A global spatiotemporal analysis of inland tropical cyclone maintenance or intensification


  • Theresa K. Andersen,

    Corresponding author
    1. Department of Geography/Atmospheric Sciences, University of Georgia, Athens, GA, USA
    • Correspondence to: T. K. Andersen, Department of Geography/Atmospheric Sciences, University of Georgia, Athens, GA, USA. E-mail: tkande@uga.edu

    Search for more papers by this author
  • J. Marshall Shepherd

    1. Department of Geography/Atmospheric Sciences, University of Georgia, Athens, GA, USA
    Search for more papers by this author


Forecasting tropical cyclone (TC) intensity changes over land is complicated by interactions of various surface and atmospheric features. Due to generally unfavorable conditions, many TCs weaken and decay soon after landfall. In some cases, TCs may also transition to extratropical cyclones (ETs). Despite the absence of oceanic forcing, a number of TCs have been observed to maintain or increase strength inland, termed “tropical cyclone maintenance or intensification’ (TCMIs). This study identifies the environments and characteristic features of TCMIs and explores physical processes that may help to produce an atmosphere conducive for tropical systems. The objectives are to compile an inland TC dataset over a 30-year period, quantify TC traits that may relate to maximum strength over land, and analyse surface and atmospheric conditions leading up to intensification. Of 227 inland TCs globally, 45 maintained or increased strength inland: 17 cold-core (ET), 16 warm-core (TCMI), and 12 hybrid cases. Analysis of synoptic conditions indicates that TCs persist when low-level temperature gradients are weak. Soil moisture gradients were in the vicinity of the cyclones at the time of intensification and may be forcing the TCMIs via increased surface latent heat flux (LHF). The area-averaged LHF threshold is found to be around 70 W m−2 for TCMI occurrence. In the 2 weeks leading up to each TCMI, the LHF tends to be higher than average over the intensification regions and provides further evidence of land surface forcing.

1. Introduction

Current climate change may impact the frequency, intensity, and distribution of tropical cyclones (TCs) as the water cycle accelerates and weather patterns shift (Shepherd and Knutson, 2007). Forecasting TC intensity changes has proved difficult, especially over land where tropical systems may interact with a variety of synoptic, mesoscale, convective, and topographic features. Most tropical systems weaken over land owing to increased friction, moisture loss, and general baroclinicity. However, there have been cases of TCs actually maintaining or increasing strength inland far from the preferred oceanic energy source. Such events have occurred in the United States (Arndt et al., 2009), India (Chang et al., 2009), China (Chen, 2012), and Australia (Emanuel et al., 2008). It is hypothesized that these TCs are forced from below by soil moisture and energy fluxes, termed the ‘brown ocean’ effect.

This study identifies the environments and characteristic features of TCs maintaining warm cores inland and explores physical processes that may help to produce a moisture-rich boundary layer. Specifically, the objectives are to (1) compile an inland TC dataset over a 30-year period and further classify events based on intensity and thermal structure, (2) quantify path length, over-ocean intensity, and other cyclone traits that may relate to maximum strength over land, and (3) analyse surface and atmospheric conditions to understand physical mechanisms associated with maintaining a warm-core tropical low pressure system.

The remainder of the paper is organized into four sections. Section 'Background' provides background information on TC classification and soil moisture–atmospheric feedbacks. Section 'Data and methodology' lists the datasets used and methodology for assessing inland TCs. The results of the three objectives are presented in Section 'Results', followed by a discussion and concluding remarks in Section 'Discussion and conclusion'.

2. Background

Over land, the defined spectrum of TC types is limited as there is no comprehensive post-landfall scale in place. The traditional Saffir-Simpson scale, based on wind speed, does not sufficiently account for total damage following landfall (Senkbeil and Sheridan, 2006; Kantha, 2010). A supplemental classification system for post-landfall hurricanes in the United States has been proposed owing to the increasing urbanization and population along the Atlantic coast (Senkbeil and Sheridan, 2006). An updated hurricane classification system would use air pressure, maximum sustained winds, gust score, storm surge, duration, and precipitation for post-landfall identification of hurricane types. Other studies have focused on standardizing wind speed scales for global applicability (Dotzek, 2009). Additionally, research contributed to the World Meteorological Organization (WMO) Seventh International Workshop on Tropical Cyclones (Gyakum, 2010) describes several types of tropical systems that deviate from the Saffir-Simpson scale (i.e. hybrid, subtropical, and frontal). Research from the present study may add another type, termed tropical cyclone maintenance or intensification (TCMI). TCMIs describe tropical systems that maintain tropical characteristics over land where forcing is not from the ocean.

Soil moisture, an important component of the water cycle, has a direct effect on the atmosphere through moisture and energy exchanges (Mahmood et al., 2006). The moisture supply and instability of the planetary boundary layer (PBL) are key elements for convective storm development. Studies have shown that differential heating caused by wet–dry soil boundaries can initiate a thermal circulation in a conditionally unstable atmosphere (Rabin et al., 1990; Hong et al., 1995). These wet–dry boundaries can form from previous rainfall, deforestation boundaries, or alternating bare soil-vegetation areas. Land cover boundaries, such as urban to agricultural, have been linked to convective cloud mass clustering (Brown and Arnold, 1998). Other studies have shown that wet soil alone, because of sensible and latent heat fluxes (LHFs), can transport energy and moisture into the atmosphere to enhance convective available potential energy (CAPE) (Clark and Arritt, 1995; Bosilovich and Sun, 1999; Lynn et al., 1998). These fluxes create ‘land–land breezes’ that help lift parcels to initiate convection (Hanesiak et al., 2004). Moisture transfer through plant evapotranspiration can also cause differential heating in the atmosphere, leading to local thermal circulations and convective instability (Chang and Wetzel, 1991; Pan et al., 1996; Hanesiak et al., 2004; Alonge et al., 2007; Iwasaki et al., 2008; Frye and Mote, 2010).

The energy contributed by the land surface may help sustain tropical systems through similar processes. Arndt et al. (2009), Kellner et al. (2012), and Evans et al. (2011) suggest that anomalously wet soils in Oklahoma aided in the reintensification of Tropical Storm Erin (2007) by creating a favourable thermodynamic environment with enhanced moist static energy and latent heat. Chang et al. (2009) found, using the weather research and forecast (WRF) model, that monsoon depressions (MDs) are sensitive to soil moisture (i.e. heavy rainfall events 1 week prior to landfall are associated with longer sustained MD intensity over land). Emanuel et al. (2008) cite that large vertical heat fluxes from moistened, hot sands in northern Australia can help intensify landfalling TCs. Sandy soil must be wetted from the cyclone itself, very recently wetted from prior rain events, or exhibit moisture persistence in order to impact TC intensity. Kishtawal et al. (2012) state that soil heat flux is an important consideration in assessing inland decay of TCs due to heat and moisture transfer. Chen (2012) found that frictional effects over land can increase radial wind convergence and convection helping to sustain typhoons that would otherwise decay over China. Alternatively, large reservoirs, lakes, and wetlands can supply energy to typhoons post-landfall.

Over land, synoptic-scale features interact with tropical systems often resulting in weakening of the cyclone or a transition to an extratropical system. The National Hurricane Center (NHC) defines extratropical transition (ET) by subjective satellite interpretation of storm structure within 1–2 d of transition (Hart and Evans, 2001). Transitioning cyclones cause damaging winds, floods, and coastal erosion hundreds of kilometres beyond the storm centre. An extratropical transitioning storm will develop an increasingly asymmetric temperature field and cold-core structure (geostrophic wind speed increasing with height) in response to the baroclinic atmosphere. In the North Atlantic, ETs are suggested to be more common than in the Pacific, particularly near the United States east coast and Canadian maritime areas. Superstorm Sandy is a recent example of tropical to extratropical transition as documented in the study by Shepherd (2012). The situations in which inland TCs are reinforced instead of transformed may occur when the environment mimics barotropic conditions (i.e. low shear and uniform temperature), or when peripheral weather systems contribute moisture and energy to the system. Low-level jets transport moisture and momentum, upper-level jets enhance outflow, and mesoscale vortices can merge with and strengthen TCs in some cases (Chen, 2012).

3. Data and methodology

3.1 Data

Geographic Information Systems (GIS) software, along with satellite-based datasets, provide the tools essential to spatially and temporally analyse TCs globally. The tracks and attributes (i.e. location, wind speed, and pressure) of TCs are obtained from International Best Track Archive for Climate Stewardship (IBTrACS, http://www.ncdc.noaa.gov/oa/ibtracs/) during the satellite era (1979–2008). IBTrACS merges multiple datasets from international meteorological centres into one cohesive archive for public use. The originating data centres applicable to this study are the US National Hurricane Center, Australian Bureau of Meteorology, Hong Kong Observatory, China Meteorological Administration, and India Meteorological Department.

Atmospheric and environmental data prior to and during TC events are obtained from NASA Modern Era Retrospective-Analysis for Research and Applications (MERRA, http://gmao.gsfc.nasa.gov/research/merra/intro.php). MERRA integrates multiple observing systems with numerical models to produce gridded variables. The atmospheric assimilation emphasizes the hydrological cycle, which helps minimize uncertainty in precipitation and inter-annual variability. The data include 3-hourly 600 and 900 hPa geopotential heights (m), 300 hPa maximum eastward wind component, monthly surface LHF (W m−2), monthly fractional topsoil layer wetness, and 3-hourly surface LHF (W m−2).

3.2. Methodology

The NHC defines landfalling as ‘the intersection of the surface center of a TC with a coastline’. The central points along the paths of TCs can be found in data records making it straightforward to determine the time of landfall. The time of reaching ‘inland’ is less well-defined because of the varying sizes of hurricanes and disorganized nature of weaker TCs. In determining TC size, Frank (1977) found the cyclonic, convergent inflow to be within a 4°–6° radius of the centre and Evans and Hart (2003) found the potential vorticity signature to be well within a 500 km radius. To ensure the energetics are not being driven by the ocean but still capture all potential TCMI cases, inland tropical cyclones (ITCs) are defined as those landfalling storms that subsequently have a central pressure measurement at least 350 km away from the nearest coast. This radius corresponds to the typical edge of outer rainbands based on research by Frank (1977).

ITC types are determined by the cyclone phase spacing technique of Evans and Hart (2003). Cyclone phase space is a useful tool for retroanalysis of ET (Kofron et al., 2010) and can help differentiate cold-core versus warm-core systems. This method computes the thermal wind for assessing the thermal nature of the storm at any given time. Here it is computed during maintenance (steady pressure and/or wind speed since landfall) or intensification (lowering pressure or increasing wind speed inland) (termed ‘MIs’). The warm-core and cold-core structures are distinguished with the 600–900 hPa thermal wind measurement:

display math(1)

where zmax is the maximum height value and zmin is the minimum height value at a pressure level between 600 and 900 hPa within the storm's radius. The radius is measured from centre of the surface cyclone at a given time. The method for sub-categorizing TCs is summarized in Figure 1.

Figure 1.

Schematic showing tropical cyclone sub-categorization.

After TCMI events are identified, the spatiotemporal distribution, maximum sustained wind speed, minimum central pressure, over-ocean intensity, and path length are assessed for common features using analysis of variance (ANOVA). Atmospheric pre-existing conditions for ET and TCMI are examined with the 850 hPa temperature range and 300 hPa maximum u-component wind speed using two-sample t-tests. Surface conditions are assessed with fractional soil wetness and surface LHF to better understand the physical mechanisms associated with maintaining TCs over land.

4. Results

4.1. Climatology

Of 227 ITC cases, 45 maintained or increased strength inland (MIs) (Table 1). Figure 2 shows the spatial distribution of the neutral, cold-core (ET), and warm-core (TCMI) cyclones based on thermal wind calculations. There were 17 ET events, which occurred mainly over North America, 16 TCMI cases occurring primarily over Australia, and 12 neutral/hybrid cases occurring over the United States, India, and China. India has the fewest events overall.

Table 1. The maximum sustained wind speed (knots), minimum central pressure (hPa), and 600–900 hPa thermal wind measurements for cyclones during maintenance or intensification inland
CycloneUTC/dateWind speed (knots)Pressure (hPa)Δ900–Δ600 (hPa)Core type
  1. Cold-core indicates the storm transitioned from tropical to extratropical cyclone (ET). Warm-core indicates the storm maintained tropical characteristics (TCMI). Neutral indicates the storm is a hybrid of both types.

Hope0600 8-9-197925<10Neutral
Freda0000 8-9-198417996−13Cold
Nelson1800 8-25-198519997<10Neutral
Winifred0600 2-3-198621999+14Warm
Bonnie0000 6-28-1986151014<10Neutral
Connie0600 1-22-198739992−26Cold
Irma0000 1-22-198730+20Warm
Gilbert1200 9-19-198825999−19Cold
Hugo0600 9-23-198940990−71Cold
1990226N190880000 8-16-199027995<10Neutral
1990232N220891200 8-24-199027994−12Cold
Daphne0000 2-20-199120+10Warm
1991208N210900000 7-30-199130988+25Warm
1991234N200891200 8-23-199125992<10Neutral
Nat0600 10-3-19911008+10Warm
Tim0600 7-12-199429992+10Warm
Annette0600 12-19-199435995−16Cold
Opal1800 10-5-199540986−15Cold
Fran0000 9-8-199630999+10Warm
Amber0600 8-31-19971002< 10Neutral
Olga0600 8-5-1999996+14Warm
1999219N210890600 8-8-199920992<10Neutral
Dennis1800 9-8-1999201005−10Cold
Sam1200 12-12-200044+17Warm
Terri0000 2-1-200141990−30Cold
Winsome0000 2-13-200125988<10Neutral
Wylva1200 2-18-2001988+21Warm
Abigail0000 3-3-200125992+10Warm
Barry0000 8-8-2001101016<10Neutral
2003206N210891200 7-26-200330988−10Cold
2003279N170840600 10-9-2003251000+23Warm
Fritz0600 2-14-200425992−19Cold
2004163N160901200 6-14-200425990<10Neutral
Frances1200 9-9-2004301001<10Neutral
Haima0600 9-16-200429993−25Cold
Arlene0600 6-14-2005201003−26Cold
Dennis0000 7-18-2005101010−18Cold
Emma1200 2-28-200629988+10Warm
Sepat0600 8-23-2007191002−10Cold
George0600 3-10-200730986+24Warm
2007172N150880000 6-23-2007201007+10Warm
Erin0600 8-19-200750995+19Warm
Gustav1800 9-4-2008201000−15Cold
Ike1200 9-14-200840987−19Cold
Two1800 9-18-200825996<10Neutral
Figure 2.

Tropical cyclones maintaining or increasing strength inland (1979–2008). Thermal wind analysis finds 17 ET (stars), 16 TCMI (squares), and 12 hybrid (circles) cases.

The years with the most ITCs do not necessarily produce the most TCMI cases. The 4 years with the highest counts of inland TCs (i.e. 1996, 1999, 2005, and 2008) actually have relatively low TCMI counts (Figure 3). Pearson's correlation coefficient is r = 0.22 for global ITCs versus all MIs and r = 0.12 for global ITCs versus TCMIs. A linear fit suggests that ITC frequency is increasing by 1 every 11 years possibly from better reporting or the positive phase of the Atlantic multidecadel oscillation (AMO).

Figure 3.

Time series of the total number of global inland TCs (solid), TCs maintaining strength or intensifying (dashed), and TCs maintaining/intensifying with warm-cores (TCMIs, dotted) between 1979 and 2008. Years with relatively high tropical cyclone frequency (peaks) do not necessarily produce more intensification events (r = 0.22) nor TCMIs (r = 0.12).

China and the United States have the most TCs trekking over land, but Australia exhibits the most TCMI cases over the study period (Figure 4). Therefore, TCMIs are not a function of the total number of storms in the region. The TCMIs at time of intensification inland over the four study regions, represented by hurricane symbols, are shown in Figure 5. The hurricane symbol size represents pressure fall (0 to 9+ hPa) and tone represents sustained wind speed increase (0 to 15+ knots) during the intensification process. Each cyclone is numbered and can be referenced to Table 2.

Figure 4.

Graph of global inland TCs and TCMIs during 1979–2008 by region. The United States and China have high overall counts but low TCMI counts, indicating the number of TCMIs is not a function of the total number of cyclones trekking over the region.

Figure 5.

TCMIs over the United States (top left), India (top right), China (bottom left), and Australia (bottom right). The hurricane symbol represents the location at time of inland maintenance or intensification. The symbol size represents the pressure fall (0–9 hPa) and tone represents the sustained wind speed increase (0–15+ knots) since landfall or previous maximum over land.

Table 2. Global tropical cyclone maintenance/intensification events (TCMIs) during 1979–2008. The pressure drop and wind speed increase during the inland intensification process (ranging 6–42 h) are listed
 CycloneTCMI location

(Lat, Lon)

Span of intensification (hours)Pressure change (hPa)Wind speed change (knots)
  1. The number in the first column corresponds to the cyclone shown in Figure 5.

1Winifred−20.9, 141.460+5
2Irma−17.3, 132.460
3Daphne−18.5, 131.86+5
41991208N2109022.5, 80.06−20
5Nat27.9, 116.86−4
6Tim29.5, 115.1600
7Fran42.9, −80.16−10
8Olga50.9, 131.16−4
9Sam−20.5, 130.018+11
10Wylva−18.6, 130.518−7
11Abigail−19.7, 124.742−60
122003279N1708424.0, 86.3600
13Emma−24.2, 117.16−20
14George−24.0, 122.012−20
152007172N1508818.0, 77.56−3+5
16Erin35.6, −98.812−12+30

Previous studies have defined rapid intensification as +30 knots per day based on 95th percentile of wind speed change (Kaplan and DeMaria, 2003) or −42 hPa d−1 based on 75th percentile pressure change (Holliday and Thompson, 1979). These rates are ocean-based calculations that may not necessarily apply to inland intensification; however, they help to gauge the magnitudes of values given in Table 2. For example, a 30 knot per day increase equals a 7.5 knot per 6-h increase, indicating at least one of the TCMIs rapidly intensified (i.e. TS Erin).

4.2. TCMI characteristics

The characteristics of TCMIs and paths are examined for any possible trends that could relate to strength over land. The over-ocean maximum wind speed for the dataset ranged from Tropical Depression to Category 3 hurricane. The pre- and post-landfall pairs of values for minimum pressure and maximum sustained wind speed indicate maximum strength over ocean does not control the strength over land (Figure 6). The mean path length for all TCMIs is 3820 km. Broken down by regions, mean path length is 5396 km (China), 3413 km (Australia), 2180 (India), and 5536 km (the United States). The regional path lengths correspond to the vastness of the originating ocean basin, and do not correspond to intensification strength. Generally, ET cases have the longest treks, TCMI cases have a wide range of track lengths, and neutral cases have the shortest treks. However, ANOVA reveals the path length differences between regions (F ratio = 2.23 and F crit = 4.81) and path length differences among the three cyclone types (F ratio = 2.40 and F crit = 4.30) are best explained by chance.

Figure 6.

Pre- and post-landfall values for minimum pressure (left) and maximum sustained wind (right) for TCMIs. TCMIs do not appear to be a function of storm strength over the ocean.

It has been suggested that landfalling TCs intensify over areas with large upward heat fluxes and enhanced thermal conductivity such as warm, swampy terrain, hot sands, and recently moistened soils (Emanuel et al., 2008). The soil types (NRCS, http://soils.usda.gov/use/worldsoils/mapindex/order.html), major biomes (NRCS, http://soils.usda.gov/use/worldsoils/mapindex/biomes.html), and land cover (NASA, http://earthobservatory.nasa.gov/Newsroom/view.php?id=22585) of the primary TCMI regions are as follows:

  • Eastern China: ultisols (red clay), temperate humid biome, and croplands and mixed forest.
  • South-central United States: mollisols, temperate semi-arid and temperate humid biomes, and grasslands.
  • Northern Australia: entisols (sandy), tropical semi-arid and desert temperate biomes, and open shrublands and savannas.
  • India: alfisols, tropical semi-arid and temperate semi-arid, and croplands.

Although China experiences the most hurricane landfalls, Australia has the most TCMIs. Northern Australian soil is sandy and may have relatively high sensible and LHFs during the hurricane season. Soils, climates, and vegetation vary regionally, and more in-depth modelling studies are needed to determine if surface–atmospheric feedbacks are similar across these regions.

4.3. TCMI environments

To maintain warm-core structures, it is hypothesized that TCMIs do not encounter strong wind shear and temperature gradients post-landfall (i.e. the atmosphere is baroclinically weak). To test this hypothesis, ET and TCMI atmospheric conditions (within a 750 km radius) at the time of intensification are analysed using two-sample t-tests. Analysis of the low-level temperature gradients indicates that ET environments exhibit significantly greater 850 hPa temperature ranges than TCMI environments (p = 0.01). It is proposed that ITCs persist when temperature gradients are weak, while ET is encouraged by local frontal boundaries. Analysis of the upper level winds indicates that the 300 hPa maximum u-wind component is on average 12 knots higher for ET environments, but not significantly different from TCMI environments (p = 0.16) (Table 3).

Table 3. To quantify baroclinicity, the 300 hPa eastward wind component and 850 hPa temperature range (over 1500 km in diameter) between ET and TCMI cyclones are statistically analysed
 300 hPa u-windmax850 hPa temprange
  1. The two-sample t-test for equal variances indicates that at the 95% confidence level, the 300 hPa maximum wind is not significantly different between ETs and TCMIs; however, the 850 hPa temperature range is significantly greater for ET cases than for TCMIs. These results support the hypothesis that TCMI events are more likely to occur in environments with relatively weak temperature gradients (i.e. lack of frontal boundaries).

Standard deviation19.727.52.96.1
p-value0.16 (>0.05)0.01 (<0.05)

TCs derive their energy from ocean surface evaporation. Latent heat release within the storm sustains the ‘heat engine’, allowing it to last up to several weeks (Kelley and Halverson, 2011; Shepherd, 2012). To remain warm-core structures over land, a similar moisture-rich environment may need to exist. Surface water content and associated LHF help to quantify the near-surface moisture and energy availability. The 1-month antecedent surface LHF (W m−2), 1-month antecedent fractional top soil layer wetness, and LHF at time of TCMI exhibit similar spatial patterns, an indication that LHF is a good approximation of surface moisture and that these trends are persistent on a weekly to monthly scale. Soil moisture gradients are evident in the vicinity of intensification regions, which is consistent with the ‘brown ocean’ concept (Figure 7). Over north-central Australia, three strong intensification events occurred consecutively: Sam in December 2000 (#9), Wylva in February of 2001 (#10), and Abigail in March 2001 (#11). Here, TCs are a primary source of precipitation and may provide the surface wetting to force subsequent TCMI events.

Figure 7.

One-month antecedent fractional top soil layer wetness for TCMI events. The arrow denotes the cyclone direction. Soil moisture gradients are evident in the vicinity of the intensification region.

Minimum, maximum, and area-averaged LHF for each TCMI at the time of maintenance/intensification are presented in Figure 8. The difference between the maximum and minimum values over the intensification region is over 400 W m−2 in some cases and can be used to quantify the LHF gradient. Although some cases appear to have low LHF, the magnitude should only be considered relative to the particular area. To put the magnitudes in context, time series of daily LHF over the intensification region (6° × 6°) in the 3 weeks leading up to the TCMI are examined. The mean value over the associated hurricane season is plotted for comparison. It can be seen from Figures 9 and 10 that these TCs are sustained following high LHF days to weeks prior. The high flux values are consistent with the peak of the hurricane season (when most TCMIs occur) as summer rainfall and abundant surface heating promote evaporation. The total surface precipitation flux (kg m−2s−1) confirms that significant rainfall events precede TCMIs.

Figure 8.

Maximum, minimum, and area-averaged surface latent heat flux (W m−2) for TCMIs at time of maintenance/intensification.

Figure 9.

Daily area-averaged surface latent heat (squares) and precipitation flux (shading) leading up to TCMI (cyclones 1–8) and associated hurricane season-averaged latent heat flux (dashed line) for comparison. The final time is TCMI occurrence.

Figure 10.

Same as in Figure 9, but for cyclones 9–16.

For each TCMI, the area-mean LHF over the 3-week antecedent period is compared to that of all weakening TCs that passed within a 3° radius of the TCMI latitude, longitude over the study period. Only TCs that occurred in a similar time of year (within 2 weeks) as the TCMIs are considered to facilitate a better comparison. Some regions experience more tropical activity overall; therefore, the number of TCs to compare vary among the TCMIs. The results indicate that North Atlantic (NA) and South Pacific/South Indian (SP/SI) regions at the time of TCMI have the highest or above average flux values as compared with instances of a weakening TC (Figure 11). West Pacific (WP) and North Indian (NI) TCMIs tend to exhibit average or below average LHF compared to weakening cyclones. On the basis of the results, it is suggested that there is an area-averaged LHF threshold near 70 W m−2 for TCMI occurrence.

Figure 11.

Area-averaged LHF (W m−2) for the 16 TCMI regions at the time of intensification (black squares). The grey line indicates the range of LHF for weakening tropical cyclones recorded in a similar area and time of year to the TCMI, respectively, between 1979 and 2008. The originating ocean basin is indicated to the right of each TCMI region. All South Pacific/South Indian and North Atlantic intensification regions during TCMIs exhibit the highest or higher than average LHF compared with times of weakening TCs. West Pacific and North Indian regions do not show any clear trend in LHF magnitude with respect to TC strength. It is suggested that there is an area-averaged LHF threshold near 70 W m−2 for TCMI occurrence.

5. Discussion and conclusion

As TCs make landfall, they often weaken owing to increased friction, wind shear, loss of moisture source, and temperature gradients. Those that remain strong tropical systems inland (i.e. TCMIs) are not well understood and are the focus of this study. An inland TC (ITC) database was derived from IBTrACS and further analysed on the basis of the observation that a system maintained or increased strength at least 350 km inland from the coast. Of 227 ITC cases, 45 maintained or increased strength inland: 17 cold-core (ET), 16 warm-core (TCMI), and 12 neutral/hybrid cases. Analysis of synoptic conditions at the time of intensification indicates that ITCs persist when temperature gradients are weak, while ET is associated with frontal boundaries. The upper-level winds are not significantly different between ET and TCMI events. Despite the wind shear, the jet stream may aid inland TC outflow in some cases (Chen, 2012).

The primary energy source for hurricanes is latent heat provided by the ample moisture originating from the ocean. Post-landfall, it is suggested that land surface feedbacks may provide the energy to continue sustaining the tropical system under certain scenarios. In previous studies, soil moisture gradients have been shown to increase the potential energy of the boundary layer, consequently enhancing instability and convection. In this study, spatial analyses of surface conditions revealed that soil moisture gradients, and associated LHF, were in the vicinity of the cyclones at the time of intensification. Analysis of LHF leading up to each TCMI reveals an overall tendency of higher than average values over the intensification regions. A TC may be more likely to intensify inland during the peak of the hurricane season when other TCs or rainfall events directly precede it.

Comparisons of each TCMI to weakening TCs found over the same region between 1979 and 2008 reveal that the intensification region consistently has higher surface LHF during an intensifying cyclone than during a weakening cyclone in the United States and Australia. These regions may exhibit stronger land surface–atmospheric feedbacks that relate to TC maintenance inland; however, comprehensive analysis of atmospheric conditions would be required to isolate the role of surface LHF. The WP and NI weakening cyclone events used here may have experienced strong synoptic forcing that encouraged TC decay despite high LHF. For all regions, an important finding is the area-averaged LHF threshold near 70 W m−2 for TCMI occurrence.

The criteria for ‘warm-core’ and ‘inland’ used here limit the sample size of TCs for statistical analysis and may exclude events that were influenced by land surface forcing. However, the 16 TCMI cases identified showed some clear trends and commonalities that merit additional research. Future objectives using a finite element soil heat flow model and mesoscale numerical weather prediction model will help to differentiate land surface effects from larger atmospheric effects on inland TCs.


This research was funded by the NASA Earth and Space Science Fellowship.