Methods

WATER AVAILABILITY ESTIMATES

Figure 5. SWAT's Hydrologic Processes

Schema of the processes simulated by the Soil and Water Assessment Tool (SWAT) hydrologic model. These processes represent the water cycle. They are calculated by adapting the water balance equation to each process. They are spatially defined using Geographic Information System (GIS) interfaces, such as QGIS or the ArcGIS Desktop suite. The SWAT's water balance equation is used for water yield calculation (Equation 1).

Source: Adapted from [14].



  • Hydrologic Model

The current course project will consider the water availability estimated through a Soil and Water Assessment Tool (SWAT) model. SWAT is a comprehensive and computationally efficient model which visually represents the hydrologic cycle processes (Figure 5) through spatial features (i.e., maps) and time series. SWAT has several documented applications in water quantification, with water yield simulation as the most common output variable for estimating water availability at the sub-basin scale [9]. The main advantage of SWAT is its capacity to spatially and temporally model water cycle processes, which allows its adaptation to different areas of study by changing the input data and applying calibration and validation processes with in-situ measurements.

  • Water Yield

Water yield data was modelled through the SWAT model created by a Master of Science thesis [9]. To estimate water yield, SWAT utilizes the concept of water balance, also known as water budget or hydrological balance, which is the balance that occurs over a period of time between the inputs and outputs of the hydrologic cycle together with the water that is already stored in any watershed. Equation 1 shows the adaptation of the water balance equation in the SWAT model to estimate the water yield.


As a result, an annual average water yield value is obtained for each sub-basin within the drainage area for 40 years between 1980 - 2019.

Equation 1. Water Yield Balance

Adaptation from the water balance equation to estimate the variable of interest in this project, the Water Yield, in terms of the water that leaves every sub-basin and enters the main channel.

Source: [14].

STUDY AREA

Figure 6. ACG's Natural Richness

Examples of the natural richness at the Conservation Area of Guanacaste (ACG)-Costa Rica.

Source: Ⓒ Oscar Baron-Ruiz

  • The World Heritage Site

The Conservation Area of Guanacaste (ACG) is this project's area of interest. In 1999, the United Nations Educational, Scientific and Cultural Organization (UNESCO) declared the ACG as a World Heritage Site because it harbours very complex biological and ecological dynamics. It connects the Pacific Ocean and the lowlands of the Caribbean. Astonishingly, different ecosystems of humid and dry tropical forests, savannas, and mangroves coexist in the ACG. This natural richness encompasses iconic landscape features such as volcanoes, coral reefs, cobble beaches, among others (some examples in Figure 6). More than 235,000 species have been identified in this conservation region comprising about 65% of the total biodiversity species estimated in Costa Rica and 2.6% of global biodiversity [10, 11].




  • The Drainage Area

The study area is the drainage area of the ACG delineated by the SWAT model mapped in Figure 7. This drainage area (2,861.9 km²) is divided into 107 sub-basins, which can be grouped into four major basins identified in the official cartographic information of Costa Rica as: 1. The Nicaragua Lake watershed (1,143.3 km²), 2. The Santa Elena Bay (208.4 km²), 3. The Papagayo Gulf (144.4 km²), and 4. The Tempisque River basin (1,365.8 km²). These basins comprise 39.9%, 7.3%, 5.0%, and 47.7% of the total drainage area, respectively.


All the data used in this project is calculated as an absolute value for each sub-basin.

Figure 7. Study Area

This course project's study area, equivalent to 2,861.9 km², encompasses the drainage area within the ACG and the 107 sub-basins simulated by the SWAT model. Four main basins colour this area: Nicaragua Lake (purple), Santa Elena Bay (blue), Papagayo Gulf (red), and Tempisque River (green). This area's Costa Rican hydrometeorological network comprises one streamflow station (blue drop symbol) and four weather stations (clouds and sun symbols). Daily weather information of precipitation, temperature, wind speed, and relative humidity was used to generate the SWAT model. The streamflow data was used for calibration and validation of water yield simulations.

RESEARCH QUESTION 1


The weather data was spatially limited in the study area (Figure 7). Therefore, to assess the relationship between climate factors and water availability, annual weather information with better spatial representation is estimated for each sub-basin, as described below.


Mean Annual Precipitation (MAP) and Mean Annual Temperature (MAT) were calculated from the Climate-South America (ClimateSA) software in version 1.12. ClimateSA provides grids of historical interpolated data [12]. A 1 km2 grid was generated for the ACG drainage area, and each cell's centre value was used to extract the annual climate data from 1980 to 2019. Then, cell values were averaged to the boundary of each of the 107 sub-basins to estimate the absolute average MAT and MAP in each sub-basin (Figure 8).

Figure 8. MAP and MAT

Schema of weather data generation for each sub-basin at the drainage area within the ACG. Left: Map of interpolated weather data available in ClimateSA software. Middle: 1 km² grid generated to extract climate information for the ACG's drainage area. Right: Maps of multiannual absolute average values of precipitation and temperature of 107 sub-basins.

RESEARCH QUESTION 2

The land cover data to address the second research question of this project was extracted from the land cover maps generated by Chen [13]. These maps were clipped to the drainage area within the Conservation Area of Guanacaste (ACG). This information provides the area in km² for the years 1979, 1997, and 2015 of the following land covers: Agricultural Lands (AGRL), Deciduous Forest (FRSD), Evergreen Forest (FRSE), Pasture/Hay (PAST), and other land covers, such as urban areas and mangroves (OTHER)ᵃ.


The land cover distribution in the ACG's drainage region was geographically depicted for 1979, 1997, and 2015 (Figure 9). When analyzing the most recent data (2015), almost 65% (1,846.8 km²) of ACG's drainage area is occupied by forests, either deciduous (4.9%) or evergreen (59.7%). Agricultural activities and pastures encompass 13% and 17.9%, respectively. Other (2.8%) and unidentified (1.8%) land covers represent 4.6% (132.5 km²) of the study area.


According to the previous values, the analysis will delve into forest changes as it is the matrix in the ACG's landscape. In addition, it directly impacts water yield generation [8].

a

b

c

Figure 9. Land Cover

Land Cover maps of the ACG's drainage area for the years 1979 (a), 1997 (b), and 2015 (c). The tabulated data summarizes the area in km² and % by the major basins. Colours in the table serve as the legend of the land covers identified in the maps: Agricultural Lands (AGRL), Deciduous Forest (FRSD), Evergreen Forest (FRSE), Pasture/Hay (PAST), Other Land Cover (OHTER), and unidentified land covers due to the presence of clouds (NO DATA).

STATISTICAL ANALYSIS

Linear regression analysis is selected as the statistical tool to answer the two research questions.


The linear regression models will be used to address the first research question, seeking to predict changes over time in both water availability and climatic variables in order to establish whether there is a similar trend that could explain the behaviour of water yield. It is expected to find a statistical significance where the predictor variable (time) generates an effect on the response variables (water and climate).


This statistical analysis is also applied to land cover data, first analyzing the 10 sub-watersheds with the most significant changes in land cover and water availability over the entire monitoring period. It is expected that the difference in water availability at the decadal scale between 1980-1989 - 2010-2019 can be explained by the change between 1979 and 2015 in land cover, mainly forest.


The four major basins, Nicaragua Lake, Papagayo Gulf, Santa Elena Bay, and Tempisque River, have been defined as a covariate to increase the statistical power of this project's regression analyses.

ᵃ This project only focuses on the vegetation land covers. Therefore, land covers classified as "OTHER" are not considered as they, in total, represent less than 3% of the total study area in all the years analyzed.