (AUS), and (VII) the South Pacific Ocean (SPO). To assess the TC contribution over land, we selected six continental areas combining one or two TC basins which include (1) North and Central America and the Caribbean (NCA: 50°W–130°W, 5°N–50°N), (2) East Asia (EAS: 100°E–160°E, 5°N–50°N), (3) South and West Asia (SWA: 40°E–100°E, 5°N–50°N), (4) Oceania (OCE: 100°E–180°E, 5°S–50°S), and (5) East Africa (EAF: 25°E–65°E, 5°S–50°S). These domains correspond loosely to the six oceanic domains used in other studies [Lonfat et al., 2004]. Differences rest in the fact that the Americas domain combines the North Atlantic Ocean and the Northeast Pacific Ocean basins. Table 1 summarizes the correspondence between the five over land areas selected and the major TC basins according to the World Meteorological Organization (WMO) classification. As we focused on over land precipitation at continental scale, the Southwest Pacific Ocean basin, which consists of isolated islands of the same order of magnitude as the resolution of TMPA, was not included in this study.

Figure 1.

Annual and average monthly number of TCs for the five domains considered (NCA: North and Central America, EAS: East Asia, SWA: South and West Asia, OCE: Oceania, EAF: East Africa). The TC tracks (1205 TCs) for 1998–2009 are from the IBTrACS database [Knapp et al., 2010]. For the 1998–2009 and monthly averages, the bars indicate one standard deviation. The 1998–2009 TC activity is compared with the 1950–2009 (60 year) and 1980–2009 (30 year) periods.

Table 1. Characteristics of the Over Land Domains Considered in This Study Along With the Observed TC Activity for the Period 1998–2009a
DomainTC Activity (1998–2009)TC Basin (WMO)
  1. a

    Correspondence between the domains and the TC basins delineation according to the World Meteorological Organization (WMO). Acronyms for the domains considered: North and Central America (NCA), East Asia (EAS), South and West Asia (SWA), Oceania (OCE), and East Africa (EAF).

NCAJan/Apr–DecNAO & NEPO (North Atlantic & Northeast Pacific Ocean)
EASJan–DecNWPO (Northwestern Pacific Ocean)
SWAJan–DecNIO (northern Indian Ocean)
OCEJan–Jun/Oct–DecAUS & SPO (Australian Region & South Pacific Ocean)
EAFJan–Jul/Sep–DecSWIO (South-West Indian Ocean)

[10] During the period 1998–2009, the North America (NCA) domain experienced the most TCs with an average of 32.25 TC yr−1; ranging from 27 TCs (1998) to 44 TCs (2005). East Asia (EAS) had a lower yearly average with 22.75 TC yr−1 and ranged from 16 TCs (1998) to 28 TCs (2004). The other domains displayed comparable TC activity with 9.75, 10.5, and 7.6 TC yr−1 for SWA, OCE, and EAF respectively. The same trend was found for the minimum TC activity with a minimum of 5 (2007), 6 (2007), and 4 (2006) TCs for SWA, OCE, and EAF respectively. More differences were observed for the maximum TC activity with 15 (2005), 17 (1999), and 12 (2003) TCs for the same domains. Please note that the TC activity for each domain corresponds to TCs whose center crosses the domain as well as TCs that travel within 500 km from the limit of these domains. The reported TCs frequencies are partly a function of the size of the domains and might be different from the TC activity within the entire TC basin (Figure 1).

[11] For the monthly TC activity, characteristic bell-shaped curves were observed with maximum for August–October (ASO) for the Northern Hemisphere (NCA, EAS) and January–March (JFM) for the Southern Hemisphere (OCE, EAF). Most of the TC activity (60–70%) occurred over the three aforementioned months. For South and West Asia (SWA), the TC activity in the North Indian Ocean basin (NIO) was divided between the monsoon transition months in May–June (21%) and the end of wet season September–November (46%). Moreover, while the TC activity was mostly concentrated over the 6 month period of June–November (JJASON) for the Northern Hemisphere (95% for NCA, 84% for EAS) and of November–April (NDJFMA) for the Southern Hemisphere (94% for OCE, 92% for EAF), it was only 65% during the typical TC season (January/May–June/October–December) for the South and West Asia domain. Therefore, a significant number of TCs occurring between January and April would not be accounted for if considering only the duration of the TC season (January/May–June/October–December) for the North Indian Ocean domain [Jiang and Zipser, 2010].

3. Tropical Cyclones Rainfall and Contribution to the Total Precipitation Budget

3.1. Spatial Distribution of Tropical Cyclone Seasonal Contribution for Each TC Basin

3.1.1. North and Central America and Caribbean (NCA)

[12] Figure 2 displays the annual rainfall (Figure 2a) and the rainfall associated with TCs (TC rain) (Figure 2b). The band of heavy precipitation with rainfall totals above 2900 mm yr−1 (>8 mm day−1) along 10°N, corresponds to the Inter-Tropical Convergence Zone (ITCZ). In the Atlantic high rainfall accumulations greater than 2200 mm yr−1 (>6 mm day−1) were found along the Gulf Stream (Figure 2a). Over land, higher rain totals were found over Central America with maximum precipitation (8328 mm yr−1) near Quibdo in Colombia (wettest place of the Americas: http://www.ncdc.noaa.gov/oa/climate/ globalextremes.html). Significant rainfall was observed over the Yucatan peninsula, the Caribbean, and the Eastern United States (East of 95°W) that displayed higher precipitation than the Western counterpart (Figure 2a). The highest TC rain was found off the Pacific coast of Mexico with rain totals above 360 mm yr−1 (>1 mm day−1) (Figure 2b). The Caribbean Sea displayed TC rain between 240 and 320 mm yr−1, which was slightly higher than observed (180–260 mm yr−1) over the Gulf of Mexico and along the Atlantic coast (Figure 2b). Over land, the geographical extent of TC rain included the northern part of South America, the greater and lesser Antilles, central America, Mexico (south, central, and coastal Pacific, Baja California), eastern United States and up to Canada (Figure 2b). A maximum of 308.5 mm yr−1 was found near the coastal town of Motin de Oro in Western Mexico (Figure 2b). Locations of higher TC rain between 120 and 220 mm yr−1 included Mexico's Pacific Coast for the western seaboard of the NCA domain. Comparable amounts of TC rain were observed for the islands of the Caribbean (Cuba, Jamaica, and Hispaniola), the Yucatan Peninsula, Florida, and the eastern seaboard of the United States near the Carolinas (Figure 2b).

Figure 2.

(a) Total rainfall and (b) TC rainfall. (c) TC contribution for North and Central America and the Caribbean for 1998–2009. The gray box indicates the domain considered in Prat and Nelson [2013].

[13] The maximum TC contribution (ratio of the TC rain over the total annual rainfall) corresponded to locations with the higher TC rainfall (Antilles, Gulf coast and Florida peninsula, western Mexico seaboard) and/or annual rainfall was relatively low (Figure 2c). The western part of the NCA domain presented higher TC contribution from the TC of the Northwestern Pacific Ocean basin than the eastern counterpart impacted by TCs originating in the North Atlantic Ocean. TC contribution of about 15% was found along Mexico's Pacific Coast while a slightly lower contribution (10–15%) was observed for parts of the Yucatan Peninsula, the Antilles, Florida, and down to 8–10% for the coastal Texas/Louisiana and the eastern seaboard of the US over the Carolinas region. The maximum TC contribution that is above 40% of the annual precipitation budget corresponded to the Baja California Sur (below 28°N). This important contribution was due to the desert-like nature of the Baja California with a very low amount of annual precipitation of about 156 mm yr−1 among which 59 mm yr−1 were linked to cyclonic activity for the period 1998–2009. Furthermore, near the southernmost point of the Baja California, 61% of the annual rainfall was related to TC activity (Figure 2c). As discussed in Prat and Nelson [2013] focusing over the Southeastern United States (gray subdomain in Figure 2a), seasonal TC contribution derived from TMPA agrees reasonably well with comparable studies in terms of overall spatial patterns and quantitative TC induced seasonal rainfall [Rodgers et al., 2001; Larson et al., 2005; Knight and Davis, 2007; Shepherd et al., 2007; Jiang and Zipser, 2010]. Differences observed between studies are possibly explained by differences between sensors, period of observations, and/or the limitation to landfalling TCs only. It should be noted that unlike other studies that consider a seasonal basis (June–November), this work quantified TC contribution with respect to annual precipitation, thus explaining lower TC contribution.

3.1.2. East and Southeast Asia (EAS)

[14] For the EAS domain, higher rainfall accumulations were found over the ITCZ starting from the Philippines (>2900 mm yr−1) and along the Kuroshio Current (>2000 mm yr−1) (Figure 3a). Over land, the total maximum rainfall (4164 mm yr−1) was found over the Siargao Island in the Caraga region of the Philippines. High rain totals (>2500 mm yr−1) were observed over Malaysia, the Philippines, East Thailand, the region of the Cardamom Mountains in Cambodia, and Southern Japan (Figure 3a). Higher TC rain was concentrated over the central part of the domain with rainfall above 360 mm yr−1 for the Visayas and Luzon Island (Philippines), Taiwan, and coastal China with a maximum (865 mm yr−1) located on the east coast of the Luzon Island in the Philippines (Figure 3b). Taiwan presented a TC contribution greater than 30% of the annual rainfall with a maximum (40%) found near the town of Taimali in Eastern Taiwan. The maximum TC contribution (>30%) was located over the eastern half of the island while the western half displayed lower TC contribution (20–30%). Overall, TC provided about one third (31%) of the islands total rainfall, which was consistent with the 20–30% for most of Taiwan and the 40–50% for southernmost Taiwan [Ren et al., 2006]. A significant contribution (20%) was observed for the Hainan Island and corresponded to the lower limit (20–40%) reported in other studies [Ren et al., 2006; Yu et al., 2009]. Differences can be explained by the time periods considered (1998–2009 versus late fifties/early sixties to mid-2000s), measurement method (satellite versus ground based stations), or the continuation of the four-decade decreasing trend in the number of TCs impacting Hainan Island and Southeast China in general [Ren et al., 2006; Yu et al., 2009]. For southern China, important TC contribution (>20%) was concentrated along the coastline and indicated a rapid decrease of TC rainfall when moving inland. Overall, the TC rain and TC contribution spatial distributions were quantitatively similar to those obtained with ground based measurements [Lee et al., 2010]. Similarly, the Luzon Island (Philippines) displayed significant TC contribution (>15%) which is about half of that found over Taiwan (Figure 3c) despite a comparable TC induced rainfall (Figure 4b). Finally, Southern Japan (Kyūshū, Shikoku and Honshū Islands) presented a TC induced rainfall of about 10% (Figure 3c).

Figure 3.

(a) Total rainfall and (b) TC rainfall. (c) TC contribution for East Asia for 1998–2009.

Figure 4.

(a) Total rainfall and (b) TC rainfall. (c) TC contribution for South and West Asia for 1998–2009.

3.1.3. South and West Asia (SWA)

[15] For the over land domain corresponding to the North Indian Ocean basin, the higher rainfall accumulation was observed over India and Bangladesh with the characteristic precipitation regime dominated by summer Monsoon (Figure 4a). Higher rainfall (2500–5000 mm yr−1) corresponded to the western slopes of the Sahyadri Mountains on the western coast of India, Northeastern India, the Ganges Delta, and southern Burma. The maximum (5900 mm yr−1) was found near Mawsynram in Northeastern India, which corresponds loosely to the location of the highest average annual rainfall for Asia (http://www.ncdc.noaa.gov/oa/climate/globalextremes.html). The TC rain was mostly significant (110–210 mm yr−1) for areas of western India and Bangladesh bordering the Bay of Bengal with maximum TC rain (209 mm yr−1) found near the city of Balasore on the East Coast of India (Figure 4b). Although very low (<20 mm yr−1), the TC rain between the Gulfs of Oman and Aden was of the same order of magnitude as non-TC precipitation (Figure 4b). TC contribution was moderate for western India (10–15%), which corresponded to the location of the greater average TC rainfall (Figure 4c). The highest TC ratio was located over the Arabian Peninsula, where despite a sporadic cyclonic activity, TC contribution was important due to the arid nature of the environment with an average annual rainfall below 50 mm yr−1 (Figure 4a). A maximum annual TC contribution of 26% was found near the city of Sur in Oman. This maximum corresponded to the passage of the cyclone Gonu in 2007 (Figure 4c).

3.1.4. Oceania (OCE)

[16] In the south Pacific and Oceania, higher rain totals (>2900 mm yr−1) were found over Indonesia, Papua New Guinea with a maximum of 5781 mm yr−1 in Papua New Guinea (Figure 5a). Although the coastal areas of northern and eastern Australia experience average rainfall of 900–1900 mm yr−1, the majority of the continent exhibited limited rainfall (<450 mm yr−1). The cyclonic activity was concentrated over the coasts of northern and western Australia, and between Vanuatu and the Salomon Island in the Pacific with TC precipitation between 120 and 240 mm yr−1 (Figure 5b). The Melville Island near Darwin (Australia) experienced the maximum TC rainfall (252 mm yr−1). In addition, TC activity provided occasional rainfall to New Caledonia, and more occasional rainfall amounts to Indonesia and Papua New Guinea (Figure 5b). Figure 5c indicates that the TC contribution matched the spatial distribution of TC rainfall (Figure 5b). Significant TC contribution (10–20%) was found along the coast of western and northern Australia with the highest TC contribution (≈30%) over the coastal regions of Canning and Pilbara in Western Australia and a maximum (38%) at the mouth of the De Grey River north of Port Hedland (Figure 5c). Globally, these results are in good agreement with the recent findings of Dare et al. [2012] that looked at the seasonal (November–April) TC rainfall contribution over Australia. The seasonal [Dare et al., 2012] and annual (this study) data show higher TC rainfall along the northern coastline between 120°E and 150°E, while the highest TC contribution (>15% of annual rainfall) is found west of 125°E. In addition, TCs proportionally provided less rainfall (5–15%) to the northern Territories and Queensland due to higher annual rainfall associated with the tropical type of climate. The extent of the area displaying the highest TC contribution corresponded to a dry Western Australia associated with low rainfall totals and relatively high average intensity and therefore confirmed the role of TCs on precipitation patterns for the area as hypothesized [Smith et al., 2008]. Furthermore, the arid areas of the northern Territories and Western Australia benefited from the few TCs traveling further inland contributing to about 5–15% of the annual rainfall (Figure 5c). On a yearly basis, TC contributed a great amount of rainfall to the semiarid and tropical areas of western and northern Australia respectively, and brought significant precipitation contribution (>20%) over the desert of central Australia. This observation is consistent with previous studies reporting over 30% of rainfall for January–March coming from TC over a large portion of Western Australia with more than 50% for drier areas [Milton, 1980].

Figure 5.

(a) Total rainfall and (b) TC rainfall. (c) TC contribution for Oceania for 1998–2009.

3.1.5. East Africa and Madagascar (EAF)

[17] In the South Indian Ocean, Madagascar displayed the highest accumulation of precipitation along the east coast with maximum precipitation (3171 mm yr−1) and maximum TC rain (343 mm yr−1) found in the Alaotra-Mangoro region in eastern Madagascar (Figures 6a and 6b). Over Ocean, TC rain was the highest over western Madagascar, along the Mozambique Channel, and in the Indian Ocean north of the Islands of La Réunion and Mauritius (Figure 6b). TCs provided a substantial amount of rainfall for countries of southwest Africa such as Mozambique, and Zimbabwe, and to a lesser extent Malawi and Zambia (Figure 6b). The highest contribution (>30%) was observed over the Mozambique Channel and the western seaboard of Madagascar (Figure 6c). The islands of La Réunion and Mauritius received on average 18–22% of annual rainfall from TCs. At the northern end of the Mozambique Channel, Comoros experienced about 10–14% annual rainfall from TCs. The greater TC contribution (33%) was found near the town of Itampolo in the southwestern Madagascar despite a lower TC rain than the eastern seaboard of the Island (Figure 6c). Overall, higher TC contribution (15–25%) was located over the coast of the Nampula province in Mozambique, and along southwest Madagascar between the Massifs of Ivakoany (East) and Isala Roiniforme (North), a region with arid-like characteristics (Figure 6a). Due to higher non-TC rainfall, eastern Madagascar displayed a lower TC contribution (10–17%), while in central Madagascar TCs contributed a moderate amount (5–10%) of the total precipitation (Figure 6c).

Figure 6.

(a) Total rainfall and (b) TC rainfall. (c) TC contribution for East Africa for 1998–2009.

3.2. Influence of Land/Ocean Transition on TC Contribution

[18] Because coastal areas are vulnerable to TC rainfall, it is important to investigate the effect of ocean/land transition. Figure 7 displays the TC contribution (left column: Figure 7-n-a, where n=1,5), TC rain (middle column: Figure 7-n-b), and non-TC rain (right column: Figure 7-n-c) for each domain. The dash-dotted gray line represents the probability density function of the domain physical characteristic (land versus ocean) with respect to the distance from the coastline. The probability density function is computed by summing all the pixels as a function of the distance from the coast from the farthest ocean pixel to the farthest land pixel. The percentage of the ocean|land extent associated with each domain is 64|36, 67|33, 29|71, 78|22, and 70|30 for the NCA, EAS, SWA, OCE, and EAF domains respectively. The TC contribution decreased during the transition from ocean to land and was particularly pronounced for the first 100 km over land. (Figure 7-n-a). The decrease was the most significant for the NCA (Figure 7−1-a) domain with a TC contribution dropping from 70% near the coast to about 10% for distances greater than 150 km. For the other domains EAS (Figure 7−2-a), SWA (Figure 7−3-a), OCE (Figure 7−4-a), and EAF (Figure 7−5-a), the maximum contribution over coastal areas was close to 40%. Similarly, the TC rain decreased when moving inland which was more important for domains skewed toward a predominant ocean area like NCA (Figure 7−1-b), EAS (Figure 7−2-b), OCE (Figure 7−4-b), and EAF (Figure 7−5-b), and less significantly for the northern Indian Ocean domain (Figure 7−3-b). One possible explanation is that TCs impacting the NCA, EAS, OCE, and EAF domains originated far from the coasts thus were most likely to gain strength and generate higher rain rates at landfall or when reaching the coasts. Furthermore, the EAS domain (Figure 7−2-b) displayed maximum TC rain over the coast (80 mm yr−1) that is three-fold the coastal TC rain for the other domains. The geographical configuration of the EAS basin that includes a multitude of various size islands (Luzon, Visayas, Taiwan, Hainan, and Kyūshū) and elongated coasts (East China) located over the trajectories of TCs explains the comparatively elevated TC rain amount despite a lower TC activity of about a third in comparison of the NCA domain (Figure 1). For non-TC rainfall, higher accumulation due to sea breeze effects were found over coastal land areas (Figure 7-n-c). Globally, and at the exception of higher precipitation observed near the coast, non-TC rain displayed comparable magnitude over land regardless of distance from the coast (Figure 7-n-c). For the OCE domain, however, the amount of precipitation decreased when progressing inland due to the fact that about 60% of the Australian territory is semiarid or desert (Figure 7−4-c).

Figure 7.

(a) TC contribution, (b) TC rainfall, and (c) non-TC rainfall as a function of the distance from the coast for the TC domains (1–5). Data are for the period 1998–2009. Acronyms for the domains considered: North and Central America (NCA), East Asia (EAS), South and West Asia (SWA), Oceania (OCE), East Africa (EAF).

[19] We recognize that there is additional uncertainty related to the land/ocean transition due to sensor characteristics and retrieval techniques. The TMPA product optimally blends rainfall estimates derived from various sensors that use different retrieval algorithms over land and over ocean. Among them, the Goddard Profiling Algorithm (GPROF) experiences difficulties in retrieving precipitation over land and coastal areas due to the fact that only higher frequency channels are available. In addition, the retrieval over coastal areas (i.e., transition from ocean to land) and small islands tends to be delicate due to the fact that coastal pixels can simultaneously include land and ocean features [Adler et al., 2009].

3.3. TC Rainfall Contribution With Respect to the Distance From the Coast

[20] More quantitatively, Figure 8 displays the TC contribution, TC rain, and non-TC rain for each domain during the 12 year period of observation. Results are presented for the entire domain over land regardless of the local TC contribution (case 1) or computed over areas experiencing cyclonic activity (i.e., area impacted (AI) by at least 1 TC: case 2) and as a function of the distance from the coast (dcoast); that is for immediate coastal areas (dcoast<100 km: case 3), intermediate zone (100 km<dcoast<300 km: case 4), and inland (300 km<dcoast: case 5). For each one of the quantities presented: TC contribution (Figure 8a), TC rain (Figure 8b), and non-TC rain (Figure 8c), results for the entire domain (1) were noticeably lower than results for the subareas experiencing TC activity (2). The comparison of cases (1–2) provided an indication of the non-TC rain variability within the entire domain (1) compared with areas of preferential TC activity (2) (Figure 8c). The differences for non-TC rain were significant (NCA, EAS, SWA, OCE, EAF) with respect to the entire/TC impacted area criterion and particularly important for the SWA domain that included the arid areas of the Arabian Peninsula (Figure 4c) and the EAS domain that displayed important spatial rainfall variability (Figure 3b). For subdomains impacted by at least one TC, the TC contribution decreased with increasing distance from the coast for NCA, EAS and EAF domains but remained relatively constant for SWA and OCE (Figure 8a: cases 3–5). Moreover, the EAF domain displayed a larger variability for the TC contribution and TC rain than any of the other domains (Figures 8a and 8b). For domains characterized by the presence of islands and archipelagos over the path of TCs (EAS, EAF, NCA), the TC rainfall seemed to decrease faster with increasing distance from the coast while this decrease was less pronounced for wide continental areas (SWA, OCE) (Figure 8b).

Figure 8.

(a) TC contribution, (b) TC rainfall, and (c) non-TC rainfall over land for the five domains. Results are presented as a function of the distance from the coast (1–2: dcoast>0 km; 3: dcoast<100 km; 4: 100 km<dcoast<300 km; 5: dcoast>300 km) for the entire domain (1) and for areas experiencing cyclonic activity (2–5). Acronyms for the domains considered: North and Central America (NCA), East Asia (EAS), South and West Asia (SWA), Oceania (OCE), East Africa (EAF).

[21] Generally, the evolution of the TC contribution with respect to the distance from the coast depended on the geophysical features (presence of islands or archipelagoes), the local characteristics of non-TC rainfall (e.g., sea breezes and monsoon), and TC track variability. The annual TC induced rainfall for land|ocean|overall contribution was 2.2|6.5|5.1%, 4.4|10.6|9.2%, 1.4|3.8|2.5%, 3.0|3.2|3.2%, and 1.9|4.3|3.5% for NCA, EAS, SWA, OCE, and EAF respectively. Differences in time frame (annual versus TC season) and domain delineations (selected domain versus entire TC basins) explained the differences in contribution when compared to others that reported seasonal TC contribution of 8–9%, 7%, 11%, 5%, 7–8%, and 3–4% for the ATL, EPA, NWP, NIO, SIO, and SPA basins [Jiang and Zipser, 2010].

4. Annual and Monthly Variability of TC Contribution for 1998–2009

4.1. Annual Rainfall Contribution of Tropical Cyclones and Interannual Trends

[22] Figure 9 displays the annual TC accumulation (Figure 9a), TC rain (Figure 9b), and total rainfall (Figure 9c). Rainfall amounts were normalized with respect to the domain impacted by at least one TC for a given year to account for the geographical extent of the cyclonic contribution (Figure 9d). On a year-to-year basis, the minimum|mean|maximum annual TC contribution was 3.7|5.5|7.2%, 5.9|7.5|8.9%, 3.0|6.0|11.8%, 6.0|9.5|12.8%, and 3.2|8.9|17.9% for the NCA, EAS, SWA, OCE, and EAF domains respectively (Figure 9a). The EAS domain displayed the higher consistency while the EAF domain in the southwest Indian Ocean presented the higher TC contribution (2004) and the most important variability (Figure 9a). Globally, the lowest TC contribution was found for the SWA domain in the northern Indian Ocean (Figure 9a). Regarding TC induced rainfall, the EAS domain experienced the higher average totals despite a 30% lower TC activity (Figure 1a) when compared to the entire NCA domain (Figure 9b). Similarly, the domains OCE and EAF experienced roughly a comparable average TC rainfall than the NCA with a TC activity about a third of the NCA domain. Finally, the SWA domain systematically displayed the lowest TC rain over the 12 year period (Figure 9b).

Figure 9.

(a) TC contribution for the annual precipitation budget. (b) Yearly TC rain accumulation over land for each of the five domains. (c) Land area impacted (AI) by TCs. (d) Yearly TC rain accumulation over land normalized with respect of the AI. For the 1998–2009 average, the bars indicate one standard deviation. Acronyms for the domains considered: North and Central America (NCA), East Asia (EAS), South and West Asia (SWA), Oceania (OCE), East Africa (EAF).

[23] The NCA domain consistently had the highest geographical extent of TC activity followed by the EAS domain with a comparable extent for years 1999–2002, 2004, 2006–2007, and 2009 (Figure 9c). The EAF and SWA domains represented the lowest extent of TC activity. The EAF domain had the highest year-to-year variability compared with the EAS domain, which has higher consistency throughout the years. The difference between the minimum and the maximum annual extent for cyclonic activity over each domain varied between a two-fold to a five-fold ratio; given by 2.4, 2.7, 3.1, 4.2, and 5.2 for EAS, OCE, NCA, SWA, and EAF respectively. For NCA and EAS, annual trends were comparable regarding TC rain for the entire over land domain (Figure 9b) and TC rain reported over the AI on an annual basis (Figure 9d). In contrast, the normalized TC rainfall for domains SWA, OCE, and particularly EAF emphasized the influence of the local TC contribution on the annual precipitation budget. The EAS domain experienced the higher normalized TC rain of 118 mm yr−1 ranging from 86 to 161 mm yr−1. The NCA domain exhibited a lower normalized TC rain of 65 mm yr−1 (45–91 mm yr−1) comparable with SWA (87 mm yr−1: 43–162 mm yr−1) and OCE (82 mm yr−1: 49–115 mm yr−1), and was found significantly lower than the EAF domain (108 mm yr−1: 39–274 mm yr−1). The higher variability observed for the EAF domain can be partially explained by the geographical configuration (Madagascar and la Réunion Islands on the path of TCs) and by the characteristic of TC activity (TCs less frequently reach the coast of eastern Africa). For instance a higher total TC rain (Figure 9b) was associated with a more widespread TC activity (Figure 9c) for 2003 (that is 2002–2003 South-West Indian Ocean cyclone season). This was in contrast to 2004 that had three times the normalized TC rain (Figure 9d) due to the impact of two very intense TCs (Elita and Gafilo) that made landfall and rotated over Madagascar in early 2004 [Chang-Seng and Jury, 2010]. The largest AI was for 2000 when the TC Leon-Eline (1999–2000 SWIO TC season) crossed the entire Indian Ocean and traveled far inland impacting Mozambique, Zimbabwe, and South Africa.

4.2. Monthly Rainfall Contribution of Tropical Cyclones

[24] To compare the TC contribution between each TC basins, the TC ratio was computed over areas impacted that were defined as the over land domain that displayed rain attributable to TCs or in other words the geographical extent of the TMPA rainy pixels associated with TC rainfall for the period 1998–2009. This avoided averaging TC contribution over large areas never experiencing TCs (Figures 2c–6c) and accounting for different land/ocean geographical characteristics between domains (Figure 9). The maximum contribution of 11% occurred in September for NCA, and EAS (Figure 10a). For the SWA domain, the maximum TC contribution of 8% happened later in November. In the Southern Hemisphere for OCE, the maximum TC contribution (10%) happened in March, 1 month later than the maximum TC activity (Figure 1). More surprisingly for the EAF domain, the maximum ratio (9%) happened in May, which is after the conventional TC season (November–April) for the South-West Indian Ocean (Table 1). Slightly different results were found for the contribution computed over the entire domain (not shown). For NCA and EAS, September remained the month with higher TC contribution but with a lower value of 9% and 10% respectively due to the inclusion of a larger domain. Similarly for the OCE domain, maximum TC contribution (9%) remained in March at the end of the summer wet season (December–March) over tropical Australia [Suppiah, 1992]. Changes were found for the SWA domain with October displaying the higher contribution (4%). For the EAF domain the maximum TC contribution was found again in May (4%) and was higher than for February–March (3%).

Figure 10.

(a) TC contribution for the monthly precipitation budget. (b) Monthly TC rain accumulation over land for each of the five domains. (c) Monthly rainfall over land for each of the five domains. d) Land AI by TCs. Rainfall totals are computed over land areas impacted by at least one TC for the period 1998–2009. Acronyms for the domains considered: North and Central America (NCA), East Asia (EAS), South and West Asia (SWA), Oceania (OCE), East Africa (EAF).

[25] The TC contribution, which is the ratio of TC rain over total rainfall, depended both on TC activity and seasonal rainfall characteristics. Results showed that TC rain (amount of rain) over land matched TC activity (number of TCs) within each domain with maximum monthly TC rain (Figure 10b) corresponding globally to the peak of TC activity (Figure 1). Similarly, for domains NCA, EAS, and OCE, the maximum TC contribution occurred during the months of maximum TC activity. However, for the NCA domain despite comparable TC activity (Figure 1), September presented higher TC rain than August (+160%) due to a higher number of landfalling TCs in September [Prat and Nelson, 2013]. In addition, the variability of the TC contribution (Figure 10a) depended on the monthly total rainfall (Figure 10c). For instance, while August and September presented comparable TC rain for the EAS domain (Figure 10b), the TC contribution (Figure 10a) was higher in September due to lower total precipitation in August (Figure 10c). Furthermore the bell-shaped curve for the annual rainfall with maximum precipitation in July did not reveal intraseasonal fluctuations observed locally from the southern to the northern parts of the domain [Chen et al., 2004; Yihui and Chan, 2005]. On the other hand, for SWA and EAF, the maximum TC contribution was found later in the TC season after the peak of TC activity. Despite a TC rainfall amount (2.5 mm yr−1) about half the average for October (5.5 mm yr−1), the SWA maximum TC contribution (8%) was found at the end of the wet season (November) because total rainfall decreased concurrently from 96 mm yr−1 to 31 mm yr−1 (Figure 10c). More dramatic was the situation prevailing for the EAF domain with a maximum average TC contribution observed in May during the transition from the wet (November–April) to the dry season (May–October). While TC activity was occasional in May with TC rain representing only 17% of the January maximum (1.6 versus 9.3 mm yr−1), the transition from a wet to a dry regime caused a higher TC contribution (9% versus 5%). Regardless of the domains, the maximum monthly TC contribution occurred later in the TC season and depended on the TC activity, the TC induced rainfall, and the characteristic of the annual precipitation regime. This phenomenon was more emphasized for domains with strong annual cycles of precipitation such as the contrast between wet/dry seasons (EAF) or a monsoon regime (SWA). However, previous results have to be taken carefully due to the fact that a 12 year period is too short to derive robust statistics for average monthly TC contribution, particularly for months experiencing marginal TC activity and for which results are highly dependent on isolated storms. A better sense of the variability of the monthly TC activity can be seen with the extent of the monthly AI (i.e., area impacted by 1 TC on a given month of a given year) that displays a very important year-to-year variability particularly for months with marginal TC activity within each basin (Figure 10d).

[26] Furthermore, computing the maximum monthly TC contribution over the AI for a given month on a year-to-year basis shines a light on the influence of isolated storms during months with limited TC activity (Figure 11a). When reported over the AI, isolated storms generated substantial precipitation (Figure 11b). For instance for the EAF domain, the maximum TC contribution (73%) was found in May after the typical TC season in the Southern Hemisphere (November–April). This corresponded to two occurrences (TC probability=0.17=2/12: Figure 11c) during the 12 year period of observation with comparable results for the years 2002 and 2003. Similarly, for the EAS domain, the months of January, April, and May (outside the typical June–November TC season) presented maximum TC contributions of 58%, 56%, and 76% respectively over the AI (Figure 11a). Furthermore the maximum normalized TC rainfall (160 mm/AI) was observed in December, which is twice the maximum observed for September (79 mm/AI) despite a lower TC activity (Figure 1). Without getting into any further details for the other basins, these results revealed the important year-to-year variability of monthly TC rainfall contribution and caution should be taken with regard to any generalization. Deriving more robust statistics for the monthly TC contribution would require a longer period of observation in particular for months with the lower TC probability that is the beginning/end of the TC season (Figure 11c). Finally, those results have shown that despite their marginal occurrence, tropical storms occurring outside the typical TC season can have an important contribution, and should therefore be considered when determining the TC rainfall contribution.

Figure 11.

(a) Maximum TC contribution for the monthly precipitation budget computed over the land AI by at least one TC. (b) Maximum monthly TC rain accumulation over the impacted land area for each of the five domains. (c) TC probability of occurrence over the 12 year period of observation. The AI is defined as the over land domain experiencing at least one TC for each given month of each given year. Acronyms for the domains considered: North and Central America (NCA), East Asia (EAS), South and West Asia (SWA), Oceania (OCE), East Africa (EAF).

5. Quantification of TC Rainfall for Selected Urban Areas

[27] The previous sections provided results for the domains as an ensemble. Because the results were presented at a continental scale, the local TC contribution over highly populated areas affected by cyclonic activity was difficult to interpret. Here results are presented for 50 medium to large urban areas (see Appendix A for the abbreviation of each urban area) that are repeatedly or exceptionally impacted by TCs. Places with important TC activity (and TC rain) like Taiwan (TAIP), and the Philippines (MANI) for instance are found toward the upper right corner (Figures 12a and 12b). For specific years, a few urban areas (Key West: KEYW, La Paz: LPAZ, Port-au-Prince (Haiti): HAIT, Taipei: TAIP, Manila: MANI, Karachi: KARA, Masqat: MASQ, Port Hedland: PHED, La Réunion: REUN) received more than 40% of their annual rainfall from TCs (Appendix A). The cyclonic activity can also represent an important source of annual rainfall for arid areas like the Baja California (LPAZ) or a sudden inflow of fresh water and subsequent flooding to other arid places rarely impacted by TCs such as the Arabian Peninsula (MASQ) and Pakistan (KARA) (Figure 12a). Studies have shown that over arid zones, TC rainfall provided comparable groundwater recharge to regular long-term events and was dependent upon the geological, structural, and hydrological nature of the domain impacted [Abdalla and Al-Abri, 2011]. Displaying comparable maximum TC contribution (>40%), La Réunion Island (REUN) represented an intermediate situation due to the combination of average TC rainfall and comparatively low non-TC rainfall (Figure 12a). For La Réunion, the years 2006–2007 reflected an intense cyclonic activity with an annual TC contribution of about 45%. The highest TC contribution (47.1%) was found for 2007 when the TC Gamede impacted the island [Quetelard et al., 2009], while the highest TC rainfall (226 mm) was found for 2006.

Figure 12.

(a) TC contribution for 50 selected cities (Appendix A) around the world and (b) same as previous with a zoom over a specific domain. The bars represent one standard deviation for the rainfall accumulation (horizontal lines) and the TC rain (vertical lines). Acronyms for the domains considered: North and Central America (NCA), East Asia (EAS), South and West Asia (SWA), Oceania (OCE), East Africa (EAF).

[28] A closer look at selected subareas can further quantify the importance of TCs as a contribution to the annual precipitation budget (Figure 13). Over the 12 year period, Baja California Sur (arid climate) and Taiwan (marine tropical climate) indicated comparable TC contribution (34% and 31% respectively) with Baja California Sur representing a higher annual variability with minimum-maximum TC contribution of 8–61% versus 14–42% for Taiwan (Figure 13a). In Baja California Sur, TC rainfall was of the same order as non-TC precipitation (Figure 13b). Comparatively, TC rain consisted of a more constant source of water complementing non-TC precipitation over Taiwan (Figure 13c). However, moderate non-TC precipitation combined with low TC activity (2002) can have noticeable impacts on water resource availability resulting in generalized water shortages while increased TC activity (2003) was not necessarily linked with drought relief (Figure 13c). On the other hand, low TC activity (2006) was not necessarily synonymous with water shortage assuming that non-TC rain remained at a sufficient level (Figure 13c). Although this is beyond the scope of this study, these results indicate that TC rainfall consists of an important component of the annual water budget and that a low TC activity can have major impact on water resource availability in TCs prone areas. More work would be necessary to elucidate the link between TC activity and ecosystems resilience particularly over arid areas.

Figure 13.

(a) TC contribution for Baja California Sur (Mexico) and Taiwan. TC rainfall, non-TC rainfall, and total rainfall, for (b) Baja California Sur and (c) Taiwan.

6. Summary and Conclusion

[29] The contribution of TCs over land was quantified using long-term satellite precipitation data (TMPA) for the major basins around the world using TC track information from IBTrACS. The North America continent presented the most important cyclonic activity with an average of 32 TC yr−1 (Figure 1). For East Asia, the TC activity was about 70% of the North America activity, while for the other basins (northern Indian Ocean, Australia, South-West Indian Ocean) it represented between 1/4 and 1/3 of North America activity. Despite an activity 30% lower than North America (Figure 2), East Asia displayed the higher TC rainfall over land (>360 mm yr−1) for coastal areas and islands located within the triangle defined by the seas of south and east China, and the Philippines Sea (Figure 3). The other basins (SWA, OCE, and EAF) experienced comparable maximum TC rain solely over more localized areas (Figures 4-6). The areas with the highest TC rain did not necessarily correspond to the highest TC contribution as found for the arid Baja California that displayed a TC contribution above 30% and up to 61% (Figure 2). Furthermore, even sporadic cyclonic activity with moderate rainfall contributed a large portion of the precipitation budget (over 25%) as found for locations of the Arabian Peninsula (Figure 4). All basins with the exception of the northern Indian Ocean basin (Figure 4), represented significant subareas where TC rain accounted for more than 30% of the precipitation total (Figures 2-6).

[30] TC contribution varied mostly within the first 150 km of the coast with the most important decrease for the North America domain, which dropped from 70% near the coast to about 10% inland (Figure 7). Overall, the East Asia basin represented the most important decrease in TC rain and TC contribution at landfall due to intense TC activity as well as the presence of islands and archipelagoes on the paths of TCs (Figure 8). Over the 12 year observing period, average TC contributions for over land areas impacted by TC activity were 5.5%, 7.5%, 6%, 9.5%, and 8.9% for North America, East Asia, northern Indian Ocean, Australia, and South-West Indian Ocean respectively. More important differences between domains were found for the average extent (±σ%) of the AI by TCs with 4.1E+6(±29%), 3.3E+6(±21%), 1.3E+6(±38%), 2.0E+ 6(±33%), and 1.1E+6(±64%) km2 for NCA, EAS, SWA, OCE, and EAF respectively (Figure 9). With an average normalized TC rain of 118(±19%) mm yr−1, the East Asia domain displayed higher TC rainfall than the North and Central America (North Atlantic and Northeast Pacific Oceans), Oceania, South and West Asia (northern Indian Ocean), and South-West Indian Ocean domains with 65(±24%), 82(±20%), 87(±39%), and 108(±60%) mm yr−1 respectively. Globally, the East Asia domain represented the higher and most constant TC contribution while the Southwest Indian Ocean (EAF domain) displayed the highest year-to-year variability. Furthermore, despite a higher TC activity than the other domains, North and Central America presented a similar over land contribution as compared to West Africa and Oceania with only one third of the TC activity.

[31] While the maximum monthly TC rain corresponded typically to the peak of the cyclonic activity, the maximum monthly TC contribution occurred later in the TC season and was a conjunction between TC activity (number of TC), TC rainfall (rainfall associated with TC), and the annual cycle of the precipitation regime (Figure 10). For NCA and EAS, the peak of the TC season for the Northern Hemisphere (September) exhibited the higher TC contribution (11%). By contrast, the EAF domain displayed the highest monthly TC contribution (9%) in May, at the end of the wet season, despite a seven-fold difference in TC rainfall with the peak of the TC season in January. For that particular situation, results have shown the important year-to-year variability of monthly TC rainfall contribution and that one should be careful with regard to any generalization (Figure 11). Therefore, the derivation of robust statistics for the monthly TC contribution would require a longer period of observation.

[32] A selection of 50 urban areas (Appendix A) showed that Taipei (Taiwan), Manila (Philippines), Port Hedland (Australia), Saint Denis (La Réunion), and La Paz (Mexico) had the greatest average annual TC contribution (>20%) among all the cites selected (Figure 12). However, similar TC contributions corresponded to different situations; important TC rain for the first two (TAIP: 572 mm yr−1, MANI: 488 mm yr−1), average TC rain (PHED: 131 mm yr−1, REUN: 96 mm yr−1) and non-TC rain (PHED: 300 mm yr−1, REUN: 346 mm yr−1) for the third and fourth (PHED, REUN), and a relatively moderate TC rain (102 mm yr−1) with relatively low non-TC rainfall (164 mm yr−1) for the last located in an arid zone (LPAZ). A closer look at particular subregions such as Baja California Sur (arid) and Taiwan (marine tropical) brought useful information regarding the link between TC activity and water resource availability (Figure 13). Results seemed to indicate that TC rainfall can trigger drought relief but are not a sufficient condition as the example of Taiwan showed. However, further work would be necessary to investigate the longer-term impacts of TCs and the influence of the annual variability of cyclonic activity on water resources or droughts over cyclone-prone areas.

[33] Due to the global coverage provided by satellites, the availability of a consistent precipitation data set allowed for a systematic comparison between the different basins. There are three main sources of limitations in the precipitation data set used in this study: (1) The uncertainties related to the different land/ocean rainfall retrieval algorithms, (2) The coarse spatial-temporal resolution that cannot fully capture the variations of TC rainfall, and (3) The non-uniformity of the biases corrections that depend on the rain gauge availability. However, to the best of our knowledge, this study is the first that quantifies the TC rainfall contribution over land for the different basins around the world. Future work will consist of quantifying precipitation extremes in relation with TC activity.

Appendix: A

Table A1. Annual TC Contribution and TC Rain for 50 Urban Areas Around the World
   TC Contribution (%)TC Rain (mm yr−1)
CityIDDomAvgStdvMaxMinYear Max1AvgStdvMaxMinYear Maxa
  1. a

    If no data, the year for maximum TC rainfall is similar to that of maximum TC contribution (1). For each domain, the maximum in each category is reported in bold.

Houston (TX-USA)HOUSNCA8.68.925.71998120.0138.7420.6
New Orleans (LA-USA)NWORNCA6.85.818.62005113.2101.8311.0
Miami (FL-USA)MIAMNCA10.19.833.72005151.8164.8552.0
Key West (FL-USA)KEYWNCA15.513.146.70.42005169.6187.0662.53.7
Charleston (SC-USA)CHARNCA8.26.118.20.2200499.466.4203.02.71999
Wilmington (NC-USA)WILMNCA9.35.822.81.11999141.7134.4516.116.8
New York (NY-USA)NYCTNCA3.33.89.9200537.844.1121.8
Mexico City (Mexico)MEXINCA2.63.18.4199925.528.978.30.2
Tampico (Mexico)TAMCNCA5.06.517.1200566.8102.5279.2
Merida (Mexico)MERINCA5.34.412.80.1200064.352.3132.20.82002
La Paz (Mexico)LPAZNCA38.421.168.02.51998102.392.9252.97.62001
Puerto Vallarta (Mexico)PTVANCA10.75.618.21.31998130.180.4277.017.3
Guadalajara (Mexico)GUADNCA3.42.89.5-200631.628.299.0
Acapulco (Mexico)ACAPNCA11.77.825.11.52006181.7117.2419.829.3
Oaxaca (Mexico)OAXANCA6.05.617.10.1200347.845.9130.10.72005
Havana (Cuba)HAVANCA9.76.523.71.62005140.6117.7461.422.6
Port-au-Prince (Haiti)HAITNCA14.013.751.40.12008147.8196.1729.00.6
Santo Domingo (Dominican Rep.)SDMGNCA11.87.424.20.22008168.7121.8404.62.4
Guatemala City (Guatemala)GUATNCA2.33.713.0199837.965.9232.2
Tegucigalpa (Honduras)TEGUNCA4.14.815.5199858.380.7271.4
Managua (Nicaragua)MANANCA4.36.221.8199862.7122.1433.3
Nassau (Bahamas)NASSNCA7.98.120.2200584.275.9213.11999
San Juan (Puerto Rico)STJNNCA10.07.020.72004125.6101.0332.10.11998
Pointe-à-Pitre (Guadeloupe-France)PTPRNCA9.78.327.60.31999132.7126.2428.93.7
Fort-de-France (Martinique-France)FTFRNCA9.07.527.80.1200170.444.0138.10.6
Kingston (Jamaica)KSTNNCA14.312.839.12004204.9200.3526.62005
Osaka-Kobe-Kyoto (Japan)OSKKEAS8.16.223.51.12004135.7129.2492.520.1
Tokyo-Yokohama (Japan)TOKYEAS9.06.222.00.22005158.393.5315.13.0
Taipei (Taiwan)TAIPEAS29.914.553.06.72004572.0364.21006.5114.8
Guangzhou-Foshan-Shenzhen (China)GUFSEAS15.29.736.01.31999297.0218.5701.428.02008
Shanghai (China)SHANEAS6.27.723.7200575.688.4268.1
Fuzhou (China)FUZHEAS15.38.929.62.22005199.5120.5397.715.2
Busan (South Korea)BUSNEAS7.16.016.5200492.274.5186.92006
Seoul-Incheon (South Korea)SEOUEAS6.55.520.1199994.888.2334.60.1
Ho-Chi-Min (Vietnam)HOCMEAS1.33.110.8199820.744.3153.9
Manila (Philippines)MANIEAS20.513.240.30.12006487.9318.7919.81.02000
Guam (USA)GUAMEAS13.610.130.42004233.5212.5699.7
Dhaka (Bangladesh)DHAKSWA5.14.516.92004109.1105.6408.3
Kolkata (India)KOLKSWA7.14.715.50.42008137.986.6312.26.3
Chennai (India)CHENSWA7.78.423.3200897.9137.3407.22005
Colombo (Sri Lanka)COLMSWA0.71.55.3200516.839.6138.4
Karachi (Pakistan)KARASWA5.821.761.6199910.018.653.22004
Masqat (Oman)MASQSWA10.216.958.620079.231.9110.4
Port Hedland (Australia)PHEDOCE30.423.071.80.42003130.9115.7311.91.22007
Darwin (Australia)DARWOCE10.38.027.00.22007178.8151.9499.52.0
Port Moresby (Papua New Guinea)PMOROCE3.85.816.2200747.677.6235.9
Nouméa (New Caledonia-France)NOUMOCE6.06.918.0200370.382.9212.1
St Denis (La Réunion-France)REUNEAF21.717.047.10.1200795.877.4225.90.62006
Antananarivo (Madagascar)ANTAEAF8.76.523.42000124.691.1321.8
Beira (Mozambique)BEIREAF4.96.721.6200358.683.9251.4

Appendix: B: Comparison of Results Obtained With TMPA V7 and V6

[35] In this section we compare results obtained with the Version 7 (this study) and Version 6 [Prat and Nelson, 2013, among others]. As mentioned earlier, Version 7 uses the surface precipitation gauge analysis from the Global Precipitation Climatology Centre (GPCC). Figure B1 shows that for non-TC rainfall, TMPA V6 displays a negative bias when compared to V7 with the largest differences observed for the NCA and the SWA domains with a linear regression coefficient a=0.77 and a=0.79 respectively. Furthermore, land and coastal (i.e., pixels with 0<land<100%) pixels display a comparable bias between V6 and V7. While, important differences can be found for non-TC rainfall, the differences for TC-rainfall are less emphasized (0.85<a <1.03). This is due to the fact that the areas where the most important differences between V7 and V6 were observed (i.e., convergence of the ITCZ, Northwestern United States for NCA and Bangladesh/Burma for SWA) do not experience TC activity (NCA) or only marginal activity (SWA). However, there were important differences observed between V6 and V7 along the coastal areas surrounding the Gulf of Mexico (NCA), which were reflected by the lowest linear regression coefficient for TC rainfall (a=0.85). In addition, due to rain gauge adjustment being performed on a monthly basis, the impact of TC for a given month will be washed out. For the TC contribution, the linear regression coefficient remained relatively close to 1 (0.93<a<1.08) regardless of the domain considered. To summarize, the differences between V6 and V7 are mostly significant in terms of total and non-TC rainfall with a negative bias for V6 when compared to V7. The differences in terms of TC rainfall and TC contribution were less important.

Figure B1.

Scatterplots for non-TC rainfall, TC rainfall, and TC contribution for TMPA 3B42V7 when compared with TMPA 3B42V6. Distinction is made between land and coastal pixels. Acronyms for the domains considered: North and Central America (NCA), East Asia (EAS), South and West Asia (SWA), Oceania (OCE), East Africa (EAF). Note the different axis ranges selected for representation purpose.

Acknowledgments

[36] This research was supported by the NOAA/NCDC Climate Data Records and Science Stewardship Program through the Cooperative Institute for Climate and Satellites—North Carolina under the agreement NA09NES4400006. The authors would like to thank Alisa Young, and Carl Schreck for helping with the internal review process. The authors are grateful to three anonymous reviewers for valuable comments and suggestions.

Ancillary