Dynamical Downscaling Projections of Late 21st Century U.S. Landfalling Hurricane Activity

U.S. landfalling tropical cyclone (TC) activity was projected for late 21 st century conditions using a two-step dynamical downscaling framework. A regional atmospheric model, run for 27 seasons, generated tropical storm cases. Each storm case was re-simulated (up to 15 days) using the higher resolution GFDL hurricane model. Thirteen CMIP3 or CMIP5 modeled climate change projections were explored as scenarios. Robustness of projections was assessed using statistical signicance tests and comparing the sign of changes derived from different models. The proportion of TCs (tropical storms and hurricanes) making U.S. landfall increases for the warming scenarios (by order 50% or more). For category 1-3 hurricane frequency, a robust decrease is projected (basin-wide), but robust changes are not projected for U.S. landfalling cases. A relatively robust increase in U.S. landfalling category 4-5 hurricane frequency is projected, averaging about +400% across the models; 10 of 13 models/ensembles project an increase (statistically signicant in three individual models), while three models projected no change. The most robust projections overall for U.S. landfalling TC activity are for increased near-storm rainfall rates: these increases average +18% (all tropical storms and hurricanes), +26% (all hurricanes), and +37% (major hurricanes). Landfalling hurricane wind speed intensities show no robust signal, in contrast to a ~5% increase in basin-averaged TC intensity; basin-wide Power Dissipation Index (PDI) is projected to decrease, partly due to decreased duration. TC translation speed increases a few percent in most simulations. A caveat is the framework’s low correlation of modeled U.S. TC landfalls vs. observed interannual variations (1980-2016). but these time series do not show any signicant increases 1900. The lack of a signicant change in long TC landfalling is in contrast to the case for global mean temperature, where a clear anthropogenic warming signal has been identied (IPCC AR5). This indicates that TC landfall frequency in the above regions is clearly not a metric with a strongly detectable anthropogenic signal even on a century time-scale. A previously published Atlantic-basin-focused dynamical downscaling study (Knutson et al. 2013, hereafter K13) addressed the potential impact of future global warming on TCs, but the previous study focused on lifetime maximum intensities of TCs, by using ve-day high-resolution simulations of storms near their times of maximum intensity. Importantly, K13 did not focus on the U.S. landfalling stages, which often were outside of the 5-day window simulated with the high resolution model. The purpose of the present study is to revisit the two-step dynamical downscaling study of K13, with a specic purpose of focusing on the U.S. landfalling stages of the simulated hurricanes (for the Contiguous U.S., i.e., excluding Hawaii, Puerto Rico, Guam, etc.). Through this study, we aim to provide more societally relevant information about the damage potential impacts of the storms (in terms of intensity, frequency, rainfall) under various climate change scenarios. Previous studies on possible future changes in U.S. landfalling TCs have reported model projections of: reduced probability of TC landfall over the southeastern U.S. and increased probability over the northeastern U.S. (Murakami and Wang reduced TC occurrence over the southern Gulf of Mexico and Caribbean (Colbert et al. 2013); increased average TC rain rates over U.S. land regions (Wright al. 2015); and increased likelihood of faster-moving landfalling TCs in the Texas region (Hassanzadeh 2020). The latter study result is qualitatively in contrast to an observed nding for historical TC of a signicant reduction in propagation speed over the U.S. land regions observed result which was not reproduced in a historical forcing model Our study explores 21st century climate change projections for U.S. landfalling hurricane activity using a two-step dynamical downscaling framework, together with tropical climate change projections from multiple CMIP3 and CMIP5 climate models– the same models as used in K13. In the earlier study, the highest resolution simulations with the GFDL hurricane model (i.e., the second step of our two-step downscaling procedure) were limited to ve days in length, as we were closely following the procedures of the operational GFDL hurricane model from that time period, which had some operational limitations. Given this ve-day limitation of simulation length in K13, the focus of the high-resolution simulations was on the time period in which each hurricane was approaching its maximum intensity. Specically, in K13 each ve-day downscaling case was started from initial conditions, obtained from the 18 km grid Zetac regional atmospheric model (Knutson et al. 2007; hereafter K07), beginning two days prior to the storm’s time of maximum intensity as simulated in the Zetac model. The Zetac model simulated hurricanes only up to about 50 m s-1 intensity in terms of surface wind speed, which was one reason why the second downscaling step with the higher-resolution (9 km spacing for the inner grid) was necessary. In contrast, in the present study, where we wish to focus on the landfalling stages as well as simulate realistic storm intensity and structure throughout each storm’s life cycle, we start each high-resolution hurricane model simulation (storm case) using initial conditions from the regional Zetac model at the time when tropical storm intensity of 17.5 m s-1 is rst reached in the Zetac model, integrate the hurricane model forward for 15 days. category 4–5 TCs in these projections). Major caveats to our study include the limited skill shown by the downscaling framework in simulating the historical year-to-year variability of U.S. landfalling TC activity using SSTs and the NCEP/NCAR Reanalysis as large-scale climate forcings. There is also uncertainty in the climate change signal in the large-scale environmental parameters, which is partly reected in the spread of results across the different model-derived scenarios. The spread shown in our results cannot be assumed to represent the true condence intervals on the results in this study. Despite these limitations, it is important to test our models with such scenarios and continue to compare modeled scenarios with the growing observational database to work toward a better understanding of the changes in landfalling hurricane risk facing our society in the coming century.


Introduction
U.S. landfalling TCs (hurricanes and tropical storms) can cause major damage to coastal and inland infrastructure, and it is of great interest to better understand how landfallling TC activity may change under future anthropogenic climate change, with a particular focus on landfalling hurricanes. Relatively long records (since at least about 1900) are available for tropical storm, hurricane, and major hurricane landfalls (e.g., Knutson 2008, 2011;Klotzbach et al. 2020;Vecchi et al. 2021), but these time series do not show any signi cant increases since 1900. The lack of a signi cant change in long TC landfalling records is in contrast to the case for global mean temperature, where a clear anthropogenic warming signal has been identi ed (IPCC AR5). This indicates that TC landfall frequency in the above regions is clearly not a metric with a strongly detectable anthropogenic signal even on a century time-scale. A previously published Atlantic-basin-focused dynamical downscaling study (Knutson et al. 2013, hereafter K13) addressed the potential impact of future global warming on TCs, but the previous study focused on lifetime maximum intensities of TCs, by using ve-day high-resolution simulations of storms near their times of maximum intensity. Importantly, K13 did not focus on the U.S. landfalling stages, which often were outside of the 5-day window simulated with the high resolution model. The purpose of the present study is to revisit the two-step dynamical downscaling study of K13, with a speci c purpose of focusing on the U.S. landfalling stages of the simulated hurricanes (for the Contiguous U.S., i.e., excluding Hawaii, Puerto Rico, Guam, etc.). Through this study, we aim to provide more societally relevant information about the damage potential impacts of the storms (in terms of intensity, frequency, rainfall) under various climate change scenarios. Previous studies on possible future changes in U.S. landfalling TCs have reported model projections of: reduced probability of TC landfall over the southeastern U.S. and increased probability over the northeastern U.S. (Murakami and Wang (2010); reduced TC occurrence over the southern Gulf of Mexico and Caribbean (Colbert et al. 2013); increased average TC rain rates over U.S. land regions (Wright et al. 2015); and increased likelihood of fastermoving landfalling TCs in the Texas region (Hassanzadeh et al., 2020). The latter study result is qualitatively in contrast to an observed nding for historical TC of a signi cant reduction in propagation speed over the U.S. land regions since 1900 (Kossin 2019) -an observed result which was not reproduced in a historical forcing model simulation (Zhang et al. 2020). Levin and Murakami (2019) found that historical increases in anthropogenic climate forcing led (qualitatively) to increased frequency of U.S. major hurricane landfall in their model, although a signi cant increase in U.S. major hurricane frequency is not seen in observations since 1900 (Klotzbach et al. 2020) or since the late 19th century (Vecchi et al. 2021).
Our study explores 21st century climate change projections for U.S. landfalling hurricane activity using a two-step dynamical downscaling framework, together with tropical climate change projections from multiple CMIP3 and CMIP5 climate models-the same models as used in K13. In the earlier study, the highest resolution simulations with the GFDL hurricane model (i.e., the second step of our two-step downscaling procedure) were limited to ve days in length, as we were closely following the procedures of the operational GFDL hurricane model from that time period, which had some operational limitations. Given this ve-day limitation of simulation length in K13, the focus of the high-resolution simulations was on the time period in which each hurricane was approaching its maximum intensity. Speci cally, in K13 each ve-day downscaling case was started from initial conditions, obtained from the 18 km grid Zetac regional atmospheric model (Knutson et al. 2007; hereafter K07), beginning two days prior to the storm's time of maximum intensity as simulated in the Zetac model. The Zetac model simulated hurricanes only up to about 50 m s-1 intensity in terms of surface wind speed, which was one reason why the second downscaling step with the higher-resolution (9 km spacing for the inner grid) was necessary. In contrast, in the present study, where we wish to focus on the landfalling stages as well as simulate realistic storm intensity and structure throughout each storm's life cycle, we start each high-resolution hurricane model simulation (storm case) using initial conditions from the regional Zetac model at the time when tropical storm intensity of 17.5 m s-1 is rst reached in the Zetac model, and then integrate the hurricane model forward for 15 days. More details of methodology are provided in Sect. 2, results of the experiments are presented in Sect. 3, and our summary and conclusions in Sect. 4.

Methodology
The methodology for our study mostly follows that in K13, K07, and Bender et al. (2010), and is described in detail in those studies. Here the methodology is presented only in abbreviated form here, where we focus mainly on aspects of methodology that differ from K13 and K07. The reader is referred to these previous studies for further details.

Present-day hurricane simulations
The simulation of Atlantic hurricanes proceeds in in several stages. Before we perform climate change simulations, we rst want to assess whether our complete modeling system is able to adequately simulate Atlantic hurricane activity for a given set of environmental conditions. To explore this, we rst test our system on present-day large-scale climate as de ned by the time-evolving NCEP/NCAR Reanalysis I (Kalnay et al. 1996) of the atmosphere and observed sea surface temperature (SST) evolution for 1980-2016. For each year (1980-2016), we simulate the three-month period August-October with the regional 18km grid Zetac regional model, using the reanalysis to provide lateral boundary conditions, the atmospheric initial conditions, and the time-evolving target elds for interior spectral nudging of the very large scale (zonal and meridional wavenumber 0-2 of the regional domain) atmospheric environment, with a nudging timescale of 12 hours. The simulations were limited to the peak three months of the Atlantic TC season (Aug. -Oct.) to save on computation requirements, which may introduce some uncertainty to our results, particularly if aspects of seasonality, such as the length of the Atlantic TC season, were to change with climate warming. For example, Dwyer et al. (2015) nd that in model projections, models that project fewer TCs with climate warming also simulate shorter seasons, and vice versa for models projecting more TCs. This potential limitation of our Aug.-Oct. season approach should be kept in mind in interpreting our results. The model was run over speci ed time-evolving SSTs. The Zetac regional model is a nonhydrostatic model run here without convective parameterization (see K07 for details).
Tropical storm cases are identi ed in these three-month simulations using the automated procedure described in K07, including a requirement for warm-core structure, surface wind speeds for the storm feature of at least 17.5 m s-1, and total duration of at least 48 hours (not necessarily consecutive hours) at an intensity of at least 17.5 m s-1. Landfalling TCs for the Contiguous U.S. (CONUS) were de ned by the intersection of the surface center of the TC (de ned by the minimum in surface pressure) with the coastline (https://www.nhc.noaa.gov/aboutgloss.shtml), where the land region was part of the 48 contiguous states (excluding Alaska and Hawaii). Multiple CONUS landfalls by a single storm were counted as separate landfalls in our statistics.
Each individual tropical storm case from the Zetac regional model is then re-run as an individual 15-day case study using the GFDL Hurricane Model. The  (Bender et al. 2007) and for the present-day simulations uses a realistic present-day (U.S. Navy GDEM, or Generalized Digital Environmental Model) climatology of ocean subsurface temperature and salinity so that the hurricane model is able to accurately simulate the ocean response to the strong hurricane forcing and generate a realistic "cold wake" as it passes over the ocean (Bender et al. 2007). Following the procedure used operationally (2006-2010) for storm initialization, the ocean model state for each storm case is initialized by running the ocean model for two days using the GDEM climatology values for ocean temperature and salinity with SSTs prescribed. During this rst step, a sharpening technique is employed in order to assimilate a reasonable climatologically based Loop Current and Gulf Stream ocean structure (Yablonsky et al. 2015). Next, starting three days prior to the storm case start date, atmospheric forcing is imposed, based on the model's storm wind eld. In this second step, the SSTs as well as ocean temperatures and salinity are allowed to evolve. This procedure is used to initialize a cold wake in the SST and ocean temperature eld due to the passage of the storm. A limitation of the version of the operational GFDL Hurricane Modeling System used in this study was that the dynamical ocean domain did not cover the full North Atlantic basin but rather the coupled model used two separate ocean model domains-one covering the central to eastern Atlantic and the other the central to western Atlantic (see Fig. 5 of Bender et al. 2007). Storms traversing the central Atlantic could run off one of these grids and lose their ocean coupling. However since the two ocean grids overlapped when this occurred, we restarted the storm on the second ocean grid before the coupling was lost and continued the simulation with full ocean coupling until landfall. Since this happened only for a small subset of runs, and the transition occurred very far from U.S. landfall, and was treated the same for both present-day and warm climate experiments, we believe that this limitation of our model framework was very unlikely to affect our overall conclusions.
In addition, during the 15-day integrations, the hurricane model atmosphere tended to drift toward the model's climatological state, which differs from that of the host (Zetac) regional model and NCEP/NCAR Reanalysis. However, the average drift that occurs is found to be similar for present-day climate and future warm climate scenarios, so that any systematic effects of the drift on the storm characteristics should be similar in the present-day and warm climate cases. Figure 1 compares the tracks and intensities of simulated U.S. landfalling TCs from the Zetac regional model and the GFDL Hurricane Model for the years 1980-2013. The tracks from the higher resolution model resemble those of from the Zetac regional model, although the intensities extend to higher categories (up to category 5) in the hurricane model compared to almost no track segments above category 1 in the Zetac regional model.
One test of the simulation skill of our two-step downscaling model framework is a comparison of the year-to-year variability of August-October observed mean storm counts with that from the modeling framework for different categories of tropical storms and hurricanes (Fig. 2). For example, if the model framework, which was only provided large-scale information about the year-to-year variability of the Atlantic basin SSTs and atmospheric circulation, along with very large-scale interior spectral nudging, is still able to generate useful information about tropical storm, hurricane, and intense hurricane numbers and their yearto-year variation, this increases our con dence that the framework can translate some forms of atmospheric and SST variability into useful information about hurricane activity.
The time series in Fig. 2 show that the model is indeed useful for simulating Atlantic basin hurricane activity given speci ed large-scale atmospheric, oceanic, and SST conditions. Speci cally, with only SSTs, ocean temperature climatology, and time-varying NCEP/NCAR Reanalysis boundary forcing and very largescale interior domain nudging, the system has the following correlation with observed 37 year time series of Atlantic basin Aug-October storm counts: 1) all TCs (tropical storms and hurricanes): r = 0.77 (explained variance: 59%); 2) Category 1-5 hurricanes: r = 0.68 (explained variance: 46%); 3) major (Category 3-5) hurricanes: r = 0.56 (explained variance: 31%); 4) very intense (Category 4-5) hurricanes: r = 0.31 (explained variance: 10%). Assuming independent years, correlations above 0.329 are signi cant at the 0.05 level, so for all cases except Category 4-5 hurricanes the results show signi cant correlation. For the Category 4-5 hurricanes, the results are nearly statistically signi cant.
We note that there are rising trends evident in many of the time series in Fig. 2. The modeled trends are similar to the observed with the notable exception of basin-wide hurricane frequency (Fig. 2c) where the model has a more rapid rising trend than observed. The cause of these rising trends in observations remains a topic of research, as the time period (1980-2013) is relatively short for detection of greenhouse gas warming in uence, and since both internal variability and changes in aerosol forcing are possible contributors to such TC-related trends (e.g., Goldenberg et al. 2001 Our simulations indicate only that changes in the large-scale environment (including SSTs) help to explain the rising trends in observed TC and hurricane frequency but do not elucidate the causes of the environmental changes. However, we expect that the observed Atlantic hurricane frequency trends, and tropical Atlantic SST changes, since 1980 have multiple causes and several studies suggest the TC frequency increases are likely not primarily a response to increasing greenhouse gases alone. Thus, the over-prediction of the observed trend in hurricane frequency  evident in our model framework does not invalidate its potential use for greenhouse gas-driven future warming scenarios.
The above results provide an important caveat to our study. While the two-step model framework is relatively skillful at reproducing the year-to-year variation of basin-wide tropical storm and hurricane counts, this skill does not carry through to U.S. landfalling counts. Thus, while the basin-wide results provide modelbased evidence that the year-to-year variability in the basin-wide numbers is not random "weather noise" but rather is controlled to a large extent by large-scale environmental conditions, the U.S. landfalling count variations seem much more di cult to capture using our modeling framework. While we are not aware of many other modeling systems that can successfully simulate U.S. landfalling TC frequency, one exception is the HiFLOR model (Murakami et al. 2016), which does show some skill in predicting seasonal U.S. landfalling TC frequency over the period 1980-2015, suggesting that there are large-scale controls on this metric that are not being well-captured in our two-step model framework. We have chosen to use our modeling framework to explore future U.S. landfalling behavior under global warming to take advantage of its high-resolution for simulating hurricane structure, while keeping in mind the model's limitations for reproducing observed interannual variations of TC frequency from large-scale interannual variations of environmental conditions.

Climate Change Downscaling Experiments
Following on the 37 Atlantic hurricane seasons of present-day two-step downscaling simulations (1980-2016), we perform similar experiments, but applying a set of climate change conditions, for 27 of the 37 present-day seasons .
We rst create a series of climate change "delta" elds, which we can add to the NCEP/NCAR Reanalysis, to create a series of warm-climate perturbation experiments that use realistic conditions (i.e., the reanalysis) as the baseline case. Speci cally, we use we use changes in SST, sea level pressure (SLP), air temperature, relative humidity, and wind velocity to modify the NCEP/NCAR reanalysis elds that are used as boundary forcing and as the nudging target for the interior spectral nudging procedure with the Zetac regional model. As described in K13, for CMIP3 models, these perturbations included an 18-model temperature difference between present-day and the warm-climate condition was 1.69oC for CMIP3 vs. 1.70oC for the CMIP5 late-21st century case. In sensitivity experiments, we found that hurricane model intensity changes in the warm climate scenarios were relatively insensitive to the small increase in the ocean subsurface vertical temperature gradients associated with the SST warming scenarios (see also Tuleya et al. 2016). Therefore, following Knutson et al.
(2013) for the ocean subsurface temperature pro les in the warm climate runs, we used the 18-model average three-dimensional ocean structure change from the CMIP3 models to represent the change in ocean temperature strati cation in the warmer climate for all of the hurricane model climate change experiments.

Results Of Climate Change Downscaling Experiments
In this section, we examine the results of our climate change downscaling experiments. In the discussion below, we refer to the 13 sets of experiments as 13 "models", even though these are really the downscaling results based on input climate change signals from a particular model, or in some cases based on the ensemble mean climate change signal from a collection of models (18 CMIP3 models, and 18 CMIP5 models for the early 21st century or late 21st century scenarios).
There are many TC metrics that can be examined in our 13 different sets of experiments. To focus on results from our experiments that are relatively more robust, we present the results in a summary form showing both the level of agreement across the models for projected changes for various metrics, along with statistical signi cance tests for individual models. The statistical tests assume the 27 separate seasons (years) are independent samples. Using this summary approach, Table 1 Table 1 indicates that a very robust signal emerges for increasing TC rainfall rates, for all (basin-wide) TCs (in Table 1, see rows labeled: rain_all_TC (cat 0-5), rain_hur_(cat 1-5), rain_mhur (cat 3-5), rain_hur45) and to a lesser degree (in terms of statistical signi cance) for U.S. landfalling TCs (in Table 1: lf_rain_all_TC (0-5), lf_rain_hur (1-5), lf_rain_mhur (cat 3-5)). The rainfall rates are examined for various categories of TC. Robust increases in rainfall rates are indicated across all TC categories. For basin-wide TCs, the increase is averages about 20% regardless of TC category. Interestingly, for the U.S. landfalling TCs, the percentage increase in TC rain rate with climate warming increases monotonically with storm category, from + 19% averaged over all categories of TC (in Table 1: rain_all_TC (cat 0-5) to almost + 200% for TCs that are Category 4 or 5 at landfall (in Table 1: lf_rain_hur45). The reason for the increased sensitivity of this metric for very intense landfalling TCs is uncertain. There is a strong observed relationship between rainfall rate and intensity at landfall . As will be discussed further below, there is some indication in Table 1 for increased TC intensities with warming (basin-wide) although TC intensity at landfall shows less robust change in the simulations. The increased TC intensity over ocean regions may lead to ampli ed increased TC rain rates over oceans that may exceed the increase expected from the Clausius-Clapeyron relation (increasing water vapor with warming) alone (e.g., Liu et al. 2019). However it remains unclear why the rain rate percentage increase at landfall is larger for higher category storms (e.g., + 193% for lf_rain_hur45 vs. 18% for lf_rain_all_TC (0-5))-although these are relatively fewer in number than lower-category landfalling storms and thus the statistics may be affected by limited Another relatively robust response seen in the simulations is the prevalence of negative changes (decreases) in the upper 10 rows of Table 1, which concern basin-wide TC frequency changes for various categories of TC. The decreases are particularly robust for basin-wide: category 1 hurricanes (-42%), category 2 hurricanes (-45%), category 3 hurricanes (-43%), all tropical storms and hurricanes combined (category 0-5; -28%), all hurricanes combined (category 1-5; -34%), and major hurricanes (category 3-5; -25%). Basin-wide PDI shows a similarly robust projected decrease of -28%.
The highly signi cant and robust decreases across models seen for basin-wide frequency of TCs does not carry through very strongly to U.S. landfalling TCs.
In particular, the decreases are less robust and signi cant for the frequency of weaker U.S. landfalling storm classes. On the other hand, a moderately signi cant and robust increasing signal emerges for U.S. landfalling category 4-5 storms (in Table 1: lf_hur45, average change + 390% with at least nominal increases for 10 of 13 models, no change for the remaining three models, and statistically signi cant increases for three of the 13 models). This is noteworthy because these storms have historically caused enormous damage upon landfall. In particular, Pielke et al. (2008) found that U.S. landfalling category 4-5 TCs have historically accounted for almost 50% of normalized TC damage in the U.S., despite representing only 6% of historical TC occurrences. As a sensitivity test, we have assessed U.S. landfalling category 4-5 frequency based on surface pressure, rather the surface wind speed criteria (not shown). The statistical test results are similar though slightly less robust than for the wind speed-based results shown in Table 1.
Maximum lifetime TC intensity (in Table 1, max_wind_all_TC; max_wind_hur) is a metric which models have consistently projected to increase with climate warming (Knutson et al. 2020). The results from our experiments are less compelling for simulated U.S. landfalling TCs. For U.S. landfalling hurricanes (in Table 1: lf_maxwnd_hur), intensity at landfall increases at least nominally for about half the models (seven of 13), and in only one of 13 is the change statistically signi cant. For intensity of U.S. landfalling tropical storms and hurricanes combined (lf_maxwnd_all_TC) no increase of intensity is evident. For basin-wide hurricane intensity (i.e., in Table 1: max_wind_hur, sampling each hurricane's lifetime maximum intensity), the average intensity of hurricanes increases in 12 of 13 models, of which six of the models yield statistically signi cant increases. The one exception is the downscaling of the CMIP3 HadGEM1 model, which simulates a signi cant decrease. As discussed in K13, this particular climate model exhibits a greatly enhanced warming in the upper tropical troposphere (Atlantic main development region, 300 hPa) compared to its surface warming-a factor of 3.5 greater, which much more than the average upper tropospheric warming ampli cation factor of 1.9 to 2.7 in the other CMIP3 models. Ampli ed upper tropospheric warming compared to the surface is a known detrimental factor for TC intensi cation in the GFDL Hurricane model (Tuleya et al. 2016). The average TC intensity change across the 13 models is about 5%.
This result is relatively consistent with other high-resolution modeling studies (Knutson et al. 2020). Interestingly, this increasing signal is absent if one includes weaker (tropical storm-strength TCs) in the sample, but we consider the hurricane intensity result as the more relevant one for potential climate impacts. In summary, despite the increase in basin-wide hurricane intensity, the average intensity of landfalling hurricanes shows little signi cant change, nominally averaging about + 2%. Table 1 shows that the projected average TC duration has a clear tendency to decrease by about 13% (decreasing in 12 of 13 models, with 8 of 13 models projecting a signi cant decrease). Projected changes in TC translation speed (basin-wide average, labeled "trans_speed" in Table 1) show a weak tendency for an increase across the models, with 10 models projecting an increase (two are signi cant) and three models projecting a decrease. The slight speed-up of translation speed may be one factor for the tendency for a decrease in duration, as the lifecycle of the storm over a given track would be shortened by the faster propagation speed. However, we have not diagnosed the reasons for the decreased duration in detail.
The proportion of TCs making U.S. landfall tends to increase in the projections according to most, but not all models. For example, Table 1 (U.S. Landfall proportions section, last nine rows of table) indicates that the proportion of all tropical storms and hurricanes that make landfall (labeled "all_TC (cat 0-5)") increases on average by about 50% above its baseline (control run) value of 0.17. Some increase was simulated in all 13 models. For all hurricanes, the proportion making U.S. landfall (labeled "hur (cat 1-5)") is projected to increase by all 13 models, with an average increase of 64% above the control proportion value of 0.12. The proportion of U.S. landfalling TCs is projected to increase for major hurricanes (+ 76% over control value of 0.077, increasing for 12 or 13 models, row labeled "mhur (cat 3-5)") and category 4-5 hurricanes (+ 350% over a control run value of 0.024, with 12 of 13 models increasing, row labeled "hur45").
To visually illustrate one of the more important ndings in the table, Fig. 3 shows the tracks and intensities of the simulated U.S. landfalling hurricanes that are category 4 or 5 at landfall. The clear tendency for an increase in these very intense landfalling cases is seen across most of the different models, including

Summary And Conclusions
In this analysis, we explore future projection of U.S. landfalling TCs by examining a large number of cases generated using different climate model projections of large-scale environmental conditions, generally for the late 21st century under CMIP3 A1B scenario or the CMIP5 RCP4.5 scenario. We examined 13 different sets of projected warmed climate conditions based on 10 individual CMIP3 models or on the multi-model ensemble mean projection from the CMIP3 or CMIP5 models.
The most robust projections we simulate include an increase in the precipitation rate of U.S. landfalling TCs-a signal that increases in strength for the more intense categories of hurricanes, from + 18% averaged over all categories of TC (Cat 0-5) to almost + 200% for TCs that are Category 4 or 5 intensity at landfall. A robust reduction in TC frequency is projected for basin-wide counts (-25 to -45%) but this does not translate into a robust signal for U.S. landfalling TC frequency. Instead, a moderately signi cant and robust increasing signal emerges for the very intense U.S. landfalling TCs (category 4-5 storms). These average + 390% projected increase in frequency, with at least nominal increases projected for 10 of 13 models (with three of these projecting statistically signi cant increases) and no signi cant change for the remaining three models). Basin-wide PDI and duration are projected to decrease, while basin-wide intensity increases, but there is little signi cant signal in intensity projected for the U.S. landfalling TCs. Duration of TCs shows a signi cant decrease (averaging − 13%) while propagation speed shows a slight tendency to increase. The proportion of TCs that make U.S. landfall increases for almost all of the models, and the percentage increase in this proportion is especially large for the higher categories of TCs. It is this increase in proportion of U.S. landfalling TCs that contributes to an increase in the projected number of U.S. landfalling category 4-5 storms despite the projected decrease in overall numbers of TCs (including the weak decrease in basin-wide category 4-5 TCs in these projections).
Major caveats to our study include the limited skill shown by the downscaling framework in simulating the historical year-to-year variability of U.S. landfalling TC activity using SSTs and the NCEP/NCAR Reanalysis as large-scale climate forcings. There is also uncertainty in the climate change signal in the largescale environmental parameters, which is partly re ected in the spread of results across the different model-derived scenarios. The spread shown in our results cannot be assumed to represent the true con dence intervals on the results in this study. Despite these limitations, it is important to test our models with such scenarios and continue to compare modeled scenarios with the growing observational database to work toward a better understanding of the changes in landfalling hurricane risk facing our society in the coming century.