The effects of projected climate change on crop water availability in the U.S. Caribbean

Anthropogenic climate change particularly affects the small island developing states (SIDS) (Karnauskas et al. 2018). The rainfall variability, and the effects of low-frequency atmospheric circulation patterns on rainfall and drought events, as well as the lack of freshwater resources in the insular Caribbean, point to its vulnerability to climate change (Giannini et al. 2001; Karnauskas et al. 2018; Ault 2020; Robinson & Wren 2020). Accordingly, a predicted increase in temperature, decreased length of the rainy season, increased length of the dry season, more intense rainstorms, and increase in sea level for the Caribbean could result in the reduction of the already scarce water availability due to the increase of evapotranspiration rates, more flooding and aquifer depletion (reduced recharge), and salinity intrusion into groundwater and coastal aquifers (Cashman et al. 2010; Farrell et al. 2010; Pulwarty et al. 2010; Ault 2016; Karnauskas et al. 2016).

Because agricultural activities depend on water, this sector will be directly impacted by future drying trends, with smallholder farmers, in particular, being the most dependent upon the amount and timing of annual rainfall (Bates et al. 2008; Curtis et al. 2014). The Caribbean is already facing poverty and food insecurity together with limited land availability, which means that global climate change will further exacerbate the challenges in the agricultural sector (Trotman et al. 2009; Connell et al. 2020). Studies have found farming systems in the Caribbean to be vulnerable to projected climate change, especially changes in temperature and precipitation, due to their relatively high dependence on rainfall (Bates et al. 2008; Cashman et al. 2010; Curtis et al. 2014). Impacts of a drier and warmer Caribbean climate can include but are not limited to reduction in plant-available moisture due to increased rates of evapotranspiration, increased spread of some pests and diseases, a decrease in crop suitability, and increased stress on food productivity and sustainability (McGregor et al. 2009; Cashman et al. 2010; Curtis et al. 2014).

However, much of the research on climate change and agriculture has been more general and of regional scope (Farrell et al. 2007; Trotman et al. 2009; Pulwarty et al. 2010; Barker 2012; Karnauskas et al. 2018; Ault 2020), so there is a clear need for more studies focusing on specific islands within the Caribbean. Additionally, to our knowledge, there is no research comparing the impacts of long-term climate change on agriculture on an island in the Greater and Lesser Antilles.

Therefore, this work used a crop risk model (CROPRISK) (Batjes 1987) to analyze how water deficit and crop water stress in Puerto Rico (Greater Antilles) and the U.S. Virgin Island of St. Croix (Lesser Antilles) can be affected by future temperature and precipitation changes by the mid-century (2041–2070) and late-century (2071–2100). For this analysis, we choose to focus on the crops sweet pepper, banana, and plantain due to their agricultural importance to Puerto Rico and St. Croix, as well as their sensitivity to water deficits.

In order to evaluate how temperature and precipitation projections for mid-century (2041–2070) and late-century (2071–2100) can affect water stress and crop suitability in Puerto Rico and St. Croix, a multi-step methodology was developed, which included creating a baseline climatology and two future climatologies, using the Thornthwaite water budget model to estimate potential and actual evapotranspiration (AET) used to calculate water deficits, as well as using the CROPRISK model to assess crop suitability in current and future climates. These steps are explained below.

For future climates, we used the Coupled Model Intercomparison Project Phase 6 (CMIP6) and selected the five models based on their common and smaller grid cell size (100 km), since Puerto Rico's and St. Croix's areas are small to work with the coarser spatial resolution of the other models. Models included in this analysis were downloaded from https://esgf-node.llnl.gov/search/cmip6/ and are found in Table 1. The goal of this work was to compare the baseline climate (1981–2010) to mid-century (2041–2070) and late-century (2071–2100) climatologies, to analyze long-term future changes in climate, and not near-future changes. Therefore, we downloaded and selected both historical (1981–2010) and future (2041–2100) monthly precipitation and surface temperature data from each model. For future data, we downloaded the shared socioeconomic pathway-representative concentration pathway (SSP-RCP) scenario SSP5-8.5, which is an updated version of RCP8.5 in CMIP5 used in Curtis et al. (2014) in a similar study, and refers to the scenario with higher CO₂ emissions by the end of the century (O'Neill et al. 2016). Using the same SSP-RCP scenario from Curtis et al. (2014) allows us to compare the results of this work with another work also applied to a Caribbean island country.

Then, the data were remapped to account for the different grid systems used by the models, even though they have the same resolution. The remapping used the nearest-neighbor remapping function (remapnn) from the climate data operators (CDO) to select the grid cell corresponding to the centroid of each basin in Puerto Rico and of the island of St. Croix (Figure 1). We used the nearest-neighbor approach since other studies have indicated that this approach does not smooth the extremes of the models (Wang et al. 2019). Then, we created the 30-year monthly mean surface temperature and precipitation for historical and future climates.

After processing the climate data, we performed the sensitivity analysis of water availability in Puerto Rico and St. Croix, followed by the analysis of agroclimatic suitability for specific crops in the islands. For water availability analysis, we used the Thornthwaite Monthly Water Balance Model as developed by the USGS (McCabe & Markstrom 2007), available at https://www.usgs.gov/software/thornthwaite-monthly-water-balance-model. This model was the same used by Curtis et al. (2014) in their study focusing on crop water stress in Jamaica and is designed to work with the tabular monthly climate data used in this study. The input parameters for the model were the same as those used by Curtis et al. (2014) in their analysis of Jamaica. This model is driven by a graphical user interface in which some parameters can be specified, and the input data are monthly temperature and precipitation from a specific location. The model analyzes the components of the hydrologic cycle according to Thornthwaite (Thornthwaite 1948; McCabe & Markstrom 2007). We ran the model for each one of the four basins of Puerto Rico and the island of St. Croix using the following parameters for the five locations: 18° N as the latitude parameter; 5% as the fraction of precipitation that becomes direct runoff from infiltration-excess overflow; 150 mm as the soil-moisture storage capacity, and the runoff generation as 50% of the surplus water produced after the soil-moisture storage surpasses its capacity (Wolock & McCabe 1999; McCabe & Markstrom 2007; Curtis et al. 2014).

Finally, we used the deficit from maximum crop yield to establish the agroclimatic suitability classes defined by Batjes (1987). These classes depend on the number of years in which at least 80 and 60% of the crop's maximum yield can be obtained, as follows:

• Highly suitable (HiS): when the 80% condition of a crop's maximum yield is met at least 60% of the years and the 60% condition is met at least 80% of the years.

• Moderately suitable (MoS): when the 80% condition of the crop's maximum yield is met for at least 40% of the years and the 60% condition is met for at least 60% of the years.

• Marginally suitable (MaS): when the 80% condition of the crop's maximum yield is met for at least 20% of the years and the 60% condition is met for at least 40% of the years.

• Not suitable (NS): when the 80% condition of a crop's maximum yield is met in less than 20% of the years and the 60% condition is met in less than 40% of the years.

Both crop water stress (PET – AET) and agroclimatic suitability classes were determined on the annual cycle for the baseline climatology (1981–2010), the future climatology representing the mid-century (2041–2070), and the one representing the late-century (2071–2100) for each one of the four basins in Puerto Rico and the island of St. Croix. Through the analysis of these future climatologies, we expect to evaluate the role of temperature and precipitation changes in future agricultural suitability and water stress in both the Greater and Lesser Antilles.

In Puerto Rico, the most recent census of agriculture from the U.S. Department of Agriculture (USDA 2020a) indicated that the number of farms has decreased from 2012 to 2018, and the number of cropland harvested farms decreased from 10,008 to 4,888, representing a reduction of more than 17,806 ha. The irrigated land has also decreased its area since 2012 by at least 10,360 ha, which leaves more than 20,000 ha of harvested cropland mostly dependent on rainfall.

Puerto Rico's major crops include plantains and bananas, which were cultivated in 2,035 and 1,157 farms, respectively, in 2018 (USDA 2020a). Although the number of farms cultivating plantains and bananas has decreased since 2012, their economic value is still among the highest, with an average of the market value of products sold in 2018 equal to $31,243 per farm for plantain, and $13,521 per farm for bananas (USDA 2020a). More farms grow peppers than any other vegetable in Puerto Rico, with sweet peppers cultivated at 290 farms in 2018, and other types of peppers cultivated in 62 farms. These numbers also decreased when compared to the 2012 agricultural census, when peppers were cultivated in 603 farms (USDA 2020a). Sweet peppers also lead the vegetables with the majority number of farms (>100 farms) with a market value of agricultural products sold of >$60,000.

On the contrary, St. Croix has quintupled its area of cropland in 10 years, from 106 farms (161 ha) in 2007 to 336 farms (851 ha) in 2018. Vegetables represented the largest category of production with sales of $1.1 million (USDA 2020b). Notably, an increase in land irrigated also occurred, from 81 ha in 2007 to 223 ha in 2018 (USDA 2020b). Despite this increase, there are still at least 600 ha of cropland dependent on rainfall.

Among the vegetables produced in St. Croix, peppers also lead the number of farms as the primary crop with a total of 121 farms in 2018, within which more than 20 farms have a market value of agricultural products sold of >$10,000. Plantains and bananas are also some of the main fruits cultivated in St. Croix, based on the number of farms. In 2018, 119 farms cultivated plantains and 220 farms cultivated bananas, compared to only 33 farms cultivating plantain and 57 farms cultivating bananas in 2007. The market value of the agricultural products sold was >$10,000 for at least 9 plantain and 24 banana farms in St. Croix (USDA 2020b).

On top of the economic value, these crops also have an important cultural value for the islands. For the Puerto Rican community that lives either in Puerto Rico or in St. Croix, a particular sweet pepper, locally known as Ají dulce (Capsicum chinense), is popular and used in culinary seasoning, giving characteristic flavor to most Puerto Rican recipes (Palada et al. 2003). The historical significance of banana and plantain for the small islands in the Caribbean also make those crops relevant for our analysis. It is estimated that the production of bananas is one of the largest employers of labor and sustains thousands of small farmers in the Caribbean, providing a foundation for the economic viability and social and political stability in the region (Bernal 2020).

Besides the economic and cultural importance of banana, plantain, and sweet pepper, these crops were widely cultivated across Puerto Rico in 2016, according to the agricultural statistics interactive platform created by the USDA Caribbean Climate Hub (https://caribbeanclimatehub.org/tools-apps/agricultural statistics/). In 2016, sweet pepper, banana, and plantain were cultivated in at least 148, 156, and 430 neighborhoods, respectively, in Puerto Rico. Although we understand the importance of coffee plantations for Puerto Rico, we decided to not include coffee in this analysis because it is not cultivated in St. Croix and a recent study has examined the impact of climate change on coffee in Puerto Rico (Fain et al. 2018). Therefore, we chose to use sweet pepper, banana, and plantain as the crops to analyze the future agroclimatic suitability in Puerto Rico and St. Croix.

The values of kc and ky used for the CROPRISK model were related to each crop. Sweet pepper kc during its mid-growth stage is 1.05 and during its late-growth stage is 0.90 (Allen et al. 1998; Harmsen et al. 2003; Kisekka et al. 2010). Here we followed the approach of Curtis et al. (2014) and chose a value of 1.0 to represent the harvest time of sweet pepper and avoid negative DYs, which are possible with smaller kc values. For sweet pepper ky, we used a value of 1.10 (Batjes 1987). Because bananas and plantains are a member of the same crop group, we used the same values of kc and ky (Allen et al. 1998): kc = 1.10, based on the average kc for mid-and late-growth stages for the first and second years of cultivation, and ky = 1.27. Water stress increases on a crop with ky values greater than one (Curtis et al. 2014), which suggests that sweet pepper, banana, and plantain are all crops with increased sensitivity to water availability.

After downloading and processing the climate observations (Daymet) and CMIP6, this study analyzed how temperature and precipitation of mid-century (2041–2070) and late-century (2071–2100) climatologies will change when compared to the baseline climatology (1981–2010). Figure 2 presents the 24-month annual cycle of temperature (°C) for all the observed and projected climatologies for the four basins in Puerto Rico and St. Croix. For observed data (Daymet), the higher monthly average temperature occurred in July–August, while for simulated data (CMIP6), the higher temperatures occurred in September–October for both the historical and future climatologies. However, the future climatologies, here called ‘Delta’ to indicate they were bias-corrected projections, also peaked in July–August, but with temperatures ranging from 1.8 to 2.2 °C warmer in 2041–2070 than the baseline climatology in Puerto Rico and St. Croix. For 2071–2100, differences in temperatures ranged from 3.2 to 3.7 °C warmer in Puerto Rico and St. Croix when compared to the baseline climatology.

Among the basins in Puerto Rico, the East basin registered a slightly smaller difference between the observed temperature and the two future climatologies in July–August than the rest of the island. This smaller increase in future temperature in eastern Puerto Rico could be related to the presence of the El Yunque National Forest, a tropical rainforest located in the Luquillo Mountains with an area of around 100 km². The effect of tropical rainforests in controlling temperature is mainly related to their high evapotranspiration demands, which use the available latent heat, and consequent cloud formation, which decreases incoming solar radiation at the surface (Fetcher et al. 1985; Brovkin 2002; Lawrence & Vandecar 2015). On the other hand, the fact that St. Croix registered the smallest change in temperature between observed and future climatologies, when compared to Puerto Rico Basins, could be related to the maritime effect being more efficient in the smaller island climate, resulting in smaller temperature ranges (Granger 1985).

The observed (Daymet) 24-month annual cycle of precipitation (Figure 3) confirmed the bimodal cycle of rainfall in the eastern Caribbean, with two rainy seasons occurring in April–July (early rainfall season) and August–November (late rainfall season), separated by a drier period in June–July (mid-summer dry spell), and a dry season occurring in December–March (Taylor et al. 2002; Curtis & Gamble 2008; Angeles et al. 2010). The mid-summer dry spell is less apparent in the East Puerto Rico Basin (Figure 3(a)), probably because this region has local mechanisms of rainfall along the year, such as the orographic effect and the presence of the tropical rainforest (Sobel et al. 2011; Ramseyer & Mote 2016; Jury 2020), that make it wetter than the rest of the island.

Before bias correction was applied, the CMIP6 models did not represent the bimodal cycle in precipitation. They could not simulate the December–March dry season and the April–July rainy season, while they overestimated the August–November rainy season in Puerto Rico and in St. Croix. However, after bias correction was applied, the future climatologies are projected to have the same pattern as the observed precipitation. The 2041–2070 precipitation was predicted to have from 54 to 74 mm less than the baseline climatology, mainly from June to October, in the four basins of Puerto Rico, and 36 mm less in St. Croix over the same period. The greater difference in precipitation is expected to occur between the baseline period and 2071–2100, when the West Puerto Rico Basin is projected to have 215 mm less precipitation from June to October, and St. Croix 96 mm less precipitation. The only month when projected precipitation was expected to be greater than the observed precipitation was November, which can indicate a future shift in the peak of the hurricane season that currently occurs in September (Kossin 2008; Martinez et al. 2019).

The overall increase in temperature and decrease in precipitation by mid-to-late-century found here was also observed in past studies that used earlier versions of CMIP6 (CMIP5 and CMIP3) to assess crop suitability in Jamaica and Puerto Rico (Curtis et al. 2014; Fain et al. 2018). Although uncertainties still exist, mainly regarding the precipitation projections, those findings using different models and different bias correction methods all pointed to a likely warmer and drier insular Caribbean starting in 2041 if the worst-case CO₂ emission scenario in CMIP6 is not avoided.

The Thornthwaite water budget model was applied to each one of the four basins of Puerto Rico and St. Croix during the baseline climatology (1981–2010) and the two future climatologies representing the mid-century (2041–2070) and late-century (2071–2100). Using PET and AET outputs from the model, we computed water deficits and crop suitability for each basin and St. Croix under those three climatologies.

For the future climates, the overall crop water stress was also highest from June to August, except for the West Puerto Rico Basin where the higher values occurred in January–March. This means that even wet regions such as the East and North Puerto Rico basins, which are not currently facing water stress during the mid-summer dry spell, will start to have problems with water available for crops during these drier months after 2041. In the South Puerto Rico Basin and St. Croix, where crop water stress is already occurring in the first half of the year, the late-century climatology suggested as much as a double the current water stress during the mid-summer dry spell in St. Croix and almost three times the water stress in southern Puerto Rico. These climate projections indicate that crop risk in those areas will be high after 2041. A similar result was found by Curtis et al. (2014) when analyzing the crop water stress for Jamaica by the end of the 21st century, indicating the mid-summer dry spell as a key component for future regional water stress. They suggest that one of the causes for the mid-summer water stress increase could be associated with the Caribbean low-level jet (CLLJ), which has its relative maximum in July (Gamble & Curtis 2008; Curtis et al. 2014). The Providing REgional Climates for Impacts Studies regional model simulated the patterns of the CLLJ by the end of the 21st century and suggested that an intensification of the CLLJ will occur from May to November and reach Jamaica, Hispaniola, Puerto Rico, and the north coast of South America resulting in a drier climate for the Caribbean (Taylor et al. 2013). The CLLJ together with the North Atlantic subtropical high (NAHP) is an important atmospheric pattern responsible for the current lower values of precipitation during the boreal summer in the Caribbean, transporting moisture from the Caribbean to Central America (Gamble & Curtis 2008; Gamble et al. 2008; Cook & Vizy 2010). The period when CLLJ is projected to be more intense (Taylor et al. 2013) aligned with the months for which the future climatologies here projected less precipitation and a higher average temperature (more evapotranspiration) for all the four basins of Puerto Rico and St. Croix (Figure 3). Thus, the changes of these atmospheric variables with climate change can justify the drier mid-summer dry spell and the increase in water stress across those islands in the future.

After analyzing crop water stress, we ran the CROPRISK model to assess agroclimatic suitability for sweet pepper, banana, and plantain in the four basins of Puerto Rico and St. Croix. We ran the model for each crop and each region over the three periods of analysis. Tables 2 and 3 show the suitability classes for each month. The overall results indicate that conditions for banana and plantain cultivation will be less suitable than those for sweet pepper in the South and West Puerto Rico Basins and in St. Croix by the end of the 21st century.

Table 2

Annual cycle of agroclimatic suitability for sweet pepper in the four basins of Puerto Rico and in St. Croix during the baseline climatology (1981–2010), and the two future climatologies: mid-century (2041–2070) and late-century (2071–2100)

Highly suitable class is in green, moderately suitable in yellow, marginally suitable in orange, and NS in red.

*HiS: when 80% condition of crop's maximum yield is met at least 60% of the years and the 60% condition is met at least 80% of the years; MoS: when 80% condition of crop's maximum yield is met at least 40% of the years and the 60% condition is met at least 60% of the years; MaS: when 80% condition of crop's maximum yield is met at least 20% of the years and the 60% condition is met at least 40% of the years; NS: when 80% condition of crop's maximum yield is met in less than 20% of the years and the 60% condition is met in less than 40% of the years.

View Large

Table 3

Annual cycle of agroclimatic suitability for banana and plantain in the four basins of Puerto Rico and in St. Croix during the baseline climatology (1981–2010), and the two future climatologies: mid-century (2041–2070) and late-century (2071–2100)

Highly suitable class is in green, moderately suitable in yellow, marginally suitable in orange, and NS in red.

*HiS: when 80% condition of crop's maximum yield is met at least 60% of the years and the 60% condition is met at least 80% of the years; MoS: when 80% condition of crop's maximum yield is met at least 40% of the years and the 60% condition is met at least 60% of the years; MaS: when 80% condition of crop's maximum yield is met at least 20% of the years and the 60% condition is met at least 40% of the years; NS: when 80% condition of crop's maximum yield is met in less than 20% of the years and the 60% condition is met in less than 40% of the years.

View Large

When focusing on sweet pepper (Table 2) during the baseline climatology, the results indicated that this crop was highly suitable over the entire year for the East, West, and North Puerto Rico Basins, while in the South Puerto Rico Basin and St. Croix, the highly suitable class is limited to August–January and October–January, respectively. During the early rainfall season (April–July), sweet pepper was already moderately suitable for the South Puerto Rico Basin and marginally suitable for St. Croix from 1981 to 2010. Future climatologies, starting in 2041, did not change the high suitability of sweet pepper in the East and North Puerto Rico Basins, except for July in the northern basin, in which the crop became moderately suitable after 2071. However, a gradual decrease in sweet pepper suitability was evident in the West Puerto Rico Basin in January–April, when this region is under water stress, and each month reduced at least one suitability class. On the other hand, the South Puerto Rico Basin presented a drastic change in sweet pepper suitability, mainly from February to August when the crop that used to be moderately suitable during the baseline climatology will become unsuitable after 2071. In St. Croix, the decrease of crop suitability occurred first because sweet pepper will be unsuitable from March to August starting in 2041.

For bananas and plantains (Table 3), the agroclimatic suitability assessment was even less optimistic. During the baseline climatology, the East and North Puerto Rico Basins were still highly suitable, but some moderately suitable categories will start to occur during the mid-summer dry spell in the East basin after 2071. For the North basin, only July is moderately suitable after 2041, while after 2071, this region has a drastic change to NS for banana and plantain. In the West Puerto Rico Basin, suitability also decreased from highly to moderately suitable in the baseline climatology to marginally-NS in January–March after 2041. However, a drastic decrease in banana and plantain suitability occurs in the South Puerto Rico Basin, from January to August, where most of the months that used to be moderately to marginally suitable became NS. The exception for this region is the hurricane season, from September to November, when crop suitability is high throughout the end of this century. In St. Croix, where bananas and plantains are already NS to be rainfed from March to August in the baseline climatology, the decrease in suitability expanded to 9 months of the year (January–September) after 2071. The greater decrease in banana and plantain suitability is probably related to the fact that they are more prone to water stress than sweet pepper. As mentioned previously, the ky value greater than one indicates crops are more sensitive to water availability, and the ky for banana and plantain is greater than the one for sweet pepper.

Therefore, the overall crop suitability in the region is predicted to decay during the first half of the year (January–July), affecting the early rainfall season and the two drier seasons’ agroclimatic suitability. The effects of the crop suitability reduction will be mostly felt in St. Croix and southern Puerto Rico. On the other hand, November is the only month when all crops, independent of the period and the region, will remain highly suitable. This is most likely due to the increase in precipitation projected by the future climatologies that were previously discussed (Figure 3).

These results suggested that temperature increases have a slightly larger impact than precipitation on crop suitability and water stress. While the temperature increase is predicted to occur yearlong, the greater reduction in precipitation is projected to occur only from June to October, which is not the months with the greatest water stress and poorest crop suitability. Similarly to what Curtis et al. (2014) found in Jamaica, the effect of temperature on agriculture suitability was probably related to the projected warming in the boreal winter, which increases the evapotranspiration demand during the Caribbean dry season.

The results discussed here indicated that southern Puerto Rico and St. Croix, which are already sensitive to water deficits, will face difficulty in keeping crop suitability for sweet pepper, banana, and plantain during most of the annual cycle after 2041 (Tables 2 and 3). For short rotation crops studied in Jamaica, drought and water deficits are major problems already affecting growth rate, fertilization, and yield (Rhiney et al. 2018), which will also be the case of sweet pepper analyzed here, whose life cycle ranges from 60 to 90 days. Other studies suggested that important crops such as coffee production would have an estimated reduction in up to 84% of highly suitable growing conditions in top-producing municipalities in Puerto Rico by 2070 (Fain et al. 2018).

Therefore, one of the strategies that may be considered to adapt to climate change is finding alternatives to sweet pepper, banana, and plantain, including more drought-tolerant crops, such as sweet potato and cassava (Campbell et al. 2011; Curtis et al. 2014). Furthermore, irrigation may also be an option, as parts of the South Puerto Rico Basin were already classified as conditional farmland soils, meaning the soils there typically need irrigation (Gould et al. 2017). Another alternative could be to invest in farmer field schools with farmer-centered approaches for collaboration and participation to solve the problems and plan for the future, which may increase the chances for knowledge diffusion and collective action (Tomlinson & Rhiney 2018).

Additionally, it is important to emphasize that agriculture was highlighted as one of the human systems most impacted by climate change at the community level (Robinson & Wren 2020), while results from the Fifth Assessment Report (AR5) of the Intergovernmental Panel on Climate Change (IPCC) indicated that climate change will impact agriculture in the Caribbean mainly due to changes in temperature and water availability (Lincoln Lenderking et al. 2021). Therefore, the reduction in agroclimatic suitability assessed here (Tables 2 and 3) and in past studies reinforced the likelihood of climate change being a threat to food security in Puerto Rico and St. Croix, because food security depends not only on food availability but also on food access, utilization, and stability that are all significantly affected by climate change (Rhiney et al. 2018; Lincoln Lenderking et al. 2021).

Finally, when comparing the results of Puerto Rico and St. Croix, we can see the different impacts of climate change on the Greater and Lesser Antilles. While Puerto Rico will still have some areas remaining suitable for crops most of the year until the end of this century, St. Croix will suffer more and sooner with water stress and crop suitability, with a chance to have most of the year without the ability to rely on rainfed crops. The differences in island areas (St. Croix is at least 50 times smaller than Puerto Rico) and water resource availability can play an important role in defining the islands that are going to feel the consequences of climate change first.

This work investigated how future climate can impact water stress and crop suitability for four basins in Puerto Rico (South, North, East, and West) and the island of St. Croix using the CROPRISK model. When comparing the baseline climatology (1981–2010) and two future climatologies (2041–2070 and 2071–2100), the results indicated that the projected increase in temperature and decrease in precipitation will continually decrease crop suitability and increase water stress over most of Puerto Rico and St. Croix after 2041. Among the regions, we highlighted the South Puerto Rico Basin and St. Croix as the most sensitive to future changes due to their greater water stress and lower crop suitability when compared to the other regions. By 2041, St. Croix will not be suitable for sweet pepper, banana, and plantain for at least half of the year, while southern Puerto Rico will follow the same path after 2071. Other regions of Puerto Rico will keep crop their suitability and lower water stress for almost the entire year, with the exception of the mid-summer dry spell months (June–July), when all regions exhibit the effects of climate change in their water deficits and banana/plantain crop suitability.

Overall, results indicated that bananas and plantains were more sensitive to water deficits than sweet pepper and that the effects of future climate change on water stress and crop suitability will be worse from January to July. This indicates that the projected temperature increase will have a slightly larger impact than precipitation on crop suitability and water stress in Puerto Rico and St. Croix. While the temperature increase is predicted to occur yearlong, the greater reduction in precipitation was projected to occur only from June to October, which are not the months with the greatest water stress and poorest crop suitability in the islands.

The reduction in crop suitability and greater water stress from January to July also indicated that the two dry seasons and the early rainfall season will be the periods when planning for agriculture will be critical in order to keep crops suitable in the future. Among the options for planning and adaptation, focusing on more drought-resistant crops and/or having an irrigation plan at the government level could be alternatives. Preserving forested and highly vegetated areas would also be important, as we saw their roles in controlling temperature range and precipitation in eastern Puerto Rico. Moreover, we believe that any effort that helps to reduce CO₂ emissions should be taken into consideration to prevent a drastic increase in temperature and its consequent impact on SIDS and smallholder farmers.

For future work, we recommend the replication of the methods applied here using other crops of interest in Puerto Rico and St. Croix. We believe that further investigation is needed to assess the possible impacts of climate change on Puerto Rico and St. Croix's agriculture, as well as to better understand the spatial distribution of those impacts. It would also be important to apply a similar approach to other nations in the insular Caribbean to understand the regional impacts of future climate on crop water stress, and, hopefully, to encourage a regional action/plan to better prepare and adapt to mitigate food insecurity. Finally, we understand that the outputs from the climate models used here should be interpreted as projected climate trends for an extreme climate change scenario, rather than predictions of future weather.

The authors thank Thomas Mote for providing insightful comments and suggestions that improved this manuscript, as well as Christina Chanes and William Gould for sharing with us some agricultural data from Puerto Rico and St. Croix. The authors also thank the National Science Foundation Luquillo Long-Term Ecological Research Program (DEB1239764) for the funding provided. The authors also thank the National Aeronautics and Space Administration (NASA) Earth Data for Daymet data availability (https://daymet.ornl.gov/), the U.S. Geological Survey for the Thornthwaite Water-Balance-model (https://www.usgs.gov/software/thornthwaite-monthly-water-balance-model), and the World Climate Research Programme for the CMIP6 data availability (https://esgf-node.llnl.gov/projects/cmip6/).

This work was supported by the National Science Foundation Luquillo Long-Term Ecological Research Program (DEB1239764).

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.