Quantifying the contribution of tropical cyclones to extreme rainfall along the coastal southeastern United States



[1] Analysis of a unique satellite precipitation dataset coupled with an extensive database of storm tracks are used to develop a parameter called the “millimeter-day (MD).” MD analysis in 4 mini-basins near coastal southeastern United States reveals that September and October account for the largest number of extreme rainfall days (e.g. “wet millimeter-days” or MD > 0) during the 1998–2006 Atlantic hurricane seasons. Tropical cyclone (TC) days are more likely to produce “wet millimeter days” than non-TC days, and category 3–5 hurricane days (e.g., major hurricanes) produce the wet millimeter-days of largest magnitude. Major hurricanes produce the most extreme rainfall days, but tropical depression/storm days contribute most significantly to cumulative seasonal rainfall (8–17%, basin-dependent) due to frequency of occurrence. Thus, the influence of major hurricanes on rainfall may be most apparent in extreme daily events while weaker storms may be more critical for assessing trends in cumulative seasonal rainfall.

1. Introduction

[2] An important debate is ongoing to determine whether human-induced global warming or natural variability is leading to an increase in hurricane frequency or intensity [Shepherd and Knutson, 2006]. Pielke et al. [2005] noted that by 2050 the Intergovernmental Panel on Climate Change (IPCC) expects that every additional dollar in damage caused by global warming-enhanced tropical cyclones (TCs) will be dwarfed by an additional $22 to $60 dollars of increased damaged due to population growth and wealth alone. Rappaport [2000] found that freshwater flooding caused more than half of the 600 deaths associated with Atlantic TCs over the period 1970–1999 in the United States (US). Therefore, if TC activity in the climate system is increasing, it is reasonable to ask if this results in a small-scale acceleration of the water cycle [Webster et al., 2005] through increased cumulative rainfall. Numerical modeling studies suggest that ongoing climate change should affect TC activity and precipitation [Knutson and Tuleya, 2004], but Groisman et al. [2004] found no trends in US tropical storm and hurricane precipitation. Before we can accurately establish whether trends in tropical cyclone rainfall are evident, we must establish an understanding of TC contributions to cumulative rainfall, particularly over ocean basins where they tend to be most intense.

[3] Wu et al. [2007], using rain gauge data on China's Hainan Island, determined that TCs accounted for more than one-third of the total precipitation for the period 1962–2005. Easterling et al. [2000] noted the significant contribution of TCs to rainfall in parts of the northeastern US. Rodgers et al. [2001], using monthly passive microwave rainfall estimates over very large basins, found that that TCs contribute 4%, 3%, and 4% to the western, eastern, and entire North Atlantic cumulative rainfall, respectively. However, they noted some regional areas near Puerto Rico and Africa where contributions were as high as 30%. Larson et al. [2005] found that landfalling TCs account for 15–20% of rainfall along the US Gulf and Mexican coasts.

[4] This study (1) established a new quantitative metric for extreme rainfall days associated with tropical systems, the “millimeter-day” (MD), using the model of the heating/cooling degree-day concept, (2) provided a seasonal analysis of MD, (3) investigated the correlation between MD and tropical cyclone (TC) days, and (4) quantified the contribution of TC days and storm categories to cumulative basin rainfall during the 1998–2006 hurricane seasons.

2. Methodology

[5] Satellites provide a viable capability for monitoring TC rainfall over oceans, yet the availability of reliable and long-term datasets is limited. Since 1997, satellite datasets from the Tropical Rainfall Measuring Mission (TRMM) satellite and other systems have been providing unprecedented capabilities for precipitation measurement [Kummerow et al., 2000]. One of the standard datasets available is a 3-hourly, 0.25-degree product based on the multi-satellite precipitation analysis (TMPA) described by Huffman et al. [2007]. The daily product is composed of available microwave (e.g., TRMM microwave imager, Special Sensor Microwave Imager (SSM/I), Advanced Microwave Scanning Radiometer (AMSR) and Advanced Microwave Sounding Unit (AMSU)) and calibrated infrared (IR) estimates. We employ the daily-accumulation version of the TMPA and the National Hurricane Center's (NHC's) North Atlantic hurricane database (HURDAT) that provides “best” determination of track and intensity in a post-season analysis of the tropical cyclones.

[6] Figure 1a illustrates the 4 near-coastal mini-basins used in the study: South of Houston Mini-Basin (SHMB), South of New Orleans Mini-Basin (SNOMB), East of Miami Mini-Basin (EMMB), and South of North Carolina Mini-Basin (SNCMB). Each basin, climatologically, experiences TC activity and is near populated coastal areas that experience landfalling storms (Figure 1b). As an example, Figure 1a presents TMPA-cumulative daily rainfall associated with TCs during the anomalously active 2005 hurricane season. Using the NOAA Coastal Services Center hurricane-tracking tool interface to HURRDAT (http://maps.csc.noaa.gov/hurricanes/index.htm), a database of TC-days was constructed for the period 1998–2006. A TC-day for a basin was defined as a day in which the center of a tropical depression, tropical storm, or category 1–5 hurricane was located within 150 nautical-miles (e.g. 277 km) on the tracking tool of the center of the basin. The 277 km radius is assumed to approximate, though not exactly, the area of the 5°x 5° rainfall basins. This approach is consistent with previous approaches [Cerveny and Newman, 2000; Rodgers et al., 2001; Larson et al., 2005]. For example, after allocating infrared-based satellite rainfall estimates into 2.5° cells, Cerveny and Newman [2000] found that the center grid cell explains nearly 70% of the variance in tropical cyclone rainfall. All of these approaches assume some level of precipitation symmetry that will likely lead to errors.

Figure 1.

(a) Location of study mini-basins with 2005 hurricane season rainfall (mm) as calculated by the TRMM multi-satellite precipitation analysis. (b) Total number of hurricane strikes by county (courtesy of National Hurricane Center website, http://www.nhc.noaa.gov).

2.1. Calculation of the Millimeter Day

[7] We developed a metric called the millimeter-day (MD) to quantify extreme rainfall contributions from TCs. The millimeter-day (MD) was modeled on the concept of heating (HDD) and cooling (CDD) degree-days. HDD and CDD are derived from daily temperatures and are used to quantify demand for energy needed to heat or cool residences or buildings. A degree-day is typically computed as the arithmetic difference (18.3°C – average daily temperature). A degree-day greater (less) than zero is a HDD (CDD) and indicates that some level of heating (cooling) is required as determined by historical regression analysis between temperature and energy consumption [Quayle and Diaz, 1980]. Similarly, the MD was computed as the arithmetic difference (average daily precipitation-28.9 mm). The average daily precipitation values are computed as the area-averaged TMPA daily values in each mini-basin. The base value of 28.9 mm is the mean daily precipitation at Mt. Waialeale, Hawaii for the period October 1992–July 2007 as computed from rain gauge data at Mt. Waialeale. The data was acquired from the United States Geological Survey (USGS) National Water Information System website (http://waterdata.usgs.gov/nwis/). The median annual rainfall on Mount Waialeale is 11415 mm, which makes it one of the Earth's wettest locations [Ramage and Schroeder, 1999]. We chose this location to represent a global, absolute standard for extreme precipitation rather than basin-dependent variations from a long-term mean. The MD is a useful metric for an extreme rainfall day and is more meaningful to the non-meteorological user trying to asses whether 20 mm/day is extreme or not, for example. MD is also well-suited for investigating long-term trends in extreme rainfall events.

2.2. Monthly Classification and Statistics of Wet Millimeter Days

[8] For each mini-basin, MD values (N = 3286 days, 1 January 1998 to 31 December 2006) were computed from the satellite-based TMPA. MD values > 0 were defined as “wet millimeter-days (WMD).” The sum (WMDSUM) is the total number of WMD values > 0 in all four mini-basins. WMDSUM is binned into monthly categories. The mean (equation image) and standard deviation of monthly WMDSUM values are presented in Table 1. Analysis of variance (ANOVA) testing was used to establish if the mean monthly values are significantly different. Additionally, we partitioned the data into two samples: the sum of (equation image) values for (1) hurricane season months (June–November) and (2) non-hurricane season months (December–May) for each mini-basin. A student's t-test, assuming unequal variances, was used to determine if (equation image) during hurricane season months were statistically different from non-hurricane season months.

Table 1. Statistics for All Four Basins Aggregated by Month for Wet Millimeter Days (WMD)
MonthCumulative WMD DaysMean Number of WMD DaysStandard Deviation

2.3. Coupling Wet Millimeter Days With Tropical Cyclone Days

[9] Using MDs from all four mini-basins (N = 6588) spanning the 1998–2006 hurricane season, we conducted a statistical rank-percentile analysis to identify the MDs with the largest magnitude. The distribution of MDs during the hurricane season (1998–2006) was highly skewed towards the negative tail of the distribution with WMDs (e.g., positive MDs) accounting for only the top 1.9% of the sample. We hypothesized that TC-days account for the WMDs with the largest magnitude (e.g., the most extreme rain days). To test this hypothesis, we examined the sample of MDs to extract what TC categories were associated with extreme rain days. To differentiate the contribution of TC-days to both WMD and cumulative rainfall, we statistically quantify what storm category (tropical depression/storm (TD-TS), category 1–2 hurricane (Cat1–2), or category 3–5 hurricane (Cat 3–5)) contributes most significantly to hurricane season cumulative rainfall.

3. Results and Discussion

[10] Figure 2 displays the monthly distribution of WMDs for the study period of record for each mini-basin. September and October account for the largest percentage of WMDs for three of the four mini-basins while SNOMB reveals a secondary peak in June. Table 1 lists mean (equation image) and standard deviation of monthly WMDSUM for the four basins. ANOVA testing indicates that the high-WMD months are significantly different from the other months at the 95% level (e.g., the F-ratio, 4.6, is greater than the F-critical value, 2.06).

Figure 2.

Monthly distribution of “wet millimeter days” for each mini-basin.

[11] September and October are two of the most active months for TCs in the Atlantic Basin [Neumann, 1993] so we hypothesize that TC days account for the large number of WMDs. Student's t-test confirms that the difference in number of WMDs during the hurricane and non-hurricane season months are significant at greater than the 95% level.

[12] To further test our hypothesis, we aggregated all MDs and TC-days for the four mini-basins and ranked their values. Because the distribution was highly skewed, we chose the top quartile of the distribution as a representative threshold for analysis. We filtered the data in the top quartile (N = 1647) to create four samples MDTD/TS (MD values on tropical depression/storm days), MD12 (millimeter-day values on category 1–2 hurricane days), MD35 (MD values on category 3–5 hurricane days), and MDTC (MD values for all tropical cyclone days). It is worth reiterating that MD values greater than 0 are considered “wet millimeter days” or WMDs. The box-whisker plot (Figure 3) indicates that Category 3–5 hurricane days (e.g., major hurricanes) produce the largest magnitude WMDs in the 4-basins (median = 20.56) as a function of the MD metric. The results are consistent with Marks et al. [2002], who used a rainfall climatology and persistence model to show that stronger storms offshore tend to produce more rainfall. Lonfat et al. [2004] also found, using 260 global TCs sampled by the TRMM microwave imager, that azimuthal mean rain rates were larger (smaller) in category 3–5 TCs (tropical storms).

Figure 3.

Box and whisker plot illustrating key descriptive statistics for the top quartile of MD days (N = 1647) during different storm-category days.

[13] As hypothesized, TC (non-TC) days are also more (less) likely to produce WMDs. This result is important because recent studies [e.g., Webster et al., 2005] have shown that the frequency of major hurricanes in the Atlantic Basin has increased recently. Our results suggest that MD may be an alternative metric for quantifying extreme precipitation variability and trends related to TCs over ocean basins, at coastal locations, and inland.

[14] It is important to note that while major hurricanes are responsible for the largest magnitude WMDs, tropical depression/tropical storm days are the most significant contributor to the cumulative hurricane season (1998–2006) rainfall budget in the basins. TC-days were classified by category for each mini-basin and an analysis of cumulative rainfall per basin was conducted. Table 2 indicates that a relatively small percentage of TC-days (1.8%) in the sample (N = 1647 per basin) contributed, on average, nearly 13% of the total rainfall during the hurricane season (1998–2006). The TC-day contribution ranged from nearly 8% in the SNCMB to 17.5% in the SNOMB, which is consistent with Larson et al. [2005]. Of the TC-days, tropical depression/storm days are the most frequently occurring and thus, contribute a larger percentage of cumulative rainfall than major storms.

Table 2. Contribution of Storm Category Days to Cumulative Rainfall in the Four Mini-Basinsa
  • a

    Total rainfall sample days (N) = 1647.

Cat 1–21.4%1.6%4.1%2.5%2.4%
Cat 3–50.8%6.1%4.4%2.1%3.4%
All storms12.6%17.5%13.4%7.7%12.8%

4. Conclusions

[15] The apparent increase in intense hurricane activity in the Atlantic basin is receiving considerable attention. It is reasonable to ask if more intense hurricane activity manifests itself in increased precipitation totals, which may signal an accelerated water cycle and more freshwater flux to ocean basins. However, more information is still required to understand the net contribution of tropical cyclones to ocean basins. We developed the millimeter-day (MD) as a metric to quantify the contribution of TCs to extreme rainfall events. Our results revealed extreme rainfall days or “wet millimeter days (WMDs)” are most likely in September and October during the peak of the hurricane season. Further, we found that major hurricanes are strongly correlated with the largest magnitude WMDs (e.g. cumulative daily rainfall event) during the TC season (1998–2006). TC-days accounted for nearly 8–17% (mean 12.8%) of cumulative rainfall during the hurricane season, depending on the basin. However, tropical depression/storm days contributed most significantly to basin cumulative rainfall for the season due to frequency of occurrence. The results suggest that TCs can contribute a significant portion of rainfall in an ocean basin but an increase in major TCs may be apparent in extreme daily events rather than the cumulative seasonal totals. This may mean that a trend in TC-related rainfall may not be apparent if a weaker regime of storms like tropical depressions/storms are not increasing in frequency of occurrence.

[16] We plan to apply the MD metric to land-based rainfall measurements that span significantly longer periods than the satellite record. MD analysis could play a role in future studies investigating acceleration of regional and global water cycles. This study should also be replicated for even longer time periods over the same (and other global) basins though accurate, microwave-based precipitation data (daily, <1 degree) over the oceans are scare before the TRMM era. Additionally, the results may spur new research directions in atmosphere-ocean interactions since TC-days are an apparent source of enhanced freshwater flux to the ocean.


[17] We are grateful to GPM Project Science Office (NASA/GSFC) and the NASA Precipitation Measurement Missions program (NASA/HQ) for related funding and helpful reviews by Josh Durkee, John Frye, and Matt Lacke.