DECEMBER 2013 A N D E R S E N E T A L . 2797

Quantifying Surface Energy Fluxes in the Vicinity of Inland-Tracking Tropical Cyclones

THERESA K. ANDERSEN,DAVID E. RADCLIFFE, AND J. MARSHALL SHEPHERD University of Georgia, Athens, Georgia

(Manuscript received 18 January 2013, in final form 9 July 2013)

ABSTRACT

Tropical cyclones (TCs) typically weaken or transition to extratropical cyclones after making landfall. However, there are cases of TCs maintaining warm-core structures and intensifying inland unexpectedly, referred to as TC maintenance or intensification events (TCMIs). It has been proposed that wet soils create an atmosphere conducive to TC maintenance by enhancing surface latent heat flux (LHF). In this study, ‘‘HYDRUS-1D’’ is used to simulate the surface energy balance in intensification regions leading up to four different TCMIs. Specifically, the 2-week magnitudes and trends of soil temperature, sensible heat flux (SHF), and LHF are analyzed and compared across regions. While TCMIs are most common over northern , theoretically linked to large fluxes from hot sands, the results revealed that SHF and LHF are equally large over the south-central . Modern-Era Retrospective Analysis for Research and Applications (MERRA) 3-hourly LHF data were obtained for the same HYDRUS study regions as well as nearby ocean regions along the TC path 3 days prior (prestorm) to the TC appearance. Results indicate that the simulated prestorm mean LHF is similar in magnitude to that obtained from MERRA, with slightly lower values overall. The modeled 3-day mean fluxes over land are less than those found over the ocean; however, the maximum LHF over the 3-day period is greater over land (HYDRUS) than over the ocean (MERRA) for three of four cases. It is concluded that LHF inland can achieve similar magnitudes to that over the ocean during the daytime and should be pursued as a potential energy source for inland TCs.

1. Introduction precipitation and soil moisture were unusually high preceding most TCMI cases while atmospheric fea- Landfalling tropical cyclones (TCs) bring storm surge tures amenable to extratropical transition were absent flooding, high winds, heavy precipitation, and tornadoes (i.e., weak baroclinicity). Sixteen cases were observed to coastal regions. After landfall, a TC may dissipate or in northern Australia, eastern China, India, and the transition to an (i.e., cold-core, United States (Andersen and Shepherd 2013). Further low-pressure system) by merging with the midlatitude understanding of the environment conducive to intensi- flow. Alternatively, TCs may maintain their tropical fication events can improve forecasting and draw atten- characteristics (i.e., warm-core, low-pressure system) and tion to this oft-overlooked stage of tropical cyclones. produce hazardous conditions for inland areas. Andersen Tropical cyclones are warm-core, low pressure sys- and Shepherd (2013) have suggested that anomalously tems that develop over the tropical oceans. The North moist soils in the vicinity of the TC enhance the surface Atlantic, north and south Indian, South Pacific, and west latent heat flux (LHF), providing energy to the cyclone Pacific Oceans are ideal development regions when the that would otherwise be lacking over land. This process sea surface temperature (SST) is .26.58C and the at- is characterized as the ‘‘brown ocean’’ effect. A global mosphere is characterized by weak vertical wind shear, analysis of TCs maintaining or increasing in strength conditional instability, enhanced midtroposphere rela- inland (TCMIs) over a 30-yr period revealed that tive humidity, and enhanced lower troposphere relative vorticity (Gray 1968). Latent heat flux from the warm oceans is a primary contributor to TC intensification Corresponding author address: Theresa Andersen, Department of Geography, University of Georgia, Room 31A, Athens, GA (Kleinschmidt 1951; Miller 1958; Malkus and Riehl 30602. 1960; Liu et al. 2012). Latent and sensible heat trans- E-mail: [email protected] ferred from the ocean surface along with potential energy,

DOI: 10.1175/JAMC-D-13-035.1

Ó 2013 American Meteorological Society Unauthenticated | Downloaded 09/23/21 02:43 PM UTC 2798 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 52

FIG. 1. Schematic of the brown ocean effect. Anomalously wet soil moisture conditions enhance evaporation and latent heat release. In a weakly baroclinic environment, the energy can help to sustain or intensify an inland .

increase the moist static energy of the boundary layer. Kellner et al. (2012), and Evans et al. (2011) suggest Convection converts moist static energy to kinetic en- that saturated soils and moisture transport in Okla- ergy, thereby intensifying the primary circulation of the homa contributed to the reintensification of Tropical TC. Typically, tropical systems weaken or transition to Storm Erin in 2007 by enhancing moist static energy extratropical cyclones after landfall because of the loss and latent heat release. Emanuel et al. (2008) performed of the energy source from the ocean, increased wind simulations using a simple tropical cyclone model coupled shear, and increased friction from the land surface, to a one-dimensional soil model to determine the role among other factors (Tuleya 1994). However, in instances of large vertical heat fluxes from recently moistened of weak baroclinicity, wetter than normal soil conditions soil on northern Australia cyclone redevelopment. may provide adequate energy to sustain TCs in a similar Results indicated that warm, wet soils may help to manner to the ocean environment (Fig. 1). maintain or energize landfalling TCs through heat trans- Wetter soils tend to absorb more radiation compared fer. In Asia, freshwater bodies and wetlands are hy- to dry soils because of the high heat capacity of water. pothesized to supply energy to typhoons removed from The net downward radiation at the soil surface can be the ocean (Chen 2012). Shen et al. (2002) found that described as hurricane decay rate is a function of surface water depth. A 2-m layer of water can adequately maintain a hurri- 5 1 3 1 Rn SHF L ET JH , (1) cane’s intensity over land (e.g., 950 hPa), a 0.5-m layer of water is sufficient to reduce landfall decay (e.g., 965 hPa), 2 where SHF is sensible heat flux (W m 2), L is latent heat and no surface water leads to rapid decay (e.g., 980 hPa) 2 of vaporization (J kg 1), ET is evapotranspiration 24 h after landfall. 22 21 (kg m s ), and JH is the soil heat flux between the Herein, a numerical water and heat flow model 2 surface and a deeper layer (W m 2) (Radcliffe and (‘‘HYDRUS-1D’’) is utilized to simulate soil charac- Sim unek 2010). Wet soils enhance evaporation and LHF, teristics and surface energy fluxes over observed regions which produce a moist, unstable boundary layer (Clark of TC intensification. Four TCMI regions are simulated and Arritt 1995; Bosilovich and Sun 1999; Lynn et al. to output a time series of energy flux values two weeks 1998). Although the boundary layer under these condi- leading up to the event. Fully coupled land surface– tions is often shallow, evaporational cooling is minimal atmospheric models have been successfully employed in (Evans et al. 2011). similar studies (Emanuel et al. 2008; Evans et al. 2011; Strengthened moisture transport, rainfall reinforce- Kellner et al. 2012). However, the accuracy of noncoupled ment, and an associated latent heat trigger are identi- versus coupled models for this particular application is fied as important factors for inland tracking storm beyond the scope of this research. It has been demonstrated intensity (Dong et al. 2010; Kishtawal et al. 2012). In that HYDRUS-1D effectively adapts meteorological data model simulations, Chang et al. (2009) found that the to simulate the surface energy balance and provides many intensity and longevity of monsoon depressions (MDs) options for soil parameters (e.g., soil type, soil texture, are sensitive to soil moisture. Arndt et al. (2009), temperature and moisture profiles, plant rooting depth, and

Unauthenticated | Downloaded 09/23/21 02:43 PM UTC DECEMBER 2013 A N D E R S E N E T A L . 2799 water table). HYDRUS-1D is valuable as a meteorological central pressure, and maximum sustained wind speed application because it solves coupled equations governing are obtained from the International Best Track Archive liquid water, water vapor, and heat transport within the soil. for Climate Stewardship (IBTrACS; http://www.ncdc. The model is tested public domain code that requires noaa.gov/oa/ibtracs/index.php). IBTrACS merges TC a minimal amount of computational resources, it has information from Regional Specialized Meteorological a straightforward user interface, and it is well documented Centers, international centers, and individuals to create (Saito et al. 2006). While recent reanalysis datasets provide the most complete global best track dataset available surface energy flux data globally, using a soil model such as (Knapp et al. 2010). HYDRUS-1D allows flexibility and control over the en- Tropical Storm (TS) Erin developed on 15 August vironmental conditions. Specifically, the model is used to 2007 in the Gulf of Mexico. It achieved tropical storm 2 simulate heat and moisture fluxes over bare soil conditions. status with 35-kt (1 kt 5 0.51 m s 1) winds and sub- Vegetation moderates the LHF of an area while energy sequently made landfall near Corpus Christi, Texas. fluxes over bare soil are more variable. Previous work has Reintensification occurred over Oklahoma, with the suggested that TCMIs are unique to Australia because of strength surpassing that of its ocean counterpart. The the ability of sandy soil to suddenly release huge amounts wind was recorded as 50 kt and pressure at 995 hPa. of energy when moistened (Emanuel et al. 2008). By sim- Studies have highlighted the anomalously wet soil in ulating the energy balance of each intensification region Oklahoma at this time and possible influence on TS Erin without vegetation, the fluxes can be directly linked to soil (Arndt et al. 2009; Evans et al. 2011; Kellner et al. 2012; texture. Andersen and Shepherd 2013). The objectives of this research are to 1) simulate the Typhoon Nat developed on 14 September 1991 in soil profiles of TC intensification regions in the 2 weeks the Philippine Sea. It reached category 3 strength on leading up to TCMI events, 2) determine the trends and the Saffir–Simpson hurricane wind scale before track- magnitudes of surface energy fluxes prior to a TC ing over Taiwan and making landfall near Chaozhou, tracking over and intensifying, and 3) compare the Guangdong (China). Nat weakened over land but simulated energy flux values with typical values over maintained a relatively constant pressure near the end the tropical ocean. Although the purpose of this study of its life cycle with a slight pressure drop (minimum coincides with Emanuel et al. (2008) (i.e., quantifica- central pressure of 1008 hPa). TC best track datasets in tion of surface energy fluxes in relation to TCs), the the west Pacific suffer from discrepancies due to varying focus is on the feasibility of moist soil acting as a sub- algorithms used by the Joint Typhoon Warning Center, stitute for the ocean environment, identifying an LHF Japan Meteorological Agency, and the China Meteo- threshold for inland intensification using multiple rological Administration (Barcikowska et al. 2012). cases, and comparing flux trends across four major Therefore, the postlandfall records for Typhoon Nat are landfall regions. The remainder of the paper is divided not as reliable as more recent cases because of uncertain into three sections. Section 2 describes the inland TC or missing data near the end of its life cycle, but it proved case studies, data, and model setup. The results are to be the only case from the west Pacific that maintained presented in section 3, followed by a discussion and a warm core over land without significant meridional conclusions in section 4. flow to complicate the analysis. Considering the overall lack of TCMI cases in eastern Asia (Andersen and Shepherd 2013), it is more than likely that extratropical 2. Cases and model setup transition or dissipation dominates in this region (per- haps because of the extreme terrain). Nevertheless, it is a. Case studies still useful to include a typhoon in this study in order to Andersen and Shepherd (2013) identified 16 inland gauge possible surface energy flux differences between tropical cyclone maintenance and intensification events the four primary landfall regions. globally between 1979 and 2008. These events displayed Tropical system 2007172N15088 developed 20 June a drop in minimum central pressure and/or an increase 2007 in the Bay of Bengal. It strengthened to a tropical in maximum sustained wind speed at least 350 km inland depression and made landfall on the east coast of India from the coast while maintaining warm-core tropical near Ethamukkala, Andhra Pradesh. After initially structures (determined by thermal wind calculations). weakening, it regained strength in central India while From this dataset, the strongest intensification case (i.e., moving northwest on 23 June. The wind was recorded at greatest pressure drop) from each major hurricane basin 20 kt and the pressure at 1007 hPa. that is minimally influenced by strong baroclinicity is Cyclone Wylva developed 14 February 2001 in the chosen for analysis (Fig. 2). The TC best track, minimum Gulf of Carpentaria. It reached tropical storm strength

Unauthenticated | Downloaded 09/23/21 02:43 PM UTC 2800 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 52

FIG. 2. TCMI case studies for HYDRUS-1D model runs. Each is the strongest intensification event (i.e., greatest pressure drop inland) originating from its respective hurricane basin between 1979 and 2008 based on work by Andersen and Shepherd (2013). The arrow denotes the point of intensification. at 41-kt winds before passing near Robinson River Each model run simulates the water and energy Airport in the Northern Territory of Australia. The pres- balance of a TCMI region in the 2 weeks antecedent to sure was steady for about 2 days after landfall and then the intensification event to produce LHF, SHF, and soil began dropping as Wylva moved west (minimum central temperature time series. Note that HYDRUS produces pressure of 988 hPa). Hurricane Satellite (HURSAT; time series for one specified location. Therefore, the http://www.ncdc.noaa.gov/oa/rsad/hursat/) visible satel- heat flow and meteorological boundary conditions are lite images for each TCMI at the time of intensification taken from assimilated reanalysis data area-averaged are shown in Fig. 3. means (a 68368 geographical area centered on the storm during intensification as indicated by the best b. Model setup track data). Research has shown that the cyclonic, HYDRUS-1D (http://www.pc-progress.com/en/Default. convergent inflow is within a 48–68 radius of the TC aspx?h1d-description) is a finite-element model for simu- center and the potential vorticity signature is well lating the one-dimensional movement of water, heat, and within a 500-km radius (Frank 1977; Evans and Hart solutes in soil. It numerically solves the Richards 2003). equation for water flow and advection dispersion equa- Three-hourly near-surface air temperature (8C), near- 2 tions for heat and solutes. The heat transport equation surface wind speed (km day 1), solar radiation flux 2 2 includes heat flow by conduction, convection, and vapor (MJ m 2 day 1), 0–10-cm soil temperature (8C), and 40– phase and takes into account soil water content (Saito 100-cm soil temperature (8C) are obtained from Global et al. 2006). Land Data Assimilation System, version 2 (GLDAS-2),

Unauthenticated | Downloaded 09/23/21 02:43 PM UTC DECEMBER 2013 A N D E R S E N E T A L . 2801

FIG. 3. HURSAT visible satellite images for (a) Typhoon Nat (1991), (b) TS Erin (2007), (c) Cyclone Wylva (2001), and (d) 2007172N15088.

Noah model experiment (http://ldas.gsfc.nasa.gov/gldas/). (Scanlon et al. 2003) and is accounted for along with Near-surface relative humidity (%) and precipitation water flow: 2 rate or influx (mm day 1) are obtained from Modern-Era     Retrospective Analysis for Research and Applications ›u(h) › ›h ›T (MERRA; https://gmao.gsfc.nasa.gov/merra/). The Nat- 5 (C 1 C ) 1 1 1 (C 1 C ) ›t ›x vh ›x LT vT ›x ural Resources Conservation Service (http://soils.usda. 2 gov) provides information on regional soil types. A S(h), (2) description of the case studies is shown in Table 1. 2 TCMI occurrence is most common in Australia and least where u is the total volumetric water content (cm3 cm 3), common in the United States (Andersen and Shepherd x is depth (cm), T is temperature (K), C is the isothermal 21 2013). hydraulic conductivity of the liquid phase (cm h ), CLT Water flow and heat transport are simulated for a is the thermal hydraulic conductivity of the liquid phase 2 21 21 100-cm depth of soil. It has been shown that vapor flow (cm K h ), Cvh is the isothermal vapor hydraulic 21 is important to consider in arid and semiarid regions conductivity (cm h ), and CvT is the thermal vapor

Unauthenticated | Downloaded 09/23/21 02:43 PM UTC 2802 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 52

TABLE 1. Description of the four case studies for HYDRUS runs. The last day of each simulation is the time of intensification. Pressure refers to minimum central pressure, and wind speed refers to maximum sustained wind speed.

Case 1 Case 2 Case 3 Case 4 TCMI case Typhoon Nat Tropical Storm Erin Cyclone Wylva 2007172N15088 Region Southeast China 258–318N, South-central United States North-central Australia Central India 158–218N, 1148–1208E 338–398N, 968–1028W 168–228S, 1288–1348E 758–818E Simulation period 20 Sep–3 Oct 1991 6–19 Aug 2007 5–18 Feb 2001 10–23 Jun 2007 Landfall 1200 UTC 1 Oct 1200 UTC 16 Aug 1200 UTC 16 Feb 0600 UTC 22 Jun Intensification Pressure dropped by 4 hPa Pressure dropped by 12 hPa Pressure dropped by Pressure dropped by in 6 h and wind speed increased 7 hPa in 18 h 3 hPa and wind by 30 kt in 12 h speed increased by 5ktin6h Soil Ultisols (red clay) Mollisols (organic and clay) Entisols (sand) Alfisols (clay) Climate Humid subtropical Humid subtropical Desert Semiarid to humid subtropical

2 21 21 hydraulic conductivity (cm K h ). Heat flow with where rs is the water vapor density at the soil surface 23 23 vapor transport is calculated as (kg m ), ra is the atmospheric vapor density (kg m ),   › ›u › › › ra is the aerodynamic resistance to water vapor flow T y T T 21 C (u) 1 L 5 l(u) 2 C q (h cm ), and rs is the soil surface resistance to water p ›t 0 ›t ›x ›x w ›x 2 vapor flow (h cm 1). Further information on numerical › › qyT qy solutions and equations is provided in the HYDRUS-1D 2 Cy 2 L , (3) ›x 0 ›x manual by Sim unek et al. (2009). where Cp is volumetric heat capacity of the porous me- 2 2 2 dium (kg cm 1 h 2 K 1), u is volumetric water con- 3 23 3. Results tent (cm cm ), uy is volumetric water vapor content 3 23 (cm cm ), L0 is the volumetric latent heat of vapor- a. HYDRUS 21 22 ization of liquid water (kg cm h ), Cw is volumetric The soil temperature and sensible and latent heat 21 22 21 heat capacity of the liquid phase (kg cm h K ), q is fluxes from the model runs are used to assess the surface 21 Darcian fluid flux density (cm h ), and qy is the vapor conditions prior to the tropical cyclone. The soil tem- 21 flux density (cm h ). The simulation time is 14 days perature profiles for bare soil show that the surface with 112 time-variable boundary records and 112 mete- is most susceptible to diurnal temperature variations orological records specifying an energy balance boundary (Fig. 4). In southeastern China, northern Australia, and condition. The van Genuchten–Mualem soil hydraulic central India, the temperatures slightly decrease with model is used with no hysteresis. The water flow pa- time, while soil temperatures in the south-central United rameters are chosen according to the soil textural class States increase. The United States and Australia have (e.g., sand, clay, or loam). The water flow boundary condi- the greatest diurnal fluctuations and highest maximum tions are atmospheric with surface runoff (upper) and zero- temperatures. The reason for this observation may be pressure head gradient or free drainage (lower), assuming related to high solar radiation during the day (sum- the water table is well below the simulated soil profile. The mertime TCMI occurrence) and radiational cooling at heat transport parameters and thermal conductivity used are night (clear skies). In Australia and India, deeper soil from the HYDRUS soil catalog (adapted from Carsel and remains warmer on average compared to the other cases Parrish 1988). The heat transport boundary conditions are owing to the tropical climate. In all regions, the daily temperature (upper) and zero gradient (lower). Solar maximum surface temperature equals or surpasses the radiation flux, solar radiation-based cloudiness factor, 26.58C SST threshold: 26.58C in China, 37.78C in the TCMI location latitude and altitude, and relative hu- United States, 39.98C in Australia, and 33.78C in India. midity are selected as meteorological parameters. For the SHF time series for each region are shown in Fig. 5. 2 soil profile, initial pressure head is set to 100 uniformly Note that the model output is in incremental time steps; (approximately field capacity). The surface LHF term is therefore, the time series plots undergo some smoothing to generate 3-hourly data. SHF magnitudes over China r 2 r 2 5 s a and India are very low, less than 30 W m 2 for the du- E 1 , (4) ra rs ration of the simulations. For much of the time, SHF

Unauthenticated | Downloaded 09/23/21 02:43 PM UTC DECEMBER 2013 A N D E R S E N E T A L . 2803

FIG. 4. HYDRUS-generated soil temperature (8C) at the surface and at 10- and 40-cm depths for (a) Typhoon Nat (China), (b) TS Erin (United States), (c) Cyclone Wylva (Australia), and (d) tropical system 2007172N15088 (India) over bare soil. is negative, implying that the ground temperature is apparent that the U.S. and Australia study regions have relatively cooler than the air. This is not surprising as both moderately high heat and moisture fluxes. Typhoon Nat occurred in early autumn and the Indian The total 2-week mean LHF indicates there is a 2 tropical system occurred during the wet monsoon season. ;70 W m 2 area-averaged threshold for potential TCMI TS Erin occurred during the hottest month of the year; occurrence consistent with previous findings (Andersen therefore, the United States features high daytime SHF and Shepherd 2013) (Fig. 7). The total 2-week maximum 2 2 (up to 220 W m 2). While Australia exhibited the high- LHF values indicate daytime fluxes reach 300 W m 2 or est soil temperatures, the SHF trails that of the United higher over the inland intensification regions. With the 2 States with daily values reaching 50–150 W m 2. exception of Nat, the TCMIs intensified during the night- The HYDRUS LHF time series and MERRA total time hours and likely encountered the residual effects of 2 precipitation flux (mm day 1)areshowninFig.6.Inthe daytime surface fluxes. Specifically, the high daytime fluxes regions where intensification is to occur, LHF is highest in increase the low-level atmospheric moisture, which is the afternoons/evenings and lowest in the mornings. As subsequently swept into the TC when it moves over. solar radiation warms the boundary layer throughout the b. Land versus ocean day, evaporation increases along with LHF. For these case studies, the LHF values are consistently higher in HYDRUS estimates of 3-hourly land surface LHF China and India. In the U.S. case, LHF is more variable antecedent to TCMI events respond to the diurnal pat- over time and spikes after precipitation events. In the tern of the boundary layer with maximum values oc- United States and Australia, the trend of LHF follows curring in the daytime. The magnitude of LHF is better most closely to the trend of precipitation because of drier understood with a comparison of values over land versus ambient conditions. The daily maximums for all regions tropical ocean. LHF within TCs are not precisely known 2 often reach 200 W m 2 or greater. The soil texture does because of the lack of accurate observations. However, not appear to have a significant effect on the LHF mag- remote sensing and algorithms have been used to cal- nitudes across regions; however, the diurnal changes are culate reasonable estimates (Guimond et al. 2011). Note sharper over Australia. Taken together with SHF, it is that in this study, LHF magnitudes over an area prior to

Unauthenticated | Downloaded 09/23/21 02:43 PM UTC 2804 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 52

22 FIG. 5. HYDRUS-generated surface sensible heat flux (W m ) over 2 weeks prior to (a) Typhoon Nat, (b) TS Erin, (c) Cyclone 2 Wylva, and (d) tropical system 2007172N15088 for bare soil (solid line) simulations. MERRA precipitation flux (mm day 1)isplotted for reference (in gray). the storm are of greater interest than those within the over land (same 68368 region used for HYDRUS runs) storm since the goal is to determine how favorable the are analyzed. ‘‘Prestorm’’ is considered to cover the 3 environment is leading up to the TC. days prior to the TC making an appearance in the re- According to Zhang and Rossow (1997), the mean an- gion. HYDRUS has a slightly lower 3-day mean LHF in 2 nual LHF peaks around 115–125 W m 2 near 158N–S. three of four intensification regions when compared to Trenberth and Fasullo (2007) estimate that the background the same land region from the MERRA dataset (Fig. 8, 2 LHF over the tropical ocean is approximately 120 W m 2. left). This is likely due to the fact that HYDRUS is High-resolution satellite-derived LHF estimates of the initialized and run with area-averaged conditions, while pre-TC ocean environment have been found to be 100– MERRA means are calculated from x–y gridded LHF 2 200 W m 2 (Liu et al. 2011). In the western North Pa- values. The two datasets match within approximately 2 cific, TC LHF maxima have been estimated between 150 45Wm 2 for all regions. When comparing HYDRUS 2 and 190 W m 2 (Gao and Chiu 2010). According to the land versus MERRA ocean fluxes, the prestorm means HYDRUS results, daytime values of LHF over the study are similar in magnitude with the exception of TCMI regions of southeastern China, south-central United States, Wylva. Prior to Wylva, the ocean area exhibited about northern Australia, and central India reach well above a 60% greater flux than the land area. Australia’s prox- these estimates. imity to the equator and the timing of the TCMI (i.e., To facilitate a meaningful comparison, MERRA hottest month of the year) might suggest that the off- LHF over ocean prestorm (68368 region adjacent to shore SSTs are relatively higher than the other three the coast where the TC is to form or strengthen) and regions and lead to enhanced LHF.

Unauthenticated | Downloaded 09/23/21 02:43 PM UTC DECEMBER 2013 A N D E R S E N E T A L . 2805

22 FIG. 6. As in Fig. 5, but for latent heat flux (W m ).

The prestorm maximum values reveal the same un- 4. Discussion and conclusions derestimation of LHF by HYDRUS in the inland re- 2 gions (up to 140 W m 2 in Australia). Despite this bias, Tropical cyclones often weaken or transition to the HYDRUS inland fluxes are remarkably higher than extratropical cyclones once reaching land. However, the ocean fluxes (.200%) in China and India. In the there are instances in which a TC has remained a warm- United States and Australia, the land and ocean fluxes core structure and intensified inland away from the are nearly equal. Near-zero nighttime values over land primary oceanic energy source (Arndt et al. 2009; Dong lower the mean, while the daytime fluxes are much et al. 2010; Evans et al. 2011; Kellner et al. 2012; higher and are quantified by the prestorm maximum Andersen and Shepherd 2013). Building upon the work values. The diurnal trend is subdued over the ocean of Andersen and Shepherd (2013), four tropical cy- where evaporation is relatively continuous. clone maintenance or intensification events were cho- The results indicate that the land surface is capable of sen as case studies to run HYDRUS-1D, a water and producing a nearly equal or greater magnitude LHF as heat flow model. The surface energy balance was sim- the ocean during the daytime hours closely preceding ulated for each intensification region leading up to the a TC. Because of the area-averaging method used with cyclone to quantify the available heat and moisture HYDRUS, the moisture fluxes produced are likely fluxes and determine if patterns are related to soil on the conservative side. Given that LHF is a primary texture. contributor to TC formation and intensification, the Time series were generated two weeks prior to each land surface may play an important role in aiding the TCMI event for a bare soil configuration over a region maintenance and intensification of TCs when soils are centered on the TCMI location at the time of inland abundantly moist and LHF is high. intensification. Soil temperature and LHF were generally

Unauthenticated | Downloaded 09/23/21 02:43 PM UTC 2806 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 52

a function of latitude, total landfalls, terrain character- istics, and atmospheric conditions. For all HYDRUS runs, the total 2-week maximum LHF values are comparable to or exceed typical ocean LHF magnitudes. The total 2-week mean LHF indi- 2 cates there is a ;70 W m 2 area-averaged threshold for TCMI occurrence. MERRA reanalysis 3-day (pre- storm) LHFs were utilized for a land versus ocean comparison. The prestorm mean LHFs prior to the TC reveal that the oceanic LHF was higher than the land LHF in three of four regions. Considering that LHF is diminished at nighttime over land, the daytime flux values are remarkably high. The 3-day maximum LHF values inland are well above those over the ocean in the China and India intensification regions, while LHF magnitudes ocean versus land were nearly equal over 22 FIG. 7. The 2-week mean and maximum LHF (W m ) antecedent the U.S. and Australia regions. This suggests that LHF to the TCMI for HYDRUS bare soil simulations. maxima over land have a similar magnitude to those found over the ocean prior to the appearance of a TCMI high in each region, with China and India exhibiting the and provide plausibility of the brown ocean effect. highest LHF and lowest SHF. The soil texture (i.e., While HYDRUS provides a unique avenue to assessing hydraulic conductivity) does not appear to have a sig- the surface energy balance by including liquid water nificant effect on the LHF magnitudes across regions. and vapor transport in soils, the 1D design is a limitation While TCMIs are most common over Australia, the to exploring the spatial distribution of LHF. It may be LHF over the study period is not noticeably greater concluded, however, that the daytime flux magnitudes over sandy soil than over clay (China and India) or output by HYDRUS are underestimations due to the loam (United States). However, the diurnal changes area-averaging method. are sharper over Australia and the high LHF is accom- Further analysis is needed to determine the relative panied by high SHF. This agrees with the notion of importance of short-term soil moisture anomalies to ‘‘sudden energy release’’ when the sandy soil is wetted. other contributing factors of TC intensification. Although For saturated soils, hydraulic conductivity is signifi- Evans et al. (2011) examined the seasonal soil moisture cantly greater for coarse-textured soils than clays or signal influence on TS Erin, precipitation recycling and loams (Radcliffe and Sim unek 2010). The U.S. region soil moisture memory warrant investigation in other exhibits a similar trend, with relatively high LHF and landfall regions where the land surface–atmospheric SHF, yet TCMIs are much less common (Andersen and feedback may be stronger or weaker. Additionally, it Shepherd 2013). TCMI frequency in Australia is likely was suggested that anomalously wet soils encouraged

FIG. 8. Three-day (left) mean and (right) maximum LHF over ocean (MERRA) and land (MERRA and HYDRUS) along TCMI tracks.

Unauthenticated | Downloaded 09/23/21 02:43 PM UTC DECEMBER 2013 A N D E R S E N E T A L . 2807 moisture transport from the Gulf (Evans et al. 2011). Dong, M., L. Chen, Y. Li, and C. Lu, 2010: Rainfall reinforcement Analysis of synoptic-scale features for additional TCMIs associated with landfalling tropical cyclones. J. Atmos. Sci., 67, could determine if this is a common trait. A thorough 3541–3558. Emanuel, K., J. Callaghan, and P. Otto, 2008: A hypothesis for investigation of soil moisture patterns, timing of rainfall the redevelopment of warm-core cyclones over northern events, and structure and speed of TCs in relation to Australia. Mon. Wea. Rev., 136, 3863–3872. intensification potential would be beneficial. Evans, C., R. S. Schumacher, and T. J. Galarneau, 2011: Sensitivity Future studies may extend this research by using a in the overland reintensification of Tropical Cyclone Erin larger sample size to compare TCMIs within and across (2007) to near-surface soil moisture characteristics. Mon. Wea. Rev., 139, 3848–3870. regions or add null cases to better understand differ- Evans, J. L., and R. E. Hart, 2003: Objective indicators of the life ences between intensifying and weakening TCs. This cycle evolution of extratropical transition for Atlantic tropical study and previous works are primarily feasibility stud- cyclones. Mon. Wea. Rev., 131, 909–925. ies for the soil moisture–cyclone feedback hypothesis. Frank, W. M., 1977: The structure and energetics of the tropical High spatial and temporal resolution modeling would cyclone. Mon. Wea. Rev., 105, 1119–1135. Gao, S., and L. S. Chiu, 2010: Surface latent heat flux and rainfall provide more refined evidence of the heat and moisture associated with rapidly intensifying tropical cyclones over the transport to the TC. Additionally, point observations western North Pacific. Int. J. Remote Sens., 31, 4699–4710, from meteorological stations are ideal input where doi:10.1080/01431161.2010.485149. available and could lead to more precise results for Gray, W. M., 1968: Global view of the origin of tropical distur- multiple TCMI events. A hurricane simulation model bances and storms. Mon. Wea. Rev., 96, 669–700. Guimond, S. R., M. A. Bourassa, and P. D. Reasor, 2011: A coupled with a sophisticated soil model can simulate a latent heat retrieval and its effects on the intensity and TCMI event and lead to better understanding of the structure change of Hurricane Guillermo (1997). Part I: brown ocean effect. The algorithm and observations. J. Atmos. Sci., 68, 1549– 1567. Kellner, O., D. Niyogi, M. Lei, and A. Kumar, 2012: The role Acknowledgments. This research is funded by a of anomalous soil moisture on the inland reintensification NASA Earth and Space Science Fellowship under of Tropical Storm Erin (2007). Nat. Hazards, 139, 1573– 1600. Grant NNX11AL86H. Kishtawal, C. M., D. Niyogi, A. Kumar, M. L. Bozeman, and O. Kellner, 2012: Sensitivity of inland decay of North Atlantic tropical cyclones to soil parameters. Nat. Hazards, 63, 1527– 1542, doi:10.1007/s11069-011-0015-2. REFERENCES Kleinschmidt, E., Jr., 1951: Gundlagen einer Theorie des tropi- Andersen, T. K., and J. M. Shepherd, 2013: A spatio-temporal schen Zyklonen. Arch. Meteor., Geophys. Bioklimatol., Ser. A, analysis of tropical cyclone maintenance or intensification 4, 53–72. inland. Int. J. Climatol., doi:10.1002/joc.3693, in press. Knapp, K. R., M. C. Kruk, D. H. Levinson, H. J. Diamond, and Arndt, D. S., J. B. Basara, R. A. McPherson, B. G. Illston, G. D. C. J. Neumann, 2010: The International Best Track Archive McManus, and D. B. Demko, 2009: Observations of the for Climate Stewardship (IBTrACS): Unifying tropical overland reintensification of Tropical Storm Erin (2007). Bull. cyclone best track data. Bull. Amer. Meteor. Soc., 91, 363– Amer. Meteor. Soc., 90, 1079–1093. 376. Barcikowska, M., F. Feser, and H. von Storch, 2012: Usability of Liu, C.-X., Q.-L. Wan, F. Liao, and Z.-K. Zhao, 2012: Surface best track data in climate statistics in the western North Pa- observations in the tropical cyclone environment over the cific. Mon. Wea. Rev., 140, 2818–2830. South China Sea. J. Trop. Meteor., 18, 263–274. Bosilovich, M. G., and W.-Y. Sun, 1999: Numerical simulation of Liu, J., J. A. Curry, C. A. Clayson, and M. A. Bourassa, 2011: High- the 1993 Midwestern flood: Land atmosphere interactions. resolution satellite surface latent heat fluxes in North Atlantic J. Climate, 12, 1490–1505. hurricanes. Mon. Wea. Rev., 139, 2735–2747. Carsel,R.F.,andR.S.Parrish,1988:Developingjointprob- Lynn, B. H., W.-K. Tao, and P. J. Wetzel, 1998: A study of landscape– ability distributions of soil water retention characteristics. generated deep moist convection. Mon. Wea. Rev., 126, 928– Water Resour. Res., 24, 755–769, doi:10.1029/WR024i005p00755. 942. Chang, H., D. Niyogi, A. Kumar, C. M. Kishtawal, J. Dudhia, Malkus, J. S., and H. Riehl, 1960: On the dynamics and energy F. Chen, U. C. Mohanty, and M. Shepherd, 2009: Possible transformations in steady-state hurricanes. Tellus, 12, 1–20. relation between land surface feedback and the post-landfall Miller, B. I., 1958: On the maximum intensity of hurricanes. structure of monsoon depressions. Geophys. Res. Lett., 36, J. Meteor., 15, 184–195.  L15826, doi:10.1029/2009GL037781. Radcliffe, D. E., and J. Simunek, 2010: Soil Physics with HYDRUS: Chen, L.-S., 2012: Research progress on the structure and intensity Modeling and Applications. CRC Press, 373 pp.  change for the landfalling tropical cyclones. J. Trop. Meteor., Saito, H., J. Simunek, and B. P. Mohanty, 2006: Numerical 18, 113–118. analysis of coupled water, vapor, and heat transport in Clark, C. A., and R. W. Arritt, 1995: Numerical simulations of the vadose zone. VadoseZoneJ.,5, 784–800, doi:10.2136/ the effect of soil moisture and vegetation cover on the vzj2006.0007.  development of deep convection. J. Appl. Meteor., 34, 2029– Scanlon, B., K. Keese, R. C. Reedy, J. Simunek, and B. Andraski, 2045. 2003: Variations in flow and transport in thick desert vadose

Unauthenticated | Downloaded 09/23/21 02:43 PM UTC 2808 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 52

zones in response to paleoclimatic forcing (0–90 kyr): Moni- HYDRUS Software Series 3, 330 pp. [Available online toring, modeling, and uncertainties. Water Resour. Res., 39, at http://www.pc-progress.com/Downloads/Pgm_hydrus1D/ 1179, doi:10.1029/2002WR001604. HYDRUS1D-4.08.pdf.] Shen, W., I. Ginis, and R. E. Tuleya, 2002: A numerical investigation Trenberth, K. E., and J. Fasullo, 2007: Water and energy budgets of land surface water on landfalling hurricanes. J. Atmos. Sci., and hurricanes and implications for climate change. J. Geo- 59, 789–802. phys. Res., 112, D23107, doi:10.1029/2006JD008304. Sim unek, J., M. Sejna, H. Saito, M. Sakai, and M. T. van Genuchten, Tuleya, R. E., 1994: Tropical storm development and decay: Sensitivity 2009: The HYDRUS-1D software package for simulating to surface boundary conditions. Mon. Wea. Rev., 122, 291–304. the one-dimensional movement of water, heat, and multiple Zhang, Y.-C., and W. B. Rossow, 1997: Estimating meridional solutes in variably-saturated media, version 4.08. University energy transports by the atmospheric and oceanic general of California, Riverside, Dept. of Environmental Sciences circulations using boundary fluxes. J. Climate, 10, 2358–2373.

Unauthenticated | Downloaded 09/23/21 02:43 PM UTC