Vol.:(0123456789) Natural Hazards (2025) 121:15899–15923 https://doi.org/10.1007/s11069-025-07427-5 ORIGINAL PAPER How rare was the 2016–2022 tropical cyclone activity near the Caribbean coasts of Nicaragua and Costa Rica? Hugo G. Hidalgo1   · Tosiyuki Nakaegawa2   · David Romero3   · Eric J. Alfaro4   · Tito Maldonado5   · Yukiko Imada6   · Kohei Yoshida2  Received: 6 December 2024 / Accepted: 26 May 2025 / Published online: 22 June 2025 © The Author(s) 2025 Abstract Tropical cyclones (TC) are one of the synoptic systems that most affect Central America, from late spring to northern autumn, because they cause many direct and indirect impacts on the isthmus. Observational data of hurricane tracks and a suite of 10 downscaled Gen- eral Circulation Models from the Inter-Sectoral Impact Model Intercomparison Project round 3 and a Large Ensemble Global Simulation along with off-line TC trajectory genera- tion and tracking algorithms in the Atlantic/Caribbean basin were used in a detection and attribution study to determine if the observed run of 8 cyclonic events in 7 years (2016– 2022) near the coasts of Nicaragua and Costa Rica, could be unequivocally attributed to anthropogenic climate change. The results showed there is a large model to model variabil- ity, but that although the event is rare, it could not be proved that anthropogenic forcings have increased the probabilities of this high run of cyclones considering the 95% confi- dence level. More studies are needed to determine the exact time of possible emergence of a stronger signal in the near future. Keywords  Southern Central America · Climate change · Climate variability · Attribution · Detection * Hugo G. Hidalgo hugo.hidalgo@ucr.ac.cr 1 Escuela de Física, Centro de Investigaciones Geofísicas y Centro de Investigación en Matemática Pura y Aplicada, Universidad de Costa Rica, San José, Costa Rica 2 Meteorological Research Institute, Japan Meteorological Agency, Tsukuba, Japan 3 Escuela Nacional de Estudios Superiores Unidad Mérida, Universidad Nacional Autónoma de México, Ucú, Yucatán, Mexico 4 Escuela de Física, Centro de Investigaciones Geofísicas y Centro de Investigación en Ciencias del Mar y Limnología, Universidad de Costa Rica, San José, Costa Rica 5 Escuela de Física y Centro de Investigaciones Geofísicas, Universidad de Costa Rica, San José, Costa Rica 6 Division of Climate System Research, Atmosphere and Ocean Research Institute, The University of Tokyo, Chiba, Japan http://crossmark.crossref.org/dialog/?doi=10.1007/s11069-025-07427-5&domain=pdf http://orcid.org/0000-0003-4638-0742 http://orcid.org/0000-0003-4821-4190 http://orcid.org/0000-0002-1722-7514 http://orcid.org/0000-0001-9278-5017 http://orcid.org/0000-0001-9774-9082 http://orcid.org/0000-0002-1270-8335 http://orcid.org/0000-0002-2422-5584 15900 Natural Hazards (2025) 121:15899–15923 1  Introduction Tropical cyclones (TCs) are responsible for considerable impacts in Central America, a social and economically vulnerable region and exposed to cyclonic events from the Pacific and Atlantic/Caribbean basins. Such impacts can be direct or indirect (Hidalgo et  al. 2020) depending whether the cyclone landfalls on the site of interest (direct) or instead, the impacts are related to changes in the atmospheric circulation associated with climate patterns that occur a few hundred kilometers from the low-pressure center of the cyclone (indirect). In general, Nicaragua suffers from both types of impacts, while Costa Rica (and southern Central America in general) is more frequently affected by indirect impacts. Examples of devastating disasters occurred during the landfall of TCs such as Eta and Iota which crossed over the Caribbean coast of Nicaragua within just days apart in Novem- ber 2020 (Bissolli et al. 2021), exposing 3 million people to its impacts. Around 160,000 people were evacuated, 60,000 were housed in shelters and 21 lives were lost. Damage associated with the event was estimated at USD 743 million which is about 6.2% of Nica- ragua’s Gross Domestic Product (GDP; World Bank 2022). About 22 years after suffering the severe impacts of hurricane Mitch, Honduras was also impacted by Eta and Iota. In this country 437,000 people were directly affected and 96,000 people had to stay in shelters, while about 3.9 million people were indirectly affected. Economic losses were estimated in this case at USD 2,060 million, or around 0.8% of Honduras GDP in 2020 and 0.3% in 2021 (BID 2021). In Costa Rica, Hurricane Otto (2016) caused damage estimated in USD 331 million; Tropical Storm Nate (2017) in USD 606 million (Blunden and Arndt 2017; Hartfield et al. 2018; Maldonado et al. 2020) and Hurricane Eta (2020) in 216 USD million (Quesada-Román 2021; Quesada-Román et  al. 2024). To put these figures into context, only the economic impacts of Tropical Storm Nate caused damage close to 1% of Costa Rica’s GDP. More information about the impacts spatial and seasonal distribution of TCs can be found in Pérez-Briceño et al. (2016). During the 7-year period between 2016 and 2022 there were at least 8 cyclone incur- sions to a region surrounding the Caribbean coastal area Nicaragua and Costa Rica by the following cyclones: Matthew (September, 2016), Otto (November, 2016), Harvey (August, 2017), Nate (October, 2017), Eta (November, 2020), Iota (November, 2020), Bon- nie (June–July, 2022) and Julia (October, 2022). Therefore, the objective of this article is (using a suite of models and observations) to determine an answer to the question: is there a human fingerprint in the climate, which is producing changes in the cyclones’ trajectories such that it increased the probability of occurring of a high run of consecutive cyclones positioned close to the coasts of Costa Rica and Nicaragua from 2016 to 2022? To answer that question, we will perform a formal detection and attribution study of the past trajecto- ries to determine if climate change has influenced the increased risk in the region. The question then is to determine if anthropogenic climate change has caused TC tracks to shift toward more southern locations in the region, affecting both countries mentioned before. Previous studies have identified a poleward migration of TC lifetime maximum intensity (LMI) over the Pacific Ocean, but over the North Atlantic the shift is southward, albeit statistically insignificant (Kosin et  al. 2014; Moon et  al. 2015). Recent research (Studholme et  al. 2022), indicate conflicting results between previous studies regarding the response of TC tracks with warming, and there is a difficulty in establishing robust results with current data and methodologies, other than there would be a wider latitudinal spread with warming in TC tracks probably related to poleward expansion. A more recent observational analysis by Cao et  al. (2025) found a robust southward shift in the North 15901Natural Hazards (2025) 121:15899–15923 Atlantic hurricane genesis location in the satellite era during which the data are more reli- able (1979–2022). This shift is consistent with a growth in hurricane frequency in the southern part (10°–20° N) of North Atlantic (Cao et al. 2025). This increasing trend of hur- ricane frequency is intimately attributable to the decreasing vertical shear of zonal wind, resulting from a decreasing north–south temperature gradient (Cao et al. 2025). Also, using observations, Martínez et al. (2023) found a significant positive trend (1970–2021) in the number of Major Hurricanes (categories 3–5 on the Saffir–Simpson scale, i.e., sustained winds ≥ 50 m s−1) near the Caribbean coast of Nicaragua. Although there are a couple of studies on detection and attribution of the 2015–2019 drought in Central America (i.e. Pascale et al. 2021; Anderson et al. 2023), and a recent study on detection and attribution of climate extremes (Hidalgo et al. 2025), to our knowl- edge, no other detection and attribution study has been produced on TC incursions near the coasts of Nicaragua and Costa Rica. The usual method of detection of climate change is through the comparison of the signal (climate trend or such as in this case a particular meteorological or climate event of interest) with the statistical distribution of natural cli- mate variability “noise” obtained from global climate model simulations (e.g. Barnett et al. 2008; Pierce et al. 2008; Das et al. 2009; Hidalgo et al. 2009; Pascale et al. 2021) or sta- tistical methods to obtain “counterfactual” climates (e.g. Armal et al. 2018; Kawase et al. 2019; Mengel et al. 2021; Anderson et al. 2023); in other words, estimations of the climate variables without human influence. Both types of noise-generating methods reconstruct the climate data as it might have evolved in the absence of human influence, providing a framework for discerning the impact of anthropogenic factors on observed climate trends or on a particular meteorological or climatic event of interest (such as the occurrence of 8 cyclonic events near the coasts of Costa Rica and Nicaragua in a 7-year period). In this work, we will use the modeling approach for estimating the natural variability noise from (preindustrial) control runs from downscaled General Circulation Models (GCMs) of the latest version available or from the natural run of a large ensemble simulation from an atmospheric model (see Data section for details). Previous seminal work specifically on the detection and attribution of TCs can be found in Knutson et al. (2010, 2019) in which it was suggested detectable TC activity changes in some regions associated with TC track changes, while data quality and quantity issues create greater challenges for analyses based on TC intensity and frequency. Knudson et al. (2019) concluded that there is at least low to medium confidence that the observed pole- ward migration of the latitude of maximum intensity in the western North Pacific is detect- able, or highly unusual compared to expected natural variability. Detection of attribution of long-term trends aims to evaluate the possible contribution of anthropogenic forcings (e.g., greenhouse gases or aerosols) on past trend-behaviors (Camargo et al. 2023). Examples of positive trend detection and attribution of climate change can be found in Murakami et al. (2020), in which they found a significant positive trend (1980–2018) in observed TC fre- quencies just off the Caribbean coast of Nicaragua (see Fig. 1D of Murakami et al. 2020), which is consistent with Martínez et al. (2023) results mentioned before. Several studies that support increases in North Atlantic TC frequencies around that period are Kosin et al. (2020), Wang and Toumi (2021), Emanuel (2021), Klotzbach et al. (2022) and Pfleiderer et al. (2022). We should caution, however, that in some of these studies there is significant multidecadal signal in the North Atlantic TC frequencies (see for example Fig. 4 of Ema- nuel 2021; Goldenberg et al. 2001) and the positive significant trends that started around 1980 and finish in recent times are usually very strongly influenced by these natural vari- ations. Loehle and Staehling (2020) mistrust North Atlantic TC frequency trends, because relatively short (multi-decadal) data series are inherently likely to yield spurious trends in 15902 Natural Hazards (2025) 121:15899–15923 the basin (Loehle and Staehling 2020; Camargo et al. 2023). In fact, Vecchi et al. (2021) concluded that internal (e.g., Atlantic multidecadal) climate variability and aerosol-induced mid-to-late-twentieth century major hurricane frequency reductions have probably masked century-scale greenhouse-gas warming contributions to North Atlantic major hurricane frequency. Each study that support the increase in North Atlantic TC frequencies in recent times, has certain focus; for example, Wang and Toumi (2021) determined that there is an increasing trend (albeit non-significant) in the North Atlantic basin of TCs to be closer to the coast and a significant negative trend of the latitudinal position of the tracks; Klotzbach et al. (2022) focused their conclusions regarding an increasing trend in North Atlantic TC frequencies based in an even shorter period (1990–2021), and Pfleiderer et al. (2022) con- cluded that Atlantic tropical cyclone activity since the 1980s can be robustly ascribed to variations in atmospheric circulation as well as sea surface temperature increase. However, another recent study (Chand et al. 2022) using the 20 century reanalysis data, suggests that when very long-term trends are considered, there is instead a declining tropical cyclone frequency under global warming in all ocean basins. Although there are uncertainties regarding the possibility of detecting spurious trends in the North Atlantic due to the influence of multidecadal natural variability, the obser- vational positive trends have been subject to detection and attribution studies. For exam- ple, the 1980–2018 observational trend was also seen in all forcings (natural and anthro- pogenic) simulations, but not on natural forcing runs in the Murakami et al. (2020) study. The model fingerprint analysis revealed that the observed spatial pattern of TC frequencies trends since 1980 (including the mentioned increases in the Caribbean Sea TC frequencies) was already detected around 2010 in the “all forcings” runs at the 95% confidence. In other words, the probability that the observed TC frequencies trends between 1980 and 2010 are due entirely to internal variability is less than 5% and therefore the observed spatial pat- tern of TC frequencies trends cannot be explained solely by internal variability (Murakami et  al. 2020). Intergovernmental Panel on Climate Change or IPCC (2023) in AR6 con- cluded there was a change distinct from natural variability (with medium confidence). Conversely, detection and attribution of individual events is a relatively recent field (Camargo et  al. 2023). Reed et  al. (2022) found that human-induced climate change increased the extreme 3-hourly storm rainfall rates and extreme 3-day accumulated rainfall amounts during the full 2020 hurricane season for observed storms that are at least tropical storm strength (> 18 m/s) by 10 and 5%, respectively. When focusing on hurricane strength Fig. 1   Domain of the data used from the ISIMIP3 repository (0°–45° N, 140°–20° W) and spatial window (5°–14° N, 90°–70° W) to compute the incursions of events (red box) 15903Natural Hazards (2025) 121:15899–15923 storms (> 33 m/s), extreme 3-hourly rainfall rates and extreme 3-day accumulated rainfall amounts increase by 11 and 8%, respectively (Reed et al. 2022). There are only a few stud- ies exploring how anthropogenic climate change might have affected a particular TC active season (Camargo et al. 2023). For example, Murakami et al. (2018) reported that the active 2017 hurricane season in the North Atlantic might have been influenced by anthropogenic warming. For this reason, in particular, we argue that our study is novel, as it provides a detection and attribution of a 7-year very active run of years that occurred near the Nicara- gua and Costa Rica coasts. In order to compute the natural variability noise needed in detection and attribution studies using models, we will be using simulations produced using long runs of simulated climate keeping the greenhouse gasses forcings fixed to a certain pre industrial level (i.e. year 1850), or alternatively by dynamically downscaled runs of an atmospheric model using only natural forcings (e.g. solar, volcanic, aerosols); and therefore not including anthropogenic contributing factors. The goal of the analysis of detection of attribution stud- ies sometimes is not circumscribed to precipitation or temperature, as other hydrometeoro- logical variables can be of interest. Examples of these types of studies are found in Barnett et al. (2008), Pierce et al. (2008), Das et al. (2009) and Hidalgo et al. (2009), which studied hydrological and meteorological variables in the western United States, and we already mentioned previous applications of detection and attribution on TC frequencies as well. In our case, for example, we are interested in the cyclone trajectories that incursion within an area close to the coasts of Nicaragua and Costa Rica, and therefore those trajectories will be estimated using off-line algorithms from the climate outputs of the models. Basic physics behind TCs can be found in Emanuel (2003). Modeling efforts of cyclones trajectories for climate change studies have evolved through time. Basically, long time ago it was established that the coarse scale of GCMs compared to the spatial scale of the climate conditions that determines cyclogenesis and tracking, required downscal- ing procedures that usually involved the synthetic generation of cyclones coupled with a deterministic model of cyclones trajectories evolution (Knutson et al. 1998; Knutson and Tuleya 2004; Emanuel 2006; Takaya et  al. 2010). More recent work in this area can be found in Balaguru et al. (2023) who used a synthetic hurricane model to show increased intensity in hurricane frequency in the United States Gulf and lower East coast regions. Recently, examples of deterministic methods for computing the trajectories of GCMs and large ensemble simulations of atmospheric models can be found in Murakami et al. (2012), Diro et  al. (2014) and Maldonado et  al. (2020). In these references, dynamically down- scaled climate model output was used to generate tracks with off-line algorithms. The studies by Diro et  al. (2014) and Maldonado et  al. (2020) are based on Central America. In particular, other authors such as Murakami et  al. (2012), Diro et  al. (2014) and Zhang and Wang (2018) found that the simulations of realistic TC tracks are sensitive to both the model convection scheme, resolution and the forcing GCM. Several authors (e.g. Diro et  al. 2014; Sugi et  al. 2017; IPCC 2023; Lopez et  al. 2024) suggest that the future frequency of cyclones in the tropical Atlantic near our area of interest decreases in a future warmer climate; and Murakami et al. (2020) projected a decrease in TC frequency almost everywhere in the tropics in response to + 1% CO2 forcing. These conclusions are consistent with other findings such as projected increases in windshear (Vecchi and Soden 2007; Lopez et al. 2024), changes in the Hadley circulation (Altman et al. 2018; Sharmila and Walsh 2018; Studholme and Gulev 2018) and a northward shift of TC tracks over the Atlantic (Diro et al. 2014; Lucas et al. 2014) among other mechanisms cited in the litera- ture. However, it is unclear whether any identified changes are due to a basin-wide change in TCs frequency, or to systematic track shifts (or both) (IPCC 2023). It is important to 15904 Natural Hazards (2025) 121:15899–15923 note that uncertainty in these aspects is large and ultimately there is not an established theory for the drivers of future changes in the frequency of TCs. In summary, it is known that the uncertainty of the determination of future TCs frequencies is larger compared to other metrics such as indices intensity-based indices, translation speed (Kossin 2018), mean latitude where TCs reach their peak intensity (Kossin et al. 2014), and others (Walsh et al. 2016; Sugi et al. 2017), which is consistent with theoretical understanding (e.g., Ema- nuel 1987) and observations (e.g., Kossin et  al. 2020). It should be mentioned, however that other studies suggest an increase in the number of TCs in the future around Central America (Arias et al. 2021), but the discrepancy may be related to other reasons such that the increase corresponds to other tropical oceanic regions not exactly in our area of interest (e.g. Oochi et al. 2006; Wehner et al. 2010), or it is related to the more robust increase in the frequency of intense hurricanes, consistently with the theoretical models of Emanuel (1987) and Holland (1997) (Diro et al. 2014). Maldonado et al. (2020) studied the rapid intensification (RI) process of hurricane Otto that severely impacted Costa Rica and Nicaragua. The authors found that the Weather Research and Forecasting—Advanced Research WRF (WRF-ARW; Skamarock et  al. 2008) regional model with boundary and initial conditions provided by the Global Fore- cast System (GFS) analysis simulated more realistically the track and RI of Otto, and that consistently with Diro et al. (2014) the tracks are sensitive to the model convection scheme used (Maldonado et al. 2020). The spatial distribution of hydrometeorological impacts in November 2016 is presented in Alfaro et al. (2018). All these studies are usually based on determining changes at the ocean basin or rela- tively large-region extent. It is known that smaller regions have greater variability and the number of tracks that incurs the small area of interest is low. This poses a challenge in terms of establishing statistical significance and increases modeling biases in relation to the observed data. For these reasons the model data in this study was biased-corrected to match the statistical distribution of observed data in interest as it will be seen in following sections. We performed a bias-correction procedure to the number of cyclone incursions to make it more like the observed counts. Next section is the Data used in the analysis, then Methodology, next Results and the last section is Discussion and Conclusions. 2 � Data 2.1 � Observed data Observed data of TCs tracks from the Atlantic basin HURDAT2 dataset were obtained from the HURDAT2 project by the United States National Oceanographic and Atmos- pheric Administration (NOAA) National Hurricane Center (Landsea and Franklin 2013; https://​www.​nhc.​noaa.​gov/​data/). These data include TCs’ latitude and longitude coordi- nates every 3–6  h, which in some instances were converted into daily average positions using Coordinated Universal Time (UTC) time zone (see Hidalgo et al. 2023) and sustained wind speed for the period 1851–2023. Nicaragua and Costa Rica use the -6 UTC. Land- sea et  al. (2008) emphasizes that the pre-satellite era (prior to the 1960s) is particularly prone to undercounts of tropical cyclones, especially weaker storms or those that did not make landfall. These observational limitations stem from the sparse ship traffic and limited coastal monitoring during much of the twentieth century, which likely led to an incom- plete record of storm occurrences. Furthermore, the authors point out that methodological https://www.nhc.noaa.gov/data/ 15905Natural Hazards (2025) 121:15899–15923 changes and reanalyses over time introduce inconsistencies in the dataset, making long- term trend analysis challenging. This uncertainty complicates efforts to attribute changes in storm frequency or intensity to anthropogenic climate change, as apparent increases in storm activity could be influenced, in part, by improved detection capabilities rather than actual climate-driven trends. As such, careful interpretation of HURDAT2 is essential, par- ticularly when using it to inform adaptation policies or model future risk scenarios. 2.2 � Inter‑sectoral impact model intercomparison project round 3 model data To generate TC trajectories (using a modified version of Diro et al. 2014; see Methods), several climate data were obtained from the Inter-Sectoral Impact Model Intercomparison Project’s round 3 database (ISIMIP3, https://​data.​isimip.​org/). These data were accord- ingly downscaled and harmonized with ISIMIP3’s bias adjustment and statistical down- scaling algorithm (ISIMIP3BASD, Lange 2019, 2022). All ISIMIP3 data are available at 0.5° × 0.5° horizontal resolution and at daily time step. Variables required to generate the trajectories are surface air pressure (ps), near-surface air temperature (tas) and near-surface wind speed (sfcwind). The data were downloaded and cropped to the entire domain of the map shown in Fig. 1. Ten General Circulation Models (GCMs) from the Coupled Model Intercomparison Project 6 (CMIP6) and processed through the ISIMIP3 framework were selected for the analysis as shown in Table  1. In all cases, data from the historical, the SSP5-8.5, and the piCONTROL (preindustrial control) runs for each mode were obtained. The historical daily data for the mentioned variables from 1951 to 2014 were concatenated with the SSP5-8.5 from 2015–2023 to produce a long-term dataset for the calculation of the trajectories for the historical (HIST) runs. The piCONTROL data consists of 500 years of daily data with anthropogenic forcings fixed at the 1850 levels. We should mention that we selected the most pessimistic concentration and social scenario (SSP5-8.5) to use the most conservative data. However, inspection of other scenarios presents little deviations from each other during the period 2015–2023 (not shown). There are undetermined uncer- tainties of the climate projections that could arise from the estimation of the social, cli- mate forcings and physical aspects of the models, but being a short period that was added within the entire historical period used (1951–2023) for computing statistical distributions we considered that these uncertainties are considerably low compared to other uncertain- ties such as the depiction of realistic climate patterns from the historical scenario itself (1951–2014) or from versions of the contrafactual climates used here (natural or picontrol scenarios). In the Supplementary Information comparisons between climatological pat- terns of reanalysis data and models are presented (see also Sect. 2.4). 2.3 � Large model ensemble TCs trajectories from Japan’s Meteorological Research Institute A large-ensemble global simulation (LEGS) of cyclones’ model trajectories (calculated according to Murakami et al. 2012; see Methodology) with a grid spacing of approximately 60  km (Mizuta et  al. 2012) were developed and obtained from Japan’s Meteorological Research Institute (MRI). The data consist of 100 ensemble members with different initial conditions and sea surface temperature (SST) perturbations. The data are part of the data- base for policy decision-making for future climate change (d4PDF) (Mizuta et al. 2017). The original GCM data to produce the trajectories consists of global climate simulations conducted using the atmospheric general circulation model MRI-AGCM3.2 (Imada et al. https://data.isimip.org/ 15906 Natural Hazards (2025) 121:15899–15923 Ta bl e  1   C M IP 6 ge ne ra l c irc ul at io n m od el s ( C G M s) u se d in th is a na ly si s N at iv e re so lu tio ns a re sh ow n, b ut th e IS IM IP p ro ce ss ed a nd d ow ns ca le d th e da ta to 0 .5 ° ×  0. 5° re so lu tio n ID M od el C lim at e ce nt er Ru n N at iv e re so lu tio n 1 G FD L- ES M 4 G eo ph ys ic al F lu id D yn am ic s L ab or at or y (G FD L) r1 i1 p1 f1 Lo nl at (2 88  ×  18 0) 2 IP SL -C M 6A -L R In sti tu t P ie rr e- Si m on L ap la ce (I PS L) r1 i1 p1 f1 Lo nl at (1 44  ×  14 3) 3 M PI -E SM 1- 2- H R M ax P la nc k In sti tu te fo r M et eo ro lo gy (M PI -M ) r1 i1 p1 f1 G au ss ia n (3 84  ×  19 2) 4 M R I- ES M 2- 0 M et eo ro lo gi ca l R es ea rc h In sti tu te (M R I) r1 i1 p1 f1 G au ss ia n (3 20  ×  16 0) 5 U K ES M 1- 0- LL U K E ar th S ys te m M od el lin g r1 i1 p1 f2 Lo nl at (1 92  ×  14 4) 6 C an ES M 5 C an ad ia n C en tre fo r C lim at e M od el lin g an d A na ly si s ( C C C M A ) r1 i1 p1 f1 G au ss ia n (1 28  ×  64 ) 7 C N R M -C M 6- 1 C en tre N at io na l d e Re ch er ch es M et eo ro lo gi qu es /C en tre 3 E ur op ee n de Re ch er ch e et F or m at io n A va nc ee s e n C al cu l S ci en tifi qu e (C N R M ) r1 i1 p1 f2 G au ss ia n (2 56  ×  12 8) 8 C N R M -E SM 2- 1 C N R M r1 i1 p1 f2 G au ss ia n (2 56  ×  12 8) 9 EC -E ar th 3 EC -E A RT H c on so rti um r1 i1 p1 f1 G au ss ia n (5 12  ×  25 6) 10 M IR O C 6 A tm os ph er e an d O ce an R es ea rc h In sti tu te (T he U ni ve rs ity o f T ok yo ), N at io na l I ns tit ut e fo r E nv iro nm en ta l S tu di es , a nd Ja pa n A ge nc y fo r M ar in e- Ea rth S ci en ce a nd T ec hn ol og y. (M IR O C ) r1 i1 p1 f1 G au ss ia n (2 56  ×  12 8) 15907Natural Hazards (2025) 121:15899–15923 2019). Data from two scenarios were downloaded: (1) the simulations with all forcings (composed of historical SSTs and sea ice thickness/concentration based on COBE-SST2 and the greenhouse gasses and solar irradiance from 1951 to 2006) (Hirahara et al. 2014) was concatenated with the RCP4.5 emission scenario (2006–2023) to obtain a continuous simulation from 1951 to 2023 and (2) the non-warming simulations (non-W), forced by the historical natural forcing agents over the same period, and also with counterfactual “natu- ral” SST and sea ice estimated by removing the warming trends observed in the twentieth century (Imada et al. 2019). The anthropogenic forcings were fixed at a value of year 1850 in this natural run (see Shiogama et al. 2016 for details). All track data from this section is available at a 6-hourly time step. 2.4 � Reanalysis data The data in this section were used to evaluate the ISIMIP3 model performance. Monthly data for variables ps, tas and sfcwind from the European reanalysis (ERA5; Hersbach et al. 2020) at 0.25° × 0.25° spatial resolution, covering the period January 1940-Decem- ber 2024, as well as the “55-year” Japanese reanalysis (JRA; Kobayashi et  al. 2015) at 0.375° × 0.375° spatial resolution, for the period October 1988-September 2021 were obtained from MRI servers. Both products were linearly interpolated to the same ISIMIP3 grid, cropped to the domain shown in Fig. 1 and also to the common full-year period from 1989 to 2020. Low-frequency variability is found to be well represented in ERA5 and from 10 hPa downwards general patterns of anomalies in temperature match those from another reanalysis, including JRA (Hersbach et al. 2020). Comparisons of June to November aver- ages of selected variables between both reanalyses are found in section S.1, while wavelet analysis of the variables are found in section S.2, both in the Online Resource. 3 � Methodology Flow charts with the methodology can be found in Suppl figures S3 to S9 in the Online Resource. All trajectory data were used to compute the incursions of the cyclones in a spa- tial window (5°–14° N, 90°–70° W; see red box in Fig. 1) close to the Nicaragua and Costa Rica Caribbean coasts. The data were subsequently transformed by computing the 7-year moving sums. The reason for this transformation is to highlight the 2016–2022 period when 8 events in 7-years incurred into the window. This would allow us to determine how rare the 2016–2022 sequence of 8 events is compared to the distribution of 7-year TC fre- quencies in HIST, piCONTROL and natural runs from the models. Also, the probabilities of 8 events in 7 years based on the HURDAT2 were computed using Poisson distribution (Elsner and Jagger 2013; Romero and Leon-Cruz 2024) along with a Monte Carlo simu- lation of 1000 iterations shuffling the annual counts to assess statistical significance. We also conducted a spatiotemporal sensitivity analysis of tropical cyclone (TC) activity pat- terns in the study area. Using quality-controlled HURDAT2 data (1851–2023), we exam- ined probability variations of experiencing ≥ 8 TCs within overlapping 7-year moving win- dows across multiple spatial configurations. Our methodology evaluates two distinct spatial perturbation approaches: (1) uniform bidirectional expansions (± 0.5°, ± 1°, ± 2°, and ± 5° in all directions) and (2) unidirectional boundary modifications (independent ± 2° expan- sions of northern, southern, eastern, and western limits). For each spatial configuration, we calculate Poisson-derived exceedance probabilities while maintaining fixed temporal 15908 Natural Hazards (2025) 121:15899–15923 parameters (start years: 1851, 1900, 1950, 1970) and intensity thresholds (all TCs ≥ 0 kt and strong TCs ≥ 35 kt). When comparing models and observations there is a need to bias-correct the models’ data to compare probabilities of real-world cyclone occurrences with the model simula- tions. Inspection of the histograms, probability density functions (pdfs) and cumula- tive density functions (cdfs) of HURDAT observations and models revealed that a quan- tile–quantile mapping using an empirical distribution was not enough to remove the bias (not shown). However, a gamma distribution was a good choice for this purpose, consider- ing the type of data of the transformed counts in the window of the region of interest. The procedure of Piani et al. (2010) was used for this type of correction. More details regarding the bias-correction procedure can be found in the Results section. To have an estimation of the skill of the 10 ISIMIP3 models in reproducing historical climate of the reanalyses, the climate data of the ISIMIP3 simulations were ranked by each model’s skill of reproducing the June to November (considered the hurricane season in the Atlantic basin) climatologies of ps, tas and sfcwind variables. To compare the (2-dimen- sional) monthly climatologies patterns between observed (ERA5 and JRA) and models (ISIMIP3 simulations of the models shown in Table 1) the Taylor (2001) skill was used: where R is the pattern spatial correlation between model and observation, ⌢𝜎f is the ratio of the standard deviation of the model ( �f  ) to the observed pattern ( �r ): ⌢ 𝜎f = 𝜎f∕𝜎r and Ro is the maximum correlation attainable (considered equal to 1 in our case). Once all the S’s related to ps, tas, and sfcwind for individual climatological months from June to November were obtained, an aggregated index that contains all the S from all variables and months was calculated for each model, by considering the euclidean distance of each patterns’ S to the perfect score (S = 1), in the following manner: The smaller the Δmodel , the better skill it has on reproducing the seasonal patterns of the reanalysis. Trajectories were calculated using two different methods: (1) For the ISIMIP3 data the calculation was performed using a simplified algorithm version of Diro et  al. (2014), as the one used in Fuentes-Franco et al. (2014, 2015) and Maldonado et al. (2020). The origi- nal version of the computer code was prepared for 6-hourly data, but it was modified to accept daily meteorological (ps, wspd and tas) records to adapt it to the time scale resolu- tion and availability of the ISIMIP3 data. Also, the thresholds for cyclogenetic and cyclone tracking regions associated with the variables ps, tas and sfcwind in the computer code were adjusted from our version to values obtained from the HURDAT2 data averaged at daily time scales. (2) For the MRI-AGCM3.2, the trajectories were calculated according to Murakami et al. (2012). In this case, 6 types of criteria were used for determining cyclo- genesis and cyclone tracking. Three different convection schemes of the models were cal- culated, and if applicable, different thresholds of the variables of the model were selected for each convection scheme. The criteria are: a maximum threshold on relative vorticity; a maximum wind speed threshold at 850 hPa; confirmation of the presence of a warm core (1) S = 4(1 + R) ( ⌢ 𝜎f + 1 ⌢ 𝜎 f )2 ( 1 + Ro ) (2) Δmodel = √ ( Sps June− 1 )2 +⋯ + ( SpsNovember − 1 )2 + ( Stas June− 1 )2 +⋯ + ( Stas November− 1 )2 + ( Ssfcwind June− 1 )2 +⋯ + ( Ssfcwind November− 1 )2 15909Natural Hazards (2025) 121:15899–15923 aloft, indexed with the exceedance of a threshold of the sum of temperature deviations at 300, 500, and 700  hPa; a check that the maximum wind speed at 850  hPa is greater than the maximum wind speed at 300 hPa; a condition for the North Indian Ocean only to remove tropical depressions which requires that the maximum wind speeds must be less than 200 km from the detected storm center and the duration of each detected storm must exceed 36 h (Murakami et al. 2012). 4 � Results 4.1 � Evaluation of TC trajectories in ISIMIP3 models Atlantic/Caribbean basin-wide climatological counts of the ISIMIP3 models compared to the HURDAT2 data are shown in Fig. 2. As can be seen, some of the model distribu- tions are generally acceptable for reproducing the counts around the peak of the hurricane Fig. 2   Monthly distribution of TCs in the Atlantic Basin from HURDAT2-OBS and 10 CMIP6 models from ISIMIP3 15910 Natural Hazards (2025) 121:15899–15923 season, but the most common error is that most of them simulate too many events in the winter and spring. There are also biases in the absolute number of events per year in the modeled data. The models that showed the distributions closest to the observations are the GFDL-ESM4, CNRM-CM6 and CNRM-ESM2-1. In order to compensate for these and other types of biases we performed the bias-correction procedures mentioned before in the counts inside the box of interest. 4.2 � Monte Carlo simulations with the HURDAT2 data Table 2 shows the probabilities of registering 8 events in 7 years, along with the 95% confi- dence level counts from a Monte Carlo (MC) simulation of 1000 runs constructed by shuf- fling the years of the cyclone counts inside the red box in Fig. 1. In general, for all types of cyclones, the probabilities decrease as the initial year is closer to the present, as in the initial part of the record, there were more frequent incursions of trajectories inside the area of interest. The given probabilities are within the 95% confidence interval except for the occurrence of tropical depressions in the period 1970–2023, with a particularly low probability. This means that, although the incidence of tropical depressions or greater in the region is relatively frequent, its con- centration has been exceptional for the period under consideration. Also, the decrease of the probabilities over the time implies that in the part of the record with less human influence, the incursions were more likely to occur. This result has to be taken with caution, as the HURDAT2 data is known to present inhomogeneities over time due to different methodologies for registering and observing the cyclones (Moon et al. 2019). However, the HURDAT2 has some correction of such types of errors since it is really a Renalysis, plus, errors in databases such as HURDAT2 and IBTrACS are more likely to be underestimations of TC occurrences before the satellite or hurricane reconnaissance aircraft era (Neumann and Elms 1993; Landsea and Franklin 2013) due to the lack of detection. Consequently, the maximum probabilities, considering a study period start- ing from 1900 with a probability of 8 events in 7 years = 0.1043, linked to a sequence of recorded TCs during the 1930s, are most likely a true reflection of the variability in cyclone activity in the region. It is also worth noting that this series, like the recent one, occurred during a positive phase of the Atlantic Multidecadal Oscillation (AMO) Table 2   Poisson probabilities of observing 8 events in 7 years inside the window of Fig. 1 Also shown are the 95% confidence probabilities from Monte Carlo simulations of 1000 members. The results are shown for different wind-speeds and initial years. Final year is 2023. CI, confidence interval Minimum category Initial year Prob. HURDAT2, 8 events in 7 years CI 95% Tropical depression or greater 1851 0.0749 0.024–0.129 1900 0.1043 0.073–0.205 1950 0.0978 0.084–0.266 1970 0.0800 0.117–0.327 Tropical storm or greater 1851 0.0353 0–0.041 1900 0.0496 0–0.066 1950 0.0000 0–0.071 1970 0.0000 0–0.08 15911Natural Hazards (2025) 121:15899–15923 (Enfield et al. 2001; Elsner 2006; Luo et al. 2023). Finally, it seems that the Tropical Depression occurrence dispersion is great after 1970. In consequence, although the cyclonic activity is very high during the period in average giving higher values of the Confidence Interval (CI), the actual probability is lower than including the previous years, becoming exceptionally low, below the confidence interval (see Table 2). The sensibility analysis revealed that probability values are significantly influenced by both edge modifications and zone extensions, with spatial factors having notably higher impact than temporal variables. For the 1970 period specifically, the North edge modification shows the strongest effect with a probability of 0.5563, substantially greater than East (0.253), South (0.08), and West (0.08) edge modifications (Suppl Fig S10). Zone extension analysis demonstrates that increasing window size from ± 0.5° (0.2212) to ± 5° (approximately 0.63) dramatically increases probability values for the 1970 dataset (Suppl Fig S11). Concerning the tropical storm wind-force events in the study area, the differences are even greater (Suppl Fig S11). The pronounced prob- ability variations across spatial parameters, compared to more modest variations across time periods, confirm that the model exhibits greater sensitivity to spatial configura- tions than temporal ranges. This spatial dependency is particularly evident in how the northern boundary modifications and larger window sizes produce the most significant probability variations, suggesting careful consideration of spatial parameters is critical for model reliability. 4.3 � Bias correction For illustration, the bias-correction of the LEGS data is found in Fig. 3. In the left and the middle subfigures, the pdf and cdf plots for the raw data are shown, while the right- most subfigure shows the cdf after bias-correction using quantiles from the gamma distribution. As can be seen, the adjustment of the historical scenario compared to the observed (HURDAT2) cdf is adequate, except for the left tail, which is not of interest in our case, as the estimation of the probabilities of the right tail is what concerns us. The natural scenario was corrected by estimating the probabilities of the natural data using gamma estimates corrected using the fitted parameters for the historical scenario (previously obtained from the adjustment of the historical scenario against observed HURDAT2 data). The data for the ISIMIP3 calculated trajectories for the 10 climate models were adjusted to obtain the probabilities of observing 8 events in 7  years in their corre- sponding historical and piCONTROL scenarios. Table 3 shows those probabilities of occurrence in the ISIMIP3 models and the LEGS simulations, while Fig. 4 shows the pdfs of the HURDAT2 data and the bias-corrected ISIMIP3 model simulations. Note that in Fig. 4 the bias-correction of the ISIMIP3 produced a very good fit of the simulated pdfs of the historical runs with respect to the HURDAT2 observations. However, due to the differing model physics and internal structure (and in a certain way this could be magnified by the TC tracking algorithm to climate variables), the picontrol pdfs show significant differences depending on the model. This results in different probabilities of observing 8 events in 7-years (the signal) for the internal variability noise from model to model. For this reason, in the next section, we explore the possibility of ranking the mod- els based on their suitability to produce realistic climate patterns and produce alternative results using a subset of only the best qualified models. 15912 Natural Hazards (2025) 121:15899–15923 4.4 � Evaluation of climate variables in ISIMIP3 models Since some of the risk ratios (probabilities observing a run of 8 events in 7 years in the box of Fig. 1 for the historical scenario divided by the probability for the piCONTROL or natu- ral run) for the ISIMIP3 simulations are greater than one in some cases and less than one in others, we decided to evaluate the generating variables of the trajectories to determine which of the models can simulate those variables more realistically. As ground-truth we used the ERA5 and the JRA reanalyses. Comparison of the June to November climatological means between both reanalyses show small differences (Suppl Fig. S1). The ISIMIP3 models were evaluated over the climatological months using the skill score index of Eq. 1, while the com- bination of the skills for different variables and models was integrated using the ∆ index of Eq. 2. Results show that the average risk ratio for the best-ranked 5 models is 1.69, but there is great variability between the models. On the other hand, the LEGS results suggest a mean risk ratio of 0.93 (the same as the ISIMIP3-MRI model). Like Pascale et al. (2021), in Fig. 5 a bootstrap-with-replacement using 1000 simulations was performed on the risk ratios. As can be seen, 6 out of 10 models have risk ratios statistically indistinguishable from 1 and from the 5 best ranked models, 3 showed ratios greater than 1. However, both the multi-model ensem- ble (MME) of all ISIMIP3 models and the ensemble of the best 5 ranked ISIMIP3 models Fig. 3   Probability density functions (PDFs; left) and cumulative density functions (CDFs; middle) for his- torical period of 7-year sums counts of cyclones that incursion in the window of Fig. 1 (5°–14° N, 90°–70° W). Three lines are shown: the historical (observed) HURDAT2 data (blue dashed line), the raw historical (anthropogenic + natural forcings) simulated period from the MRI-AGCM3.2 model (red solid line), and the raw natural simulation (natural forcings only) from the same model (green solid line). The subfigure of the right depicts the bias-corrected version (using a gamma distribution) of the historical simulation (red line with circles), the bias-corrected version of the natural simulation (dashed green line), and the value of the 7-year running sums in the HURDAT2 data for the period of interest (2016–2022) 15913Natural Hazards (2025) 121:15899–15923 (BEST) concludes that the risk ratio is indistinguishable from 1 and therefore, the 2016–2022 run of 8 consecutive cyclonic events near the coasts of Costa Rica and Nicaragua was proba- bly not significantly (at the 95% level) exacerbated by anthropogenic causes. This is consistent with the result found for the LEGS estimation. Although in the analysis shown here the results were calculated using daily averages (to match the time resolution of the ISIMIP3 data) of the cyclone counts, if the analysis is repeated using 6-hourly data for the HURDAT2 cyclone counts resulted in probabilities of approximately the double as those of Table 3. But, since the doubling effect is observed for all types of runs (the piCONTROL, natural and historical cases), the risk ratios are approximately the same and the conclusions derived from Fig. 5 do not change. Note however, that the historical probabilities of the models and the HURDAT2 probabilities of observing 8 events in 7-years is around p = 0.068, so although the occurrence of such high quantity of storms near the Costa Rica and Nicaragua coast was probably not exacerbated by anthropogenic causes, it was a relatively rare occurrence of events. Table 3   Gamma probabilities of observing 8 cyclones in 7 years in the box shown in Fig. 1 The RANKS REANALYSIS columns correspond to the rank of the ISIMIP3 models in simulating the June to November climatologies of surface pressure, surface temperature and wind speed, input variables that are used in the Diro et al. (2014) algorithm for determining cyclones trajectories. ERA5: ERA5 reanalysis and JRA: Japanese reanalysis. Risk ratio: probabilities observing a run of 8 events in 7 years in the box of Fig. 1 for the historical scenario divided by the probability for the piCONTROL or natural run. Data from 1950 to 2023 Model Prob (X ≥ x) Risk ratio Ranks reanalysis piCONTROL Historical ERA5 JRA piCONTROL and historical noise IPSL 0.01080 0.06710 6.21 8 8 UKESM1 0.16030 0.06750 0.42 9 9 MRI 0.07400 0.06770 0.91 5 5 CanESM5 0.32760 0.06760 0.21 1 1 EC-Earth3 0.01680 0.06750 4.02 4 3 MIROC6 0.10210 0.06720 0.66 6 6 MPI 0.03850 0.06730 1.75 2 2 GFDL 0.02000 0.06710 3.36 3 4 CNRM-CM6-1 0.07490 0.06740 0.90 7 7 CNRM-ESM2-1 0.08970 0.06700 0.75 10 10 Model Natural Historical Risk ratio Natural noise (100 ensemble members) alternative trajectory method MRI-AGCM3.2 0.0781 0.0675 0.86 HURDAT2 data Obs 0.0675 15914 Natural Hazards (2025) 121:15899–15923 5 � Discussion and conclusions In this article, probabilities and risk ratios from two types of models (ISIMIP3-CMIP6 MME and LEGS) were used to determine if the sequence of 8 TCs in 7-years that occurred in 2016–2022 was exacerbated by anthropogenic climate change. We should mention that ISIMIP3-CMIP6 MME and LEGS are fundamentally different in the type of Sea Surface Temperatures (SSTs) that are used in the models. The MME simulations predict possible realizations of SSTs in their simulations, while LEGS simulations use prescribed SSTs. The SST anomalies in the 7-years in the MME simulations is likely to be different from the observed anomalies used in the LEGS simulations, but the comparison of probabilities and risk ratios in all the analysis presented here is based in the comparison of the observed sig- nal of 8 cyclones in 7-years recorded in HURDAT2, compared to the bias corrected noise of both types of models. There are large uncertainties in determining cyclone tracks’ frequency distribution using model outputs in off-line trajectories algorithms. While global averages of the count of cyclones are difficult to estimate due to ignoring important transient local effects (Wehner 2021), the problem does not disappear when the area of interest is relatively small and there is a small number of events entering the subregion to perform statistics. For this Fig. 4   Probability density functions (pdfs) of bias-corrected 7-year running sums of the number of cyclones in the red box shown in Fig. 1 for 10 models from ISIMIP3. Obs: HURDAT2 observations (dotted blue line with circles), sim: historical simulations from the model with all forcings (red thick line); piCONTROL: control run data with greenhouse gasses fixed at preindustrial levels (i.e. levels fixed at year 1850) (thick green line) and the 2016–2022 value of 7-year events from the HURDAT2 observed data (dotted black vertical line) 15915Natural Hazards (2025) 121:15899–15923 reason, we approached the problem using 7-year moving sums and a bias-correction pro- cedure that was able to correct the shape of the distribution of incursions in the area of interest in the Caribbean coast of Costa Rica and Nicaragua. However, inspection of the large different model to model estimated probabilities indicates there are still great uncer- tainties remaining in the results. Moreover, consistently with other studies (e.g. Knutson et al. 2019; IPCC 2023), we also determined that (even after bias-correction), there is great inter-model variability in terms of their estimation of the risk factor. However, the fact that the results for the MME, the ensemble of the best 5 models and the LEGS are consistently implying that the risk factor is less than 1, but statistically indistinguishable from 1, gives us confidence that the conclusion of the run of 8 incursions in 7 years that occurred from 2016 to 2022 (although relatively rare in the historical and HURDAT2 data) was probably not sufficiently influenced by anthropogenic causes. This result is consistent with other studies (e.g. Dominguez et al. 2021) suggesting, if any, there is a reduction in the number of TCs near the region of interest in the future (compared to the historical epoch, a period more exposed to anthropogenic influence). Other studies that support a reduction of future frequency of TCs are summarized in the 6th Assessment Report of the Intergovernmental Panel on Climate Change (2023). A seminal paper by Gray et  al. (1984) identified the relationship between El Niño- Southern Oscillation (ENSO) and tropical cyclones in the Caribbean/Atlantic basin. Basi- cally, the warm phase of ENSO (El Niño) is associated with stronger shear and fewer events, while the opposite is true for the cold phase of ENSO (La Niña). In addition, Fig. 5   Risk ratios (probabilities observing a run of 8 events in 7 years in the box of Fig. 1 for the histori- cal scenario divided by the same kind of probabilities but for the piCONTROL in the ISIMIP3 models of Table 1 or natural run in the case of LEGS) and their 95% confidence intervals. MME: Multi-model ensem- ble of the 10 ISIMIP3 models. BEST: Ensemble of the best 5 ranked ISIMIP3 models according to Table 3. LEGS: large-ensemble global simulation from MRI-AGCM3.2 model 15916 Natural Hazards (2025) 121:15899–15923 Enfield and Alfaro (1999) work identified that the relative state of Pacific and Atlantic sea surface temperature anomalies (SSTas) reinforces ENSO teleconnections when both oceans present SSTas of different sign (a dipole) than when they are of the same sign. Hidalgo et al. (2015, 2019) presented connections between Pacific and Caribbean/Atlantic climate through the connection of ENSO with the strength of the Caribbean trade winds. These types of connections may have been altered (or other phenomena may have become important) in recent times due to the effect of anthropogenic climate change in the region. For example, it is unknown if more recent extreme rainfall in the Pacific coast of Costa Rica in the Hurricane Season 2024, is related to the frequent occurrence of episodes of the Central American Gyre (Papin et al. 2017), probably coupled in synoptic terms with persistent warming in the Caribbean/Atlantic basin. The Gyre may also have favored a few cyclones in the Caribbean (e.g. cyclones Rafael and Sara in 2024). This type of increase of the Gyre episodes during La Niña episodes and associated with warming in the Caribbean/ Atlantic basin deserves more research. Although we could not find detection, the TC run of 8 events in 7-year was rare in the historical observations and models. This could be some indication of the suggested previous shift of the TC tracks toward the Caribbean, consist- ently with previous work cited and discussed in the introduction. However, as previously mentioned also in the introduction, the results for small regions such as the ones studied here present challenges due to the difficulty of models to reproduce small scale climate fea- tures due to their relatively coarse resolution and also to generate realistic TC tracks from these climate output. Those are limitations that should be taken into consideration when analyzing the results. Overall, even though this climatological set of events, or the trend of cyclone counts in Central America (not shown in this article) has not produced exacerbation of the prob- abilities of TC frequencies in the 2016–2022 period, it may be too soon for a strong signal to emerge, given the large natural climate variability in the tropical regions. Indeed, for the Yucatan Peninsula, Romero and Leon Cruz (2024) related the increase, compared to 1980s in TC frequency in the last decades to the warm AMO phase. Considering the potential of disasters of great magnitude in the vulnerable region of Central America, a research priority should be focused on using statistical methods for determining if there is a time of emergence (e.g. Hyun et al. 2020; Ying et al. 2022) soon of such a signal. An integral assessment for risk management purposes requires historical information of TC’s frequency of occurrence (Castillo-Loeza et al. 2025). The present study provides information on whether the frequencies of TC incursions near the Costa Rica and Nicara- gua Caribbean coast in the period 2016–2022 have shown statistical certainty (at the 95% confidence level) of being altered by anthropogenic climate change. This determination is important as it will answer the question whether these frequencies would increase in the future, affecting the potential impacts and increasing the risk levels. It is known that the potential impacts not only depend on the climate hazard, but also on exposure, vulnerabil- ity and capacity (UNDRR 2018). In this work we presented the physical part of risk, but we agree that the risk management should consider those social aspects. However, we also recognize that the Caribbean coast of Costa Rica and especially Nicaragua are character- ized by high socio-economic vulnerability and therefore this type of study aims to provide useful information for decision makers in the Central America region. Changes in the frequency and tracking of TCs present significant implications for climate change adaptation in Central America. Historically, this region has been highly vulnerable to TCs, which can lead to devastating flooding, landslides, and economic dis- ruption. Although the results presented that the frequent incursion of TCs in the region during the period 2016–2022 could perfectly be explained by natural variability, it was a 15917Natural Hazards (2025) 121:15899–15923 rare event, and that may indicate that changes are underway towards shifts in the TC tracks that may affect the region. This implies that Central America may face in the future more frequent encounters with destructive storms, requiring adaptations in infrastructure design, disaster risk management, and early warning systems. In addition, altered TC tracks could expose previously less-affected areas to heightened risk, complicating regional adaptation strategies. For Central America, this could mean more direct landfalls in regions like Hon- duras and Nicaragua, or increased rainfall in mountainous areas, intensifying flood risks. Consequently, adaptation measures must integrate dynamic risk assessments that consider future variability in storm tracks, not just historical patterns. Enhancing regional coopera- tion, investing in resilient infrastructure, and incorporating climate projections into urban planning are essential to mitigate the growing threats posed by evolving TC behavior in a warming climate. A potential increase in the frequency of tropical cyclones in Central America would have profound socioeconomic implications, particularly for vulnerable communities already facing high levels of poverty and limited adaptive capacity. Increased storm activity threatens critical sectors such as agriculture, fisheries, and tourism, which are major sources of employment and national income in countries like Honduras, Nica- ragua, and Guatemala. Crop destruction and disruptions to food supply chains can exac- erbate food insecurity and inflation, disproportionately affecting low-income populations (Lozano-Gracia et al. 2009). Moreover, more frequent displacements due to storm-related destruction could overwhelm already strained housing and health systems, leading to long- term socioeconomic instability and increasing migration pressures both within and beyond the region. The costs of reconstruction and disaster response also place significant burdens on national budgets, potentially diverting funds from essential services such as education and healthcare. Therefore, the growing cyclone threat necessitates not only improved phys- ical infrastructure but also robust social safety nets and economic diversification strategies to protect livelihoods and promote resilience. Using climate projections, future research may identify the time of emergence of the climate change signal on the TC frequency of the cyclones near the coast of Costa Rica and Nicaragua. The key findings can be summarized as: • There is great uncertainty in the determination of detection and attribution of TC fre- quencies in small regions. This is related to data scarcity imposed by the historical tropical cyclone data in specific locations. Consistently with Romero et al. (2025), we found that location-specific biases can remain undetected even after bias-correction. • Consistently with Meiler et al. (2025), we found great model-to-model differences in the results, related to different physics and internal configurations. • The probability of observing 8 or more cyclones within a 7-year period is highly sensitive to how the region’s boundaries are defined, particularly its northern limit. Expanding the area significantly can substantially affect the results. Specifically, the probabilities are multiplied by ≈2 when extending eastward and by ≈6 when extend- ing northward. In contrast, expanding the region westward or southward doesn’t mean- ingfully change the probability, though a westward extension might incorporate Pacific basin tropical cyclones. The high sensitivity toward the north is due to the relatively high cyclonic activity with annual occurrences of events in the Yucatan Peninsula (Romero and León-Cruz 2024) • We proposed to bias correct the data of TC frequencies using a gamma distribution and that seemed to provide a good adjustment of the historical probabilities, however the mentioned model to model differences and unaccounted biased produced that different probabilities of the picontrol runs of the ISIMIP3 simulations depending on the model. 15918 Natural Hazards (2025) 121:15899–15923 • However, there is some consistency: the multimodal ensemble of all ISIMIP3 simula- tions, the ensemble of the 5-best ranked ISIMIP3 models and the results of the LEGS simulations all suggest that the 2016–2022 run of 8 cyclonic events near the Carib- bean coast of Nicaragua and Costa Rica can still be part of natural climate variability, although is a rare event (p = 0.0675 when considering the observed HURDAT2 data). Supplementary Information  The online version contains supplementary material available at https://​doi.​ org/​10.​1007/​s11069-​025-​07427-5. Acknowledgements  This work was partially produced when HH was on sabbatical license from UCR dur- ing a short stay in the year 2024 at the Meteorological Research Institute (MRI) in Japan, supported by the Ministry of Land, Infrastructure, Transport and Tourism of Japan. HH thanks the hospitality, production of updated LEGS simulations and valuable discussions with MRI and University of Tokyo researchers and students. HH, EA thanks the UCR School of Physics for giving us the research time to develop this study. Thanks to Andrés Gamboa Chacón for assistance with Fig. 2 and Bryan Rodríguez Nájera for the reference format assistance of the manuscript. Thank you to Marco Acosta Quesada and Dennis Jiménez Badilla for their assistance with Fig. 1. Authors’ contributions  All authors contributed to the study conception and design. Material preparation, data collection and analysis were performed by HH, DR, TM, YI, and KY. The first draft of the manuscript was written by HH in discussion with TN, EA, DR and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript. Funding  HH and EA wish to acknowledge partial funding for this study through the following Vicerrectoría de Investigación, Universidad de Costa Rica grants: B9454 (supported by Fondo de Grupos), C2103, C3991 (UCREA), A4906 (PESCTMA) and B0-810. EA, and HH were partially supported by a grant awarded by the International Development Research Centre (IDRC), Ottawa, Canada, and the Central American Univer- sity Council (CSUCA-SICA) to the Red Centroamericana de Ciencias sobre Cambio Climático (RC4) pro- ject (CR-66, C4468, SIA 0054-23, the opinions expressed here do not necessarily represent those of IDRC, CSUCA, or the Board of Governors). TN was supported by JSPS KAKENHI Grant Number JP23KK0077 and by the Ministry of Education, Culture, Sports, Science, and Technology Program for the Advanced Studies of Climate Change Projection (SENTAN; JPMXD0722680734). DR thanks the “Programa de Apoyo a Proyectos de Investigación e Innovación Tecnológica (PAPIIT)” of UNAM grant IA101823. Declarations  Competing interests  The authors have no relevant financial or non-financial interests to disclose. Open Access  This article is licensed under a Creative Commons Attribution-NonCommercial-NoDeriva- tives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduc- tion in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/. References Alfaro EJ, Hidalgo HG, Maldonado T et al (2018) A tri-dimensional approach to climate sciences: lessons from a Central American university. Caribb Q 64:26–56. https://​doi.​org/​10.​1080/​00086​495.​2018.​ 14353​33 https://doi.org/10.1007/s11069-025-07427-5 https://doi.org/10.1007/s11069-025-07427-5 http://creativecommons.org/licenses/by-nc-nd/4.0/ https://doi.org/10.1080/00086495.2018.1435333 https://doi.org/10.1080/00086495.2018.1435333 15919Natural Hazards (2025) 121:15899–15923 Altman J, Ukhvatkina ON, Omelko AM et al (2018) Poleward migration of the destructive effects of tropi- cal cyclones during the 20th century. Proc Natl Acad Sci USA 115:11543–11548. https://​doi.​org/​10.​ 1073/​pnas.​18089​79115 Anderson W, Baethgen W, Capitanio F et al (2023) Climate variability and simultaneous breadbasket yield shocks as observed in long-term yield records. Agric For Meteorol 331:109321. https://​doi.​org/​10.​ 1016/j.​agrfo​rmet.​2023.​109321 Arias PA, Bellouin N, Coppola E et al (2021) Technical summary. In: Masson-Delmotte V, Zhai P, Pirani A et al (eds) Climate change 2021: the physical science basis. Contribution of working group I to the sixth assessment report of the intergovernmental panel on climate change, pp 33–144. Cambridge University Press. https://​doi.​org/​10.​1017/​97810​09157​896.​002 Armal S, Devineni N, Khanbilvardi R (2018) Trends in extreme rainfall frequency in the contiguous United States: attribution to climate change and climate variability modes. J Clim 31:369–385. https://​doi.​ org/​10.​1175/​JCLI-D-​17-​0106.1 Balaguru K, Xu W, Chang C-C et al (2023) Increased U.S. coastal hurricane risk under climate change. Sci Adv 9:eadf0259. https://​doi.​org/​10.​1126/​sciadv.​adf02​59 Barnett J, Lambert S, Fry I (2008) The hazards of indicators: insights from the Environmental Vulnerability Index. Ann Assoc Am Geogr 98:102–119. https://​doi.​org/​10.​1080/​00045​60070​17343​15 Bissolli P, Ganter C, Li T et al (2021) Regional climates. Bull Am Meteorol Soc 102:S357–S464. https://​ doi.​org/​10.​1175/​2021B​AMSSt​ateof​theCl​imate_​Chapt​er7.1 Blunden J, Arndt DS (2017) State of the climate in 2016. Bull Am Meteorol Soc 98:Si–S280. https://​doi.​ org/​10.​1175/​2017B​AMSSt​ateof​theCl​imate.1 Camargo SJ, Murakami H, Bloemendaal N et al (2023) An update on the influence of natural climate vari- ability and anthropogenic climate change on tropical cyclones. Trop Cyclone Res Rev 12:216–239. https://​doi.​org/​10.​1016/j.​tcrr.​2023.​10.​001 Cao X, Wu R, Jiang X et  al (2025) The southward shift of hurricane genesis over the northern Atlantic Ocean. NPJ Clim Atmos Sci 8:37. https://​doi.​org/​10.​1038/​s41612-​025-​00923-2 Castillo-Loeza DE, Romero D, Quesada-Román A, León-Cruz JF (2025) Impacto socioeconómico de los ciclones tropicales en la península de Yucatán de 1970 a 2022. In: Sociedades resilientes y la reduc- ción del riesgo de desastre, 1a ed. Universidad de Quintana Roo, Chetumal, Quintana Roo Chand SS, Walsh KJE, Camargo SJ et al (2022) Declining tropical cyclone frequency under global warm- ing. Nat Clim Chang 12:655–661. https://​doi.​org/​10.​1038/​s41558-​022-​01388-4 Dominguez C, Done JM, Bruyère CL (2021) Future Changes in Tropical Cyclone and Easterly Wave Char- acteristics over Tropical North America. Oceans 2:429–447. https://​doi.​org/​10.​3390/​ocean​s2020​024 Das T, Hidalgo HG, Pierce DW et  al (2009) Structure and detectability of trends in hydrological meas- ures over the Western United States. J Hydrometeorol 10:871–892. https://​doi.​org/​10.​1175/​2009J​ HM1095.1 Diro GT, Giorgi F, Fuentes-Franco R et al (2014) Tropical cyclones in a regional climate change projection with RegCM4 over the CORDEX Central America domain. Clim Change 125:79–94. https://​doi.​org/​ 10.​1007/​s10584-​014-​1155-7 Elsner JB (2006) Evidence in support of the climate change–Atlantic hurricane hypothesis. Geophys Res Lett 33:2006GL026869. https://​doi.​org/​10.​1029/​2006G​L0268​69 Elsner JB, Jagger TH (2013) Hurricane climatology: a modern statistical guide using R. Oxford University Press, Oxford Emanuel K (2003) Tropical cyclones. Annu Rev Earth Planet Sci 31:75–104. https://​doi.​org/​10.​1146/​annur​ ev.​earth.​31.​100901.​141259 Emanuel K (2006) Climate and tropical cyclone activity: a new model downscaling approach. J Clim 19:4797–4802. https://​doi.​org/​10.​1175/​JCLI3​908.1 Emanuel K (2021) Atlantic tropical cyclones downscaled from climate reanalyses show increasing activity over past 150 years. Nat Commun 12:7027. https://​doi.​org/​10.​1038/​s41467-​021-​27364-8 Emanuel KA (1987) The dependence of hurricane intensity on climate. Nature 326:483–485. https://​doi.​org/​ 10.​1038/​32648​3a0 Enfield DB, Alfaro EJ (1999) The dependence of Caribbean rainfall on the interaction of the Tropical Atlantic and Pacific oceans. J Climate 12:2093–2103. https://​doi.​org/​10.​1175/​1520-​0442(1999)​012%​ 3c2093:​TDOCRO%​3e2.0.​CO;2 Enfield DB, Mestas-Nuñez AM, Trimble PJ (2001) The Atlantic Multidecadal Oscillation and its relation to rainfall and river flows in the continental U.S. Geophys Res Lett 28:2077–2080. https://​doi.​org/​10.​ 1029/​2000G​L0127​45 Fuentes-Franco R, Coppola E, Giorgi F et al (2014) Assessment of RegCM4 simulated inter-annual vari- ability and daily-scale statistics of temperature and precipitation over Mexico. Clim Dyn 42:629–647. https://​doi.​org/​10.​1007/​s00382-​013-​1686-z https://doi.org/10.1073/pnas.1808979115 https://doi.org/10.1073/pnas.1808979115 https://doi.org/10.1016/j.agrformet.2023.109321 https://doi.org/10.1016/j.agrformet.2023.109321 https://doi.org/10.1017/9781009157896.002 https://doi.org/10.1175/JCLI-D-17-0106.1 https://doi.org/10.1175/JCLI-D-17-0106.1 https://doi.org/10.1126/sciadv.adf0259 https://doi.org/10.1080/00045600701734315 https://doi.org/10.1175/2021BAMSStateoftheClimate_Chapter7.1 https://doi.org/10.1175/2021BAMSStateoftheClimate_Chapter7.1 https://doi.org/10.1175/2017BAMSStateoftheClimate.1 https://doi.org/10.1175/2017BAMSStateoftheClimate.1 https://doi.org/10.1016/j.tcrr.2023.10.001 https://doi.org/10.1038/s41612-025-00923-2 https://doi.org/10.1038/s41558-022-01388-4 https://doi.org/10.3390/oceans2020024 https://doi.org/10.1175/2009JHM1095.1 https://doi.org/10.1175/2009JHM1095.1 https://doi.org/10.1007/s10584-014-1155-7 https://doi.org/10.1007/s10584-014-1155-7 https://doi.org/10.1029/2006GL026869 https://doi.org/10.1146/annurev.earth.31.100901.141259 https://doi.org/10.1146/annurev.earth.31.100901.141259 https://doi.org/10.1175/JCLI3908.1 https://doi.org/10.1038/s41467-021-27364-8 https://doi.org/10.1038/326483a0 https://doi.org/10.1038/326483a0 https://doi.org/10.1175/1520-0442(1999)012%3c2093:TDOCRO%3e2.0.CO;2 https://doi.org/10.1175/1520-0442(1999)012%3c2093:TDOCRO%3e2.0.CO;2 https://doi.org/10.1029/2000GL012745 https://doi.org/10.1029/2000GL012745 https://doi.org/10.1007/s00382-013-1686-z 15920 Natural Hazards (2025) 121:15899–15923 Fuentes-Franco R, Coppola E, Giorgi F et al (2015) Inter-annual variability of precipitation over Southern Mexico and Central America and its relationship to sea surface temperature from a set of future pro- jections from CMIP5 GCMs and RegCM4 CORDEX simulations. Clim Dyn 45:425–440. https://​doi.​ org/​10.​1007/​s00382-​014-​2258-6 Goldenberg SB, Landsea CW, Mestas-Nuñez AM, Gray WM (2001) The recent increase in Atlantic hur- ricane activity: causes and implications. Science 293:474–479. https://​doi.​org/​10.​1126/​scien​ce.​10600​ 40 Gray WM (1984) Atlantic seasonal hurricane frequency. part I: El Niño and 30 mb Quasi-Biennial Oscil- lation influences. Mon Weather Rev 112:1649–1668. https://​doi.​org/​10.​1175/​1520-​0493(1984)​112%​ 3c1649:​ASHFPI%​3e2.0.​CO;2 Hartfield G, Blunden J, Arndt DS (2018) State of the climate in 2017. Bull Am Meteorol Soc 99:Si–S310. https://​doi.​org/​10.​1175/​2018B​AMSSt​ateof​theCl​imate.1 Hersbach H, Bell B, Berrisford P et al (2020) The ERA5 global reanalysis. Q J R Meteorol Soc 146:1999– 2049. https://​doi.​org/​10.​1002/​qj.​3803 Hidalgo HG, Das T, Dettinger MD et  al (2009) Detection and attribution of streamflow timing changes to climate change in the Western United States. J Clim 22:3838–3855. https://​doi.​org/​10.​1175/​2009J​ CLI24​70.1 Hidalgo HG, Durán-Quesada AM, Amador JA, Alfaro EJ (2015) The Caribbean low-level jet, the inter- tropical convergence zone and precipitation patterns in the Intra-Americas sea: a proposed dynamical mechanism. Geogr Ann Ser A Phys Geogr 97:41–59. https://​doi.​org/​10.​1111/​geoa.​12085 Hidalgo HG, Alfaro EJ, Amador JA, Bastidas Á (2019) Precursors of quasi-decadal dry-spells in the Central America Dry Corridor. Clim Dyn 53:1307–1322. https://​doi.​org/​10.​1007/​s00382-​019-​04638-y Hidalgo HG, Alfaro EJ, HernÃandez-Castro F, Pérez-Briceño PM (2020) Identification of Tropical Cyclones’ Critical Positions Associated with Extreme Precipitation Events in Central America. Atmosphere 11:1123. https://​doi.​org/​10.​3390/​atmos​11101​123 Hidalgo HG, Alfaro EJ, Valverde KT, Bazo J (2023) Probability of induced extreme precipitation events in Central America due to tropical cyclone positions in the surrounding oceans. Nat Hazards 116:2917– 2933. https://​doi.​org/​10.​1007/​s11069-​022-​05790-1 Hidalgo HG, Chou-Chen SW, McKinnon KA et al (2025) Detection and attribution of trends of meteorolog- ical extremes in Central America. Clim Change 178:95. https://​doi.​org/​10.​1007/​s10584-​025-​03940-5 Hirahara S, Ishii M, Fukuda Y (2014) Centennial-scale sea surface temperature analysis and its uncertainty. J Clim 27:57–75. https://​doi.​org/​10.​1175/​JCLI-D-​12-​00837.1 Holland GJ (1997) The maximum potential intensity of tropical cyclones. J Atmos Sci 54:2519–2541. https://​doi.​org/​10.​1175/​1520-​0469(1997)​054%​3c2519:​TMPIOT%​3e2.0.​CO;2 Hyun S, Yeh S, Song S et al (2020) Understanding intermodel diversity when simulating the time of emer- gence in CMIP5 climate models. Geophys Res Lett 47:e2020GL087923. https://​doi.​org/​10.​1029/​ 2020G​L0879​23 Imada Y, Watanabe M, Kawase H et al (2019) The July 2018 high temperature event in Japan could not have happened without human-induced global warming. SOLA 15A:8–12. https://​doi.​org/​10.​2151/​ sola.​15A-​002 Intergovernmental Panel On Climate Change (IPCC) (2023) Climate change 2021—the physical science basis: working group I contribution to the sixth assessment report of the intergovernmental panel on climate change, 1st edn. Cambridge University Press Kawase H, Imada Y, Sasaki H et al (2019) Contribution of historical Global Warming to local-scale heavy precipitation in western Japan estimated by large ensemble high-resolution simulations. J Geophys Res Atmos 124:6093–6103. https://​doi.​org/​10.​1029/​2018J​D0301​55 Klotzbach PJ, Wood KM, Schreck CJ et al (2022) Trends in global tropical cyclone activity: 1990–2021. Geophys Res Lett 49:e2021GL095774. https://​doi.​org/​10.​1029/​2021G​L0957​74 Knutson TR, Tuleya RE (2004) Impact of CO2-induced warming on simulated hurricane intensity and precipitation: sensitivity to the choice of climate model and convective parameterization. J Climate 17:3477–3495. https://​doi.​org/​10.​1175/​1520-​0442(2004)​017%​3c3477:​IOCWOS%​3e2.0.​CO;2 Knutson TR, Tuleya RE, Kurihara Y (1998) Simulated increase of hurricane intensities in a CO2-warmed climate. Science 279:1018–1021. https://​doi.​org/​10.​1126/​scien​ce.​279.​5353.​1018 Knutson TR, McBride JL, Chan J et al (2010) Tropical cyclones and climate change. Nat Geosci 3:157–163. https://​doi.​org/​10.​1038/​ngeo7​79 Knutson T, Camargo SJ, Chan JCL et  al (2019) Tropical cyclones and climate change assessment: part I: detection and attribution. Bull Am Meterol Soc 100:1987–2007. https://​doi.​org/​10.​1175/​ BAMS-D-​18-​0189.1 Kobayashi S, Ota Y, Harada Y et al (2015) The JRA-55 reanalysis: general specifications and basic charac- teristics. J Meteorol Soc Jpn 93:5–48. https://​doi.​org/​10.​2151/​jmsj.​2015-​001 https://doi.org/10.1007/s00382-014-2258-6 https://doi.org/10.1007/s00382-014-2258-6 https://doi.org/10.1126/science.1060040 https://doi.org/10.1126/science.1060040 https://doi.org/10.1175/1520-0493(1984)112%3c1649:ASHFPI%3e2.0.CO;2 https://doi.org/10.1175/1520-0493(1984)112%3c1649:ASHFPI%3e2.0.CO;2 https://doi.org/10.1175/2018BAMSStateoftheClimate.1 https://doi.org/10.1002/qj.3803 https://doi.org/10.1175/2009JCLI2470.1 https://doi.org/10.1175/2009JCLI2470.1 https://doi.org/10.1111/geoa.12085 https://doi.org/10.1007/s00382-019-04638-y https://doi.org/10.3390/atmos11101123 https://doi.org/10.1007/s11069-022-05790-1 https://doi.org/10.1007/s10584-025-03940-5 https://doi.org/10.1175/JCLI-D-12-00837.1 https://doi.org/10.1175/1520-0469(1997)054%3c2519:TMPIOT%3e2.0.CO;2 https://doi.org/10.1029/2020GL087923 https://doi.org/10.1029/2020GL087923 https://doi.org/10.2151/sola.15A-002 https://doi.org/10.2151/sola.15A-002 https://doi.org/10.1029/2018JD030155 https://doi.org/10.1029/2021GL095774 https://doi.org/10.1175/1520-0442(2004)017%3c3477:IOCWOS%3e2.0.CO;2 https://doi.org/10.1126/science.279.5353.1018 https://doi.org/10.1038/ngeo779 https://doi.org/10.1175/BAMS-D-18-0189.1 https://doi.org/10.1175/BAMS-D-18-0189.1 https://doi.org/10.2151/jmsj.2015-001 15921Natural Hazards (2025) 121:15899–15923 Kossin JP (2018) A global slowdown of tropical-cyclone translation speed. Nature 558:104–107. https://​doi.​ org/​10.​1038/​s41586-​018-​0158-3 Kossin JP, Emanuel KA, Vecchi GA (2014) The poleward migration of the location of tropical cyclone maximum intensity. Nature 509:349–352. https://​doi.​org/​10.​1038/​natur​e13278 Kossin JP, Knapp KR, Olander TL, Velden CS (2020) Global increase in major tropical cyclone exceedance probability over the past four decades. Proc Natl Acad Sci USA 117:11975–11980. https://​doi.​org/​10.​ 1073/​pnas.​19208​49117 Landsea CW, Franklin JL (2013) Atlantic hurricane database uncertainty and presentation of a new database format. Mon Weather Rev 141:3576–3592. https://​doi.​org/​10.​1175/​MWR-D-​12-​00254.1 Landsea CW, Vecchi GA, Bengtsson L, Knutson TR (2008) Impact of duration thresholds on Atlantic tropi- cal cyclone counts. J Clim 23:2508–2519. https://​doi.​org/​10.​1175/​2009J​CLI30​34.1 Lange S (2019) Trend-preserving bias adjustment and statistical downscaling with ISIMIP3BASD (v1.0). Geosci Model Dev 12:3055–3070. https://​doi.​org/​10.​5194/​gmd-​12-​3055-​2019 Lange S (2022) ISIMIP3BASD. Version 2.5.2. Zenodo. URL https://​zenodo.​org/​recor​ds/​63449​11 Loehle C, Staehling E (2020) Hurricane trend detection. Nat Hazards 104:1345–1357. https://​doi.​org/​10.​ 1007/​s11069-​020-​04219-x Lopez H, Lee S-K, West R et al (2024) Projected increase in the frequency of extremely active Atlantic hurricane seasons. Sci Adv 10:eadq7856. https://​doi.​org/​10.​1126/​sciadv.​adq78​56 Lozano-Gracia N, Piras G, Ibáñez AM, Hewings GJD (2009) The Journey to Safety: Conflict-Driven Migra- tion Flows in Colombia. Int Reg Sci Rev 33:157–180. https://​doi.​org/​10.​1177/​01600​17609​336998 Lucas C, Timbal B, Nguyen H (2014) The expanding tropics: a critical assessment of the observational and modeling studies. Wires Clim Change 5:89–112. https://​doi.​org/​10.​1002/​wcc.​251 Luo B, Luo D, Zhuo W et al (2023) Increased summer European heatwaves in recent decades: contribu- tions from greenhouse gases-induced warming and Atlantic Multidecadal Oscillation-like varia- tions. Earths Future 11:e2023EF003701. https://​doi.​org/​10.​1029/​2023E​F0037​01 Maldonado T, Amador JA, Rivera ER et al (2020) Examination of WRF-ARW experiments using dif- ferent planetary boundary layer parameterizations to study the rapid intensification and trajectory of Hurricane Otto (2016). Atmosphere (Basel) 11:1317. https://​doi.​org/​10.​3390/​atmos​11121​317 Martinez L-C, Romero D, Alfaro EJ (2023) Assessment of the spatial variation in the occurrence and intensity of major hurricanes in the western hemisphere. Climate 11(1):15. https://​doi.​org/​10.​ 3390/​cli11​010015 Meiler S, Kropf CM, McCaughey JW et al (2025) Navigating and attributing uncertainty in future tropi- cal cyclone risk estimates. Sci Adv 11:eadn4607. https://​doi.​org/​10.​1126/​sciadv.​adn46​07 Mengel M, Treu S, Lange S, Frieler K (2021) ATTRICI v1.1-counterfactual climate for impact attribu- tion. Geosci Model Dev 14:5269–5284. https://​doi.​org/​10.​5194/​gmd-​14-​5269-​2021 Mizuta R, Yoshimura H, Murakami H et  al (2012) Climate simulations using MRI-AGCM3.2 with 20-km grid. J Meteorol Soc Jpn 90A:233–258. https://​doi.​org/​10.​2151/​jmsj.​2012-​A12 Mizuta R, Murata A, Ishii M et al (2017) Over 5000 years of ensemble future climate simulations by 60-km global and 20-km regional atmospheric models. Bull Am Meterol Soc 98:1383–1398. https://​doi.​org/​10.​1175/​BAMS-D-​16-​0099.1 Moon I-J, Kim S-H, Klotzbach P, Chan JCL (2015) Roles of interbasin frequency changes in the pole- ward shifts of the maximum intensity location of tropical cyclones. Environ Res Lett 10:104004. https://​doi.​org/​10.​1088/​1748-​9326/​10/​10/​104004 Moon I-J, Kim S-H, Chan JCL (2019) Climate change and tropical cyclone trend. Nature 570:E3–E5. https://​doi.​org/​10.​1038/​s41586-​019-​1222-3 Murakami H, Mizuta R, Shindo E (2012) Future changes in tropical cyclone activity projected by multi- physics and multi-SST ensemble experiments using the 60-km-mesh MRI-AGCM. Clim Dyn 39:2569–2584. https://​doi.​org/​10.​1007/​s00382-​011-​1223-x Murakami H, Levin E, Delworth TL et al (2018) Dominant effect of relative tropical Atlantic warming on major hurricane occurrence. Science 362:794–799. https://​doi.​org/​10.​1126/​scien​ce.​aat67​11 Murakami H, Delworth TL, Cooke WF et al (2020) Detected climatic change in global distribution of tropical cyclones. Proc Natl Acad Sci USA 117:10706–10714. https://​doi.​org/​10.​1073/​pnas.​19225​ 00117 Oouchi K, Yoshimura J, Yoshimura H et  al (2006) Tropical cyclone climatology in a global-warming climate as simulated in a 20 km-mesh global atmospheric model: frequency and wind intensity analyses. J Meteorol Soc Jpn 84:259–276. https://​doi.​org/​10.​2151/​jmsj.​84.​259 Papin PP, Bosart LF, Torn RD (2017) A climatology of Central American Gyres. Mon Weather Rev 145:1983–2000. https://​doi.​org/​10.​1175/​MWR-D-​16-​0411.1 Pascale S, Kapnick SB, Delworth TL et al (2021) Natural variability vs forced signal in the 2015–2019 Central American drought. Clim Change 168:16. https://​doi.​org/​10.​1007/​s10584-​021-​03228-4 https://doi.org/10.1038/s41586-018-0158-3 https://doi.org/10.1038/s41586-018-0158-3 https://doi.org/10.1038/nature13278 https://doi.org/10.1073/pnas.1920849117 https://doi.org/10.1073/pnas.1920849117 https://doi.org/10.1175/MWR-D-12-00254.1 https://doi.org/10.1175/2009JCLI3034.1 https://doi.org/10.5194/gmd-12-3055-2019 https://zenodo.org/records/6344911 https://doi.org/10.1007/s11069-020-04219-x https://doi.org/10.1007/s11069-020-04219-x https://doi.org/10.1126/sciadv.adq7856 https://doi.org/10.1177/0160017609336998 https://doi.org/10.1002/wcc.251 https://doi.org/10.1029/2023EF003701 https://doi.org/10.3390/atmos11121317 https://doi.org/10.3390/cli11010015 https://doi.org/10.3390/cli11010015 https://doi.org/10.1126/sciadv.adn4607 https://doi.org/10.5194/gmd-14-5269-2021 https://doi.org/10.2151/jmsj.2012-A12 https://doi.org/10.1175/BAMS-D-16-0099.1 https://doi.org/10.1088/1748-9326/10/10/104004 https://doi.org/10.1038/s41586-019-1222-3 https://doi.org/10.1007/s00382-011-1223-x https://doi.org/10.1126/science.aat6711 https://doi.org/10.1073/pnas.1922500117 https://doi.org/10.1073/pnas.1922500117 https://doi.org/10.2151/jmsj.84.259 https://doi.org/10.1175/MWR-D-16-0411.1 https://doi.org/10.1007/s10584-021-03228-4 15922 Natural Hazards (2025) 121:15899–15923 Pérez Briceño PM, Alfaro Martínez EJ, Hidalgo León HG, Jiménez F (2016) Distribución espacial de impactos de eventos hidrometeorológicos en América Central. Rev Climatol 16:63–75 Pfleiderer P, Nath S, Schleussner C-F (2022) Extreme Atlantic hurricane seasons made twice as likely by ocean warming. Weather Clim Dyn 3:471–482. https://​doi.​org/​10.​5194/​wcd-3-​471-​2022 Piani C, Weedon GP, Best M et al (2010) Statistical bias correction of global simulated daily precipita- tion and temperature for the application of hydrological models. J Hydrol 395:199–215. https://​ doi.​org/​10.​1016/j.​jhydr​ol.​2010.​10.​024 Pierce DW, Barnett TP, Hidalgo HG et  al (2008) Attribution of declining Western US snowpack to human effects. J Clim 21:6425–6444. https://​doi.​org/​10.​1175/​2008J​CLI24​05.1 Quesada Román A (2021) Landslides and floods zonation using geomorphological analyses in a dynamic basin of Costa Rica. Rev Cart 1:125–138. https://​doi.​org/​10.​35424/​rcarto.​i102.​901 Quesada-Román A, Hidalgo HG, Alfaro EJ (2024) Assessing the impact of tropical cyclones on eco- nomic sectors in Costa Rica, Central America. Trop Cyclone Res Rev 13:196–207. https://​doi.​org/​ 10.​1016/j.​tcrr.​2024.​08.​001 Reed KA, Wehner MF, Zarzycki CM (2022) Attribution of 2020 hurricane season extreme rain- fall to human-induced climate change. Nat Commun 13:1905. https://​doi.​org/​10.​1038/​ s41467-​022-​29379-1 Romero D, León-Cruz JF (2024) Spatiotemporal changes in hurricane-force wind risk assess- ment in the Yucatan Peninsula, Mexico. Nat Hazards 120:4675–4698. https://​doi.​org/​10.​1007/​ s11069-​023-​06397-w Romero D, Appendini CM, Emanuel K et al (2025) Assessment of synthetic tropical cyclones in the North Atlantic basin. Preprint. https://​doi.​org/​10.​2139/​ssrn.​51229​58 Sharmila S, Walsh KJE (2018) Recent poleward shift of tropical cyclone formation linked to Hadley cell expansion. Nat Clim Change 8:730–736. https://​doi.​org/​10.​1038/​s41558-​018-​0227-5 Skamarock W, Klemp J, Dudhia J et  al (2008) A description of the advanced research WRF version 3. UCAR/NCAR​ Shiogama H, Imada Y, Mori M et al (2016) Attributing historical changes in probabilities of record-break- ing daily temperature and precipitation extreme events. SOLA 12:225–231. https://​doi.​org/​10.​2151/​ sola.​2016-​045 Studholme J, Gulev S (2018) Concurrent changes to Hadley Circulation and the meridional distribution of tropical cyclones. J Clim 31:4367–4389. https://​doi.​org/​10.​1175/​JCLI-D-​17-​0852.1 Studholme J, Fedorov AV, Gulev SK et al (2022) Poleward expansion of tropical cyclone latitudes in warm- ing climates. Nat Geosci 15:14–28. https://​doi.​org/​10.​1038/​s41561-​021-​00859-1 Sugi M, Murakami H, Yoshida K (2017) Projection of future changes in the frequency of intense tropical cyclones. Clim Dyn 49:619–632. https://​doi.​org/​10.​1007/​s00382-​016-​3361-7 Takaya Y, Yasuda T, Ose T, Nakaegawa T (2010) Predictability of the mean location of typhoon formation in a seasonal prediction experiment with a coupled General Circulation Model. J Meteorol Soc Jpn 88:799–812. https://​doi.​org/​10.​2151/​jmsj.​2010-​502 Taylor KE (2001) Summarizing multiple aspects of model performance in a single diagram. J Geophys Res 106:7183–7192. https://​doi.​org/​10.​1029/​2000J​D9007​19 United Nations Office for Disaster Risk Reduction (UNDRR) (2018) Disaster risk reduction terminology. Ginebra, Suiza Vecchi GA, Soden BJ (2007) Increased tropical Atlantic wind shear in model projections of global warm- ing. Geophys Res Lett 34:2006GL028905. https://​doi.​org/​10.​1029/​2006G​L0289​05 Vecchi GA, Landsea C, Zhang W et al (2021) Changes in Atlantic major hurricane frequency since the late- 19th century. Nat Commun 12:4054. https://​doi.​org/​10.​1038/​s41467-​021-​24268-5 Walsh KJE, McBride JL, Klotzbach PJ et  al (2016) Tropical cyclones and climate change. Wires Clim Change 7:65–89. https://​doi.​org/​10.​1002/​wcc.​371 Wang S, Toumi R (2021) Recent migration of tropical cyclones toward coasts. Science 371:514–517. https://​doi.​org/​10.​1126/​scien​ce.​abb90​38 Wehner M (2021) Simulated changes in tropical cyclone size, accumulated cyclone energy and power dis- sipation index in a warmer climate. Oceans 2:688–699. https://​doi.​org/​10.​3390/​ocean​s2040​039 Wehner MF, Bala G, Duffy P et al (2010) Towards direct simulation of future tropical cyclone statistics in a high-resolution global atmospheric model. Adv Meteorol 2010:915303. https://​doi.​org/​10.​1155/​2010/​ 915303 World Bank (2022) Global economic prospects, June 2022. The World Bank Ying J, Collins M, Cai W et al (2022) Emergence of climate change in the tropical Pacific. Nat Clim Chang 12:356–364. https://​doi.​org/​10.​1038/​s41558-​022-​01301-z https://doi.org/10.5194/wcd-3-471-2022 https://doi.org/10.1016/j.jhydrol.2010.10.024 https://doi.org/10.1016/j.jhydrol.2010.10.024 https://doi.org/10.1175/2008JCLI2405.1 https://doi.org/10.35424/rcarto.i102.901 https://doi.org/10.1016/j.tcrr.2024.08.001 https://doi.org/10.1016/j.tcrr.2024.08.001 https://doi.org/10.1038/s41467-022-29379-1 https://doi.org/10.1038/s41467-022-29379-1 https://doi.org/10.1007/s11069-023-06397-w https://doi.org/10.1007/s11069-023-06397-w https://doi.org/10.2139/ssrn.5122958 https://doi.org/10.1038/s41558-018-0227-5 https://doi.org/10.2151/sola.2016-045 https://doi.org/10.2151/sola.2016-045 https://doi.org/10.1175/JCLI-D-17-0852.1 https://doi.org/10.1038/s41561-021-00859-1 https://doi.org/10.1007/s00382-016-3361-7 https://doi.org/10.2151/jmsj.2010-502 https://doi.org/10.1029/2000JD900719 https://doi.org/10.1029/2006GL028905 https://doi.org/10.1038/s41467-021-24268-5 https://doi.org/10.1002/wcc.371 https://doi.org/10.1126/science.abb9038 https://doi.org/10.3390/oceans2040039 https://doi.org/10.1155/2010/915303 https://doi.org/10.1155/2010/915303 https://doi.org/10.1038/s41558-022-01301-z 15923Natural Hazards (2025) 121:15899–15923 Zhang C, Wang Y (2018) Why is the simulated climatology of tropical cyclones so sensitive to the choice of cumulus parameterization scheme in the WRF model? Clim Dyn 51:3613–3633. https://​doi.​org/​10.​ 1007/​s00382-​018-​4099-1 Publisher’s Note  Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. https://doi.org/10.1007/s00382-018-4099-1 https://doi.org/10.1007/s00382-018-4099-1 How rare was the 2016–2022 tropical cyclone activity near the Caribbean coasts of Nicaragua and Costa Rica? Abstract 1 Introduction 2 Data 2.1 Observed data 2.2 Inter-sectoral impact model intercomparison project round 3 model data 2.3 Large model ensemble TCs trajectories from Japan’s Meteorological Research Institute 2.4 Reanalysis data 3 Methodology 4 Results 4.1 Evaluation of TC trajectories in ISIMIP3 models 4.2 Monte Carlo simulations with the HURDAT2 data 4.3 Bias correction 4.4 Evaluation of climate variables in ISIMIP3 models 5 Discussion and conclusions Acknowledgements References