Water availability within the Conservation Area of Guanacaste (ACG), calculated through water yields over 40 years (1980 - 2015), shows a pattern of change between high and low values on average every 3 to 4 years (Figure 10a). Baron-Ruiz [9] offered a first theoretical analysis in this regard, mentioning that the simulated data demonstrate the ability of the SWAT hydrologic model to capture climate phenomena, such as El Niño-Southern Oscillation (ENSO), which determine the spatial and temporal patterns of climate in Costa Rica [15]. This is observed in the water yield's annual distribution (Figure 10b), where extreme data of high water availability coincide with the rainiest years identified as Niña periods in Costa Rica (1988-1989, 1998-1999, and 2010-2011) and extreme values of low water yields coincide with drought or Niño years (1982-1983, 1997-1998, and 2015-2016). This analysis is confirmed by the distribution of Precipitation (Figure 11b) and Temperature (Figure 12b) data, where the same trend of extreme values is visualized for the periods mentioned. However, this analysis is limited as it is only descriptive and does not allow to establish and quantify the true trend of water availability within the ACG. Therefore, an attempt was made to overcome this limitation through linear regression models.

The ACG's water yield tends to rise over time (Figure 15). It increases by approximately 138.5 mm (p-value: 5.8 X 10^-67) per decade. The Tempisque River basin shows the highest increase, with 144.9 mm (p-value: 2 X 10^-41) every ten years, followed by the Nicaragua Lake basin, with an average very close to the study area value, 133.9 mm (p-value: 8.7 X 10^-23). The watersheds towards the Pacific Ocean show the lowest decadal averages, with 128.9 mm (p-value: 5.9 X 10^-10) for Santa Elena Bay and 116.7 mm (p-value: 5.9 X 10^-7) for the Gulf of Papagayo. Interestingly, the two watersheds with the highest water yield's decadal increase show a positive trend in the recovery of evergreen forest (Figure 14c); 30% for the Tempisque River watershed from 1979 to 2015, equivalent to 136 km² of forest recovered, and 21% for the Lake Nicaragua watershed equal to 154.8 km² of additional forest area by 2015 (Figure 9).

Both precipitation (Figure 16) and temperature (Figure 17) for the whole study area show the same increasing pattern over time, with 11.7 mm (p-value: 0.03) of additional rainfall every ten years and an increase of 0.1°C (p-value: 8.4 X 10^-13) every decade; which would allow at first to conclude that climate change can explain the temporal variation of water availability. However, when discriminating this analysis by each of the four major basins, only temperature maintains significance (p-value < 0.01) in the 0.1°C temperature increase per decade for all basins. Regarding precipitation, the increase is significant (p-value: 0.016) only for the Nicaragua Lake basin, with 19.4 mm per decade. This leads to infer that other factors, not only climate, are responsible for the trend in water availability for the rest of the basins. Could it be the land cover change?

The change in water availability between 1980-1989 and 2010-2019 is not significantly explained (p-value: 0.9) by forest change, either in the sub-basins with higher deforestation or in those with higher forest cover recovery (Figure 18).

The lack of significance observed for the above regressions could be attributed to the lack of spatial representativity when analyzing a small number of sub-basins. Therefore, the four sub-basins were included as a covariate and outliers were discarded (Figure 19). It was observed that in the Nicaragua Lake (19a) and the Tempisque River (19b) basins (88% of the study area), the increase in forest generates an inverse response in the change in water availability between decades. In other words, the sub-basins where the forest has recovered the most are the ones where the slightest increase in water availability has been observed. This confirms previous studies, which concluded that deforestation increases water yield and reforestation decreases it [8]. In this case, an increase of 10% in the forest land cover could decrease the water yield by approximately 4% (p-value: 4.4 X 10^-4) in the Nicaragua Lake watershed and 0.8% (p-value: 0.04) in the Tempisque River basin.

The previous results are surprising because they are contrary to the initial findings made by Baron-Ruiz [9] for the ACG, where the dominant land cover within each sub-basin was analyzed as a categorical variable and entered into a linear model, concluding that there is an increase in water availability related to sub-basins where forest cover dominates. This, in turn, was supported in the literature, where several studies for the Central American region report positive values of water yield following forest recovery [16]. According to the results obtained in this project, there is not enough statistical evidence to conclude that the observed trend of increase in water availability (Figure 15) is directly related to the increase in forest cover for the total drainage area of the Conservation Area of Guanacaste (ACG). However, when looking into individual watersheds, a trend was observed for the Nicaragua Lake and Tempisque River basins with enough statistical significance.

• This project reflects what was learned during the lectures about the importance of thoroughly investigating statistical results. Generalities in statistics, such as a low r or r² or a p-value > 0.05, which would indicate low statistical significance, must be analyzed in the specific context of each study to make the most of what the data are meant to convey. For example, most of the linear regression models presented in this section have correlation coefficients (r), and hence coefficients of determination (r²), that subjectively can be considered low, which would initially lead to the conclusion that there is a weak relationship between the variables or that the predictor variables cannot explain the variability of water availability. However, the models were significant (p-values < 0.05), which reflects that the predictor variables do have an effect on water yield with 95% confidence but that the predictions made are not very precise due to the low correlation; however, they are in line with the reality of the study area and may be valid for decision making.

• The statistical analyses presented allow to conclude that of the two biophysical factors studied, climate is the one that explains the increasing trend in water availability for the drainage area of the Conservation Area of Guanacaste (Figure 15). The predictions of 138.5 mm of decadal growth in water yield in conjunction with an increase of 11. 7 mm of rainfall and 0.1 °C every ten years provide essential baseline data for decision-making to identify anomalies that generate warnings for preserving all ecosystem services that depend on water availability in this area of high ecological value for humanity.

• Although more statistical evidence was needed to relate the change in water availability to the change in forest cover, this topic needs further research. Costa Rica is a global example of sustainability. The establishment of novel environmental policies and changes in other socio-economic factors, such as increased investment in education and health, have encouraged nature conservation and forest recovery. Forests covered 26% of the country in the 1980s and, by the 2000s, rose to 47% [1, 15]. In addition, several studies in Central America point to the positive impact of forest restoration on water yield, as forests underpin the regeneration of all the hydrological cycle processes [16].

• Finally, this project focused on investigating two biophysical factors, climate and land cover. Still, others deserve attention, such as topography, which seems to affect the spatial stratification of climatic and hydrological variables, as evidenced in the maps presented here. It is worth remembering that several volcanoes and mountain ranges divide the Pacific and Caribbean zones in this area and therefore affect the region's climate [15].