Spatial and temporal rainfall variability in the Caribbean coast of Costa Rica

Rainfall in the Moist Tropical Caribbean Region (MTCR) in Costa Rica occurs practically throughout the year, with the quarters June–August (JJA) and December–February (DJF) concentrating over 70% of annual rainfall. Contrarily, in March–April and September–October, it rains below 100 mm per month. This seasonal rainfall behavior makes the region ideal for producing bananas (Musa spp.) and pineapple (Ananas comosus) for export (10% and 8% of total exports in 2021, respectively). A national-scale study determined that agriculture in the MTCR is one of the most vulnerable sectors to climate changes. However, the climate in this region has been poorly studied so far. This research analyzed the spatial and temporal variability of annual, monthly, and seasonal (DJF, JJA, SO) rainfall in the MTCR and how they change in the study period based on quality-checked series of daily rainfall from 28 weather stations in two periods: 1985–2009 and 1997–2019. The results show that rainfall regimes in the region are variable in space and throughout the year, with peaks occurring close to the mountain range and minimum values close to the coast. Trends were statistically significant in the period 1985–2009 with a predominance of significant positive trends in DJF and significant negative trends in SO. No significant trends (positive or negative) were observed in the period 1997–2019. JJA rainfall has uneven regional distribution and presents a positive and significant trend in the mountain region. This paper contributes to filling the knowledge gap in rainfall seasonality, variability, and trends in a region where banana and pineapple commercial plantations are fundamental to the country’s economy thus providing information to decision-making in the agri-food sector to reduce the negative impacts of changing rainfall regimes.


Introduction
The Moist Tropical Caribbean Region (MTCR) in the Caribbean watershed of Costa Rica stretches over 23,130 km 2 , i.e., almost half the national territory (51,179 km 2 ). The country is strategically located between the Pacific Ocean and the Caribbean Sea (CS) in Central America. Large-scale processes that originate in those two water bodies have a major influence on the Central American climate (Durán-Quesada et al. 2020;Maldonado et al. 2018).
In particular, the regional rainfall patterns in Central America are determined by trade winds that play a relevant role in the transport of moisture from the CS and thus in the seasonal variability of climate variables (Alfaro 2002;Durán-Quesada et al. 2017;Poveda et al. 2014;Taylor and Alfaro 2005). There is also influence from low-level winds associated with the Caribbean low-level jet (CLLJ) (Amador 1998;2008) and from low-level winds from the Eastern Tropical Pacific (ETPac) in the west that are associated with the Chocó jet (CJ) (Poveda and Mesa 2000). Their role in modulating rainfall patterns and their connectivity with large-scale features such as the Intertropical Convergence Zone (ITCZ) and El Niño-Southern Oscillation (ENSO) (Durán- Maldonado et al. 2018) have been extensively studied (Cook and Vizy 2010;Durán-Quesada et al. 2017;Hidalgo et al. 2015).
The complex topography of the Central American isthmus interacts with the northeasterly trade winds that are associated with the North Atlantic anticyclone (Amador 1998). Such interaction induces spatial climate variations that may lead to opposite behaviors over the Pacific and the Caribbean watersheds (Quesada and Waylen 2020).
Specifically, the rainfall in the Pacific watershed of Costa Rica has bimodal behavior, with a well-defined dry season in the boreal winter (December-March), which in some areas extends to April, and a rainy season with two rainfall peaks, one in May-June and the other one (generally more intense) in September-October (Alfaro 2002). Between the two rainfall peaks, a period of decreased rainfall occurs in July-August, which is called mid-summer drought (MSD) and is locally known as veranillo or canícula (Amador 2008;Magaña et al. 1999;Maldonado et al. 2016). The spatial and temporal variability of the rainfall in this slope has been extensively studied (e.g., Alfaro and Hidalgo 2021;Maldonado et al. 2018;Quesada and Waylen 2013;Retana 2012). Mainly in the last decade, progress has been made in the analysis of extreme events (AlMutairi et al. 2019;Guillén-Oviedo et al. 2020;Quesada-Hernández et al. 2020;Quesada-Montano et al. 2019) and the influence of tropical cyclones on the triggering of floods and landslides (Quesada-Román et al. 2019).
The Caribbean watershed has no defined dry season (Alfaro 2002). The rainfall does not vary much between January and the middle of October. This period accounts for approximately 60% of the annual total in this region. From mid-October onwards, there is a marked increase in rainfall accumulation of between 400 and 600 mm per month (Taylor and Alfaro 2005). This annual cycle presents a spatial variation from warm subhumid with low rainfall in the northwest of the region to excessively wet and cold climate in the mountain area (Pérez-Briceño et al. 2017). Although the region lacks a defined dry season, there are two periods of significant rainfall decrease in March-April and in September-October (Alfaro 2002;Alfaro and Hidalgo 2021;Sáenz and Amador 2016).
The Caribbean watershed has a growing industrial and port activity, likewise recreation activities. It is one of the main tourist destinations in the country, due to its landscapes (Quesada-Román and Pérez-Briceño 2019) and the natural beauty of its beaches and reefs (Piedra-Castro et al. 2021). On the other hand, it supports extensive agricultural activity (Quesada and Waylen 2020). The climate features acting on the Caribbean coastal plain give rise to ideal conditions for the production of bananas (Musa spp.) and pineapple (Ananas comosus). Both are the main export crops of Costa Rica and represent 39% and 31%, respectively of the agrifood sector export, or 10% and 8% of total exported goods (PROCOMER 2021). However, according to Bouroncle et al. (2015), the agri-food sector in this region is the most vulnerable economic sector to climate changes.
The temporal rainfall variability has not been studied in detail on the Caribbean slope. The literature includes works such as that of Quesada and Waylen (2020), which analyze the daily rainfall series from individual station and long record (Limón and1949-2017), or Alfaro (2002), who includes a few meteorological stations located in the CS for at least 20 years  in his study of Central America.
The climate variability creates conditions that may exceed the adaptive capacity of the activities that depend on it, which can lead to social vulnerability and economic losses. The Sixth Assessment Report of the Intergovernmental Panel on Climate Change mentions that human-induced climate change is already affecting many weather and climate phenomena, mainly extremes. These include an increase in the frequency and intensity of extreme precipitation events (excesses/deficits) (IPCC 2021).
The impact of climate variability and in particular rainfall on the environment and human activities depends mainly on the change in the frequency of occurrence or intensity of extreme events. According to Zhao et al. (2022), the annual rainfall over the Caribbean side is more complex in terms of the spatial variability. To address this issue, studies with high-quality data focused on the specific characteristics of the orography and climate of each region are necessary.
Under the hypothesis that the spatial and temporal variability of extreme events is greater than that of the mean values, a detailed understanding of the spatial and temporal variability of rainfall, especially of extreme events, will allow in future work to better assess the forcings (local and remote).
The objective of this research is to analyze the spatial and temporal variability of annual, monthly, and seasonal (DJF, JJA, and SO) rainfall, with special attention to extreme events and how they change in the study period. The interest in these seasons lies in the fact that they are critical periods for pineapples and banana crops (Serrano et al. 2008;Soto 2014).

Observational data
First, it is important to clarify that the MTCR in the Caribbean watershed of Costa Rica is one of the three major climatic regions determined by the National Meteorological Institute (IMN, by its Spanish acronym) of Costa Rica, which arise due to the mountainous disposition and the interaction with the trade winds (Solano and Villalobos 2001).
Thus, the MTCR is composed of the climatic regions: Northwest, Northeast, North, and South Caribbean (Fig. 1).
This paper examined daily rainfall series from initially 40 meteorological stations covering the longest period from 1943 to 2019 (Fig. 1). Data were provided by the IMN and the Costa Rican Electricity Institute (ICE, by its Spanish acronym). Quality and consistency controls made it possible to identify gaps in the time series of some meteorological stations, as well as inconsistent (negative) values and outliers. We removed negative values and checked the validity of outliers by analyzing daily and monthly data from nearby stations and information from monthly meteorological IMN newsletters (https:// www. imn. ac. cr/ bolet in-meteo rolog ico). Finally, after determining the number of missing data in each series, we excluded from the study the stations with 10-15% missing data, according to the criteria set by the World Meteorological Organization (WMO 2004). Table 1 shows the period, location, and altitude above sea level of the selected 28 meteorological stations included in the study (Fig. 1), from the 40 originally considered. The longest common period of 24 stations is 1985-2009. Four stations are located in areas of pineapple or banana cultivation and have uninterrupted observations from 1997 to 2019. Specific statistical analysis of these four stations was performed and compared to nearby stations in the 1985-2009 records. Since no particular issues were observed in the exploratory processes, the data were included in the climatological analysis of the study area.

Methodology
This study analyzed annual, monthly, and seasonal (DJF, JJA, SO) high-quality rainfall data from 28 weather stations. The climatological analysis was based on the mean, median, standard deviation, and the 25th and 75th percentiles of the rainfall series. In addition, we determined the contribution of seasonal rainfall to annual totals. The spatial analysis was made with the spline with barriers interpolation method. This minimum curvature method makes it possible to analyze information from regions with natural barriers (orography, lakes, and rivers) or surface discontinuities (Briggs 1974;Wolny et al. 2017). A non-parametric Mann-Kendall test (Siegel 1956) was used to analyze the temporal variability, with a statistical significance of 90, 95, and 99%. The test was performed on series without data gaps in the two periods of analysis: fifteen stations in 1985-2009 and five stations in 1997-2019. Finally, trends were examined in the percentage of seasonal rainfall (DJF, JJA, SO) contribution to annual rainfall.

Annual and monthly rainfall climatology
We first analyzed the spatial distribution of annual and monthly rainfall in the MTCR in Costa Rica (Figs. 2 and 3, respectively). The total mean annual rainfall in the region is around 4000 mm, with amounts ranging from 1750 to 7300 mm approximately. Geographically, the greatest amounts (above 7000 mm) fall in the part of the mountain that separates the Caribbean and Pacific watersheds, following a southwest-northeast line that coincides with the plains located close to the CS coast (Fig. 2). The lowest annual rainfalls, about 2000 mm and less occur in the north close to the border with Nicaragua and the southeast of the study area (Chacón and Fernández 1985;Fernández et al. 1996). The uneven distribution of annual rainfall in the region observed here agrees with that found by Pérez-Briceño et al. (2017) for the period 1960-2011.
The mean annual rainfall cycle in the MTCR presents monthly and spatial variations. Rainfall peaks in June-August and in November-December, mainly in the northeast and southwest of the region, with a characteristic strip appearing in December that connects the northeast of the Caribbean coastline with the mountain range in the southwest. Although more intense, this spatial pattern is also observed in July (more intense) and in August (less intense). The characteristics of the rainfall cycle of July and August are related to the MSD (Magaña et al. 1999), which results in relatively low precipitation over the Pacific watershed and intense rainfall in the Caribbean watershed in both months (Hidalgo et al. 2015). Rainfall amounts decrease towards the west of the study area in the first part of the year (January-April), with rains below 100 mm per month, mainly in March and April. Another less pronounced drop occurs in September-October along the Caribbean coast following a mountain-coast gradient, where rainfall amounts plummet from the mountain range towards the northeastern Caribbean coast, with a difference of 500 mm between both points (Fig. 3).
A spatial behavior similar to that of the annual rainfall cycle is observed in January and February, with peaks on the mountain range (mainly in the southwest) and minimum values in both the northwest extreme at the border with Nicaragua and the southeast extreme of the region (Fig. 2).

Seasonal rainfall climatology
Given the importance of seasonal rainfall variability to pineapple and banana crops, a detailed analysis was made of rainfall in the periods JJA, SO, and DJF. Quarter JJA has a spatial distribution similar to that of the climatology of annual rainfall in the MTCR, with peaks above 2000 mm in one part of the mountain range and minima in the northwest and the southeast of the region. A change in the rainfall regime is observed in SO when the smallest amounts follow the Caribbean coastline, and the peaks occur in the mountain area, close to the border with the Pacific watershed in the southwest. In DJF, the greatest amounts occur along the Caribbean coast and in some areas in the southwest, close to the Pacific watershed (Fig. 4, top). Based on this, a shift in the spatial distribution of rainfall is seen in JJA and DJF, with both quarters presenting maximum values over the coast and in sectors of the mountain area, while in SO, minimum values occur along the coastline and maximum values over the mountain range.
In addition to the analysis of-annual, monthly, and seasonal-regional rainfall behavior, we analyzed the contribution of seasonal rainfall to annual totals (Fig. 4, bottom). In general, spatial changes are observed according to the time of the year. In JJA, the contributions close to and over 40% concentrate in the north of the study region. It is worth highlighting that from 30 to over 40% of the MTCR rainfall occurs in JJA. On the other hand, around 20-30% of the annual rainfall occurs in DJF. A mountain area in the southwest close to the border with the Pacific watershed explains up to 10% of annual rainfall. This situation reverts in SO when the coastal area accounts for less than 15% of annual rainfall while contributions in the mountain range are above 35% (Fig. 4, bottom). This is evidence that approximately 70% of the region's precipitation is concentrated between the JJA and DJF quarters.
Piedades Sur station, located in the extreme southwest of the MTCR, in the mountain range, presents opposite behavior to the rest of the domain, as it concentrates 40% of the annual rainfall in SO and less than 10% in DJF. Such behavior might be explained by its closeness to the Pacific watershed, as its annual cycle is similar to that of the Pacific (Maldonado et al. 2021).
The information provided by percentile 25 is of major importance mainly at the stations located in banana and pineapple crop areas (see crop location in Fig. 1), given that rainfall below 100 mm per month has negative impacts on banana crops, according to Robinson and Galán-Sauco (2010). Such rainfall values were observed in February (north Caribbean region), March (south Caribbean region), and September, when they extended over the entire Caribbean coast (results not shown).
Pineapple crops can develop with minimum monthly rainfall ranges from 80 to 100 mm (Paull et al. 2017;Py et al. 1987). The first four months of the year (January-April) are the most critical for pineapple, when rainfall in percentile 25 in each of those months ranges from 20 to 100 mm per month. The northeastern and northwestern regions are identified as the most affected, particularly from February through April.
Following with the analysis of percentile 25, seasonally (Fig. 5 top), JJA presents rainfall amounts between 800 and 1000 mm over almost the entire region, which would potentially have no impact on crops in terms of minimum water requirements. SO presents amounts below 400 mm in the border strip with Nicaragua and the plains near the Caribbean coast where bananas are grown. Values even below 200 mm can be observed in Punta Castilla in the north extreme of the Caribbean coast, on the border with Nicaragua. In DJF, it rains less than 400 mm, and even less than 200 mm over a small strip in the northeastern region, where pineapple crops are located.
The areas with larger rainfall amounts corresponding to the 75th percentile (Fig. 5, bottom) are located in the center of the MTCR, with south-to-north latitudinal distribution and the greatest rainfalls (above 2 000 mm) concentrating in the south over the mountain area during JJA. The amounts in this period ranging from 1600 to 2200 mm can be harmful to both crops-bananas and pineapple-which are mostly located in the northern plains, from the foothills of the mountain range to the border with Nicaragua. This situation repeats in DJF, though with less rain (between 1400 and 2000 mm) and the spatial distribution displaced to the east of the domain, closer to the Caribbean coast.

Trend analysis
Annual and monthly rainfall trends were analyzed for each period (Figs. 6 and 7) and in the three selected seasons (JJA, DJF, and SO). In general, the annual trends (Fig. 6) are statistically significant at some stations in the period 1985-2009 (circles in Fig. 6); the greatest significance (99%) was observed at the stations located over the mountain range. In contrast, trends were not significant at the five stations in the period 1997-2019 (triangles in Fig. 6).
The analysis of the spatial behavior of monthly rainfall trends (Fig. 7) showed an interesting heterogeneous behavior both in space and time. For instance, although with opposite signs, September and November stand out for their high significance in the period 1985-2009 (Fig. 7). In September, negative trends prevail in the Caribbean coast, while in November, trends are positive and significant (99%) at most stations. Indeed, a decreasingalthough not significant-trend is observed in rainfall from August through October, mainly over the plains of the north and south Caribbean regions. September stands out with significant trends in the north Caribbean region. The situation reverts in November, when significant positive trends are observed at all the stations, except for the Significant positive trends concentrate over the mountain range in April and May; in June, a sign switch is seen at the stations of the North Caribbean region, although the trend is not significant. The situation in July is different with all the stations presenting increasing-non-significant-trends, except for two of them located in the plains of the eastern and the northwestern regions, respectively.
The period 1997-2019 presents few signals with significant trends. The months with significant positive trends are May, July, and September all at different stations (triangles in Fig. 7).
We analyzed the trends for the three seasons and found that the first period (1985-2009) (circles, Fig. 8, top) presents a dominance of positive trends in JJA and DJF. Most of the trends in DJF are significant. The opposite behavior is observed in SO when significant negative trends predominate. In addition, the trend behavior stands out in JJA, which presents negative, though non-significant trends at some stations, mainly near the Caribbean coast. In the second period (1997-2019) (triangles, Fig. 8, top), negative trends dominate in the three seasons although they are not significant.
JJA trends in the 1985-2009 period are positive and significant only at a couple of stations near the mountain area, while negative, though non-significant, trends are observed in the Caribbean plains (below 200 m asl) close to the coastal area. During this quarter, the behavior of June and August rainfall plays a major role in the north Caribbean region, where stations present negative trends, which switch sign in July (Fig. 7).
A dominance is seen of intense negative trends near the coast, mainly in the north during SO in the period 1985-2009 (Fig. 8, top). In the mountain area, trends are positive at some stations in the south and the southwest and negative in the southeast; however, none of them is significant. Values are significant mainly at stations close to the coast, and significance decreases towards the mountains. The five stations located in the Caribbean watershed have significant values at the 99, 95, and 90% levels. Only one station in the west has a significant trend at 90%. The comparison with the results of Fig. 7 reveals that both September Fig. 6 Annual rainfall trend in the MTCR, Costa Rica. Circles (period 1985-2009), triangles (period 1997-2019) and October present negative trends, mainly in the north and south Caribbean regions. These trends are predominantly significant over the former region in September.
Next, the DJF quarter in the period 1985-2009 shows a categorical change in the rainfall trend, which becomes positive and significant at most of the stations, particularly in the northeastern, north, and south Caribbean regions. More marked trends stand out in the mountain range as well as in the plains of the north Caribbean region. Only two stations present no statistical significance in DJF, one in the north of the study region and the other one in the south (Fig. 8, top). The comparison of the result with the trends of individual months (Fig. 7) shows that each one presents positive trends at practically all studied stations, being January the most significant.
Given the importance of these seasons for banana and pineapple crops, we analyzed the trends of the contributions of each season to total annual rainfall (Fig. 8, bottom). It is worth noticing that although total annual rainfall presents a positive trend, the trend of the contributions of seasons JJA and SO to the annual trend is negative and significant in most of the stations (mainly in SO). On the other hand, DJF presents positive trends with significant values in the mountain area. Trend behavior in the period 1997-2019 is uneven and not significant in any season.
These results can be connected to the path of the trade winds within the domain, mainly in DJF, with positive significant trends near the coast and in the mountain area. The trade winds are more intense in DJF and weaken in May; then, they become stronger in July and weaken again in October (Alfaro 2002). These winds contribute moisture to the windward watershed and have a major influence on the climate of the region .
The negative trend found for JJA, when the trade winds are stronger and cause a rainfall peak, according to the annual cycle of the region (Hidalgo 2021;Hidalgo et al. 2015;Maldonado et al. 2021) suggests the need for a more detailed analysis of this quarter and especially of July to verify whether rainfall maxima are decreasing along the years.
An interesting fact is that in the most recent 23-year period (1997-2019), rainfall trends are not significant. In the north of the country, close to the border with Nicaragua, trends are negative in SO and DJF and positive in JJA. At the stations close to the Caribbean coast and the Panama border, trends are positive, though not significant in the three seasons analyzed.

Discussion and conclusions
Contrarily to the Pacific watershed, meteorological information in the Caribbean watershed of Costa Rica is scarce, and there are few studies available. This study is to provide updated climate information of the rainfall variability in Costa Rican MTCR (Moist Tropical Caribbean Region), The aim of this study is to contribute to a better understanding of the spatial and temporal variability of rainfall in the Caribbean slope, using high-quality series from 28 stations in two periods 1985-2009 and 1997-2019. This understanding especially of extreme rainfall leads to a better assessment of their forcing in future research. Furthermore, following Zhao et al. (2022), annual precipitation over the Caribbean slope is more complex in terms of spatial variability and demonstrates a close connection with different forcings such as trade winds, ITCZ, ENSO (Durán- Maldonado et al. 2018;Martínez et al. 2019;Sáenz et al. 2022), and other regional patterns such as CLLJ, CJ, and tropical disturbances (Amador 1998;2008;Durán-Quesada et al. 2017;Hugo et al. 2013).
The main results revealed that rainfall regimes are variable both in time and space throughout the year in the very MTCR (Alfaro 2002;Maldonado et al. 2021), where the mean annual total is around 4000 mm. The spatial distribution of the annual rainfall cycle repeats in January and February, with peaks (between 500 and 600 mm) in the mountain range in the southwest and minimum values in the northwest at the border with Nicaragua (less than 100 mm). These results are consistent with previous studies (Alfaro 2002;Fernández et al. 1996;Pérez-Briceño et al. 2017).
Minimum monthly rainfall values occur in the west (mainly in the northwestern region) during the first 4 months of the year (January-April), being March the month with the lowest rainfall in the entire region (with values below 100 mm and not exceeding 400 mm). In the last 2 months of the year (November-December), peaks between 500 mm and above 700 mm per month and minimum values larger than 200 mm are observed in the east, over the coastal plains (particularly the north Caribbean region). In July and December, a southwest-northeast strip develops with abundant monthly rainfall (between 500 and 700 mm) that connects the mountain range with the north of the Caribbean coast. These results are in agreement with those of Alfaro (2002) for the period 1950-1994, Amador et al. (2013) for 1960-2011, and Maldonado et al. (2021) for 1976 Rainfall amounts in the seasons analyzed are quite different throughout the region. Although the amount of rainfall in JJA is greater than in DJF, spatial behavior in both seasons is similar to that of annual rainfall amount, whereas in SO, the pattern changes, showing a mountain range-coast gradient.
When assessing how much the seasonal total accounts for the annual total, the rainfall pattern in JJA is above 30%, and in regions to the northwest, it is close to 50%. In agreement with Sáenz and Amador (2016) in their study of 20 stations in the period 2006-2011and Villalobos and Rojas (2016 in the period 1941-2016 who analyzed only the Limón station, similar results were found-JJA accounts for the greatest concentration of annual rainfall-38%-followed by DJF-32.5%. These results highlight the contrast between rainfall regimes in the Pacific and Caribbean watersheds of Costa Rica, whose annual cycles have opposite behavior. On the one hand, the MSD in July-August causes low rainfall amounts in the Pacific (Magaña et al. 1999;Hidalgo et al. 2015;Maldonado et al. 2021); on the other hand, rainfall peaks occur in the Caribbean during those months. In addition, the Pacific rainfall regime presents the most intense peak during SO (Maldonado et al. 2021), while a decrease is seen in the Caribbean during those 2 months.
Regarding the spatial behavior of extreme values (25th and 75th percentiles), in general, JJA presents the greatest values, SO is the lowest, and an uneven behavior is seen in DJF (peaks in the east and over the coast in the north and south Caribbean regions and minimum values in the north of the northwestern region). In SO, the smallest amounts (below 400 mm) concentrate on the Caribbean coast (in the areas of banana crops) and the highest, in the mountains (1600 mm). During DJF, maximum values (between 1600 and 2000 mm) stretch in the southeast-northeast direction (mountains-north Caribbean coast), which could affect banana and pineapple crops in the north Caribbean region.
Bananas are cultivated in the north and south Caribbean regions and require a minimum rainfall of 100 mm per month (Robinson and Galán-Saúco 2010;Salvacion 2020). Lower amounts of rain cause the growth of leaves to slow and stop (Galán-Saúco and Robinson 2013). Pineapple crops are located in the north Caribbean region and require at least 80 mm of rainfall per month to thrive (Paull et al. 2017;Py et al. 1987). The effects of atmospheric water deficit on pineapple are reduced growth, longer growing cycle, and reduced fruit weight (Jiménez 1999). High rainfall can affect both plant growth and fruit weight, and waterlogging can lead to plant death (Lobo and Yahia 2017).
Based on the above, the highest risks of extreme water deficit events for banana crops occur in February, March, and SO, where percentile 25 shows rainfall amounts below 200 mm over the northernmost part of the north Caribbean region. For pineapple crops, this situation occurs in the first 4 months of the year (January-April). On the other hand, wet extremes and their impact on crops are higher in JJA for both crops and in DJF for banana plantations.
Trend analysis was carried out in two longest common periods (1985-2009 and 1997-2019). In the first one ) annual trends are positive over the entire MTCR and statistically significant along the mountain range. However, the monthly behavior is quite uneven. A decreasing pattern is observed from August to October. The most outstanding results are the rainfall reductions (negative significant trends) in September, particularly in the coastal plains of the north Caribbean region. Although the trends in August and October are slight and not significant at all the stations, they are relevant to crop growth.
Seasonal trends, in both periods, in general, present uneven behavior. For the period of 1985-2009, the result that is more outstanding is the positive and significant trend in DJF over the entire MTCR, and SO trends are negative and significant at the stations located in the coastal plains of the north and south Caribbean regions. An interesting fact is that positive and negative (annual, monthly, and seasonal) trends were not significant in the past 20-30 years .
Other authors have analyzed monthly or annual trends in the study region. However, these results may differ because they used different periods. For example, Maldonado et al. (2021) indicate that the Caribbean region shows an increase in rainfall in July and a significant decrease in September (period 1976-2015). Alfaro-Córdoba et al. (2020), for the period 1970-1999, focus the analysis in the aridity trends on the Pacific slope. Hidalgo et al. (2017), in their study for the period 1970-1999, suggest positive trends on the southern Caribbean coast of Costa Rica, which is consistent with our study in annual terms. Hannah et al. (2017) show interesting results focused on the agricultural sector for the period 1960-2017, indicating that rainfall trends in the region are highly variable spatially, which is consistent with our research.
The trends in the percentage contribution of seasonal rainfall to the annual cycle show decreasing amounts in JJA and SO, which are significant in the plains of the northeastern and north Caribbean regions during JJA and over the entire MTCR during SO. This means that decreasing trends are observed in both seasons. As a consequence, there are five consecutive months (June through October) when the rainfall contribution to the annual cycle is decreasing, the most evident period being SO. On the contrary, the contribution of DJF to the annual cycle would be increasing, particularly in the northeastern region and at some stations in the north and south Caribbean regions.
These findings imply an alert for fruit farmers, mainly in SO, since rainfall during these 2 months is between 200 and 300 mm over the entire Caribbean coast. And even more, the trend for SO is negative and significant over the plains of the north and south Caribbean regions, and the trend of the SO rainfall percentage contribution to the annual cycle is also negative and significant in the entire MTCR.
These results highlight the need for climate monitoring for pineapple and banana crops in Costa Rica, which begins yearly in September and continues until May of the following year. The most critical months are SO because of possible water deficit and DJF for water excess, as the latter period accounts for an average of 32.5% of annual rainfall (Sáenz and Amador 2016;Villalobos and Rojas 2016). Since DJF is between two low-rainfall periods-SO and March-April of the following year-it is of major importance to rainfall distribution. Reduced rainfall in this quarter could have negative impacts on the crops given that it is followed by a low-rainfall period.
Moreover, more detailed and specific studies of the changes in rainfall regimes used fixed thresholds for critical crop periods to help plan and optimize production. Such research is important as a baseline for further studies and input to decision-making in the agri-export sector of the MTCR, aimed at minimizing the negative impacts of extreme rainfall events.
Overall, this study provides a contribution to improving the understanding of rainfall (monthly, seasonal, extreme) in the Costa Rican Caribbean watershed and as an input for agricultural planning in the region, particularly pineapple and banana plantations-the most important export products-which are mainly produced in the MTCR. On the other hand, this research could serve as a basis for future studies focusing on the physical circulation mechanisms that affect the spatial and temporal distribution of rainfall in the Caribbean slope.