2. Methods
2.1 Study areas
The research took place in two forested semiarid areas (Figure 1). The Caatinga experimental area is located inside the Aiuaba Ecological Station (ESEC, 117 km²), a Federal preservation area located in the Brazilian Northeast. This space has been fully preserved since 1978 and monitored since 2003 by the Semiarid Hydro-Sedimentological Research Group (http://www.hidrosed.ufc.br/index.php).
The Caatinga experimental area is part of the Jaguaribe River basin. According to the Köppen classification, the local climate is Bsh (hot semiarid) with potential annual evaporation of 2,500 mm, average annual precipitation of 549 mm, and average annual temperature of 26 ºC (Pinheiro et al., 2016). The rainy season is concentrated from January to April. The predominant lithology is the crystalline complex of metasedimentary formation. Soils in the region are generally shallow and originated from a crystalline substrate. The predominant soil types are Luvisols, Latosols, and Argisols (de Araújo & Piedra, 2009; Farias, Medeiros, Navarro-Hevia, & de Araújo, 2019).
The Pinares forest (106 km2) is located in Tierra de Pinares, in the province of Valladolid, Spanish Northern Plateau. Local vegetation there is formed mainly by stone pine (Pinus pinea L), but the maritime pine (Pinus pinaster ) and holm oak (Quercus ilex ) can also be found (Bello, 2004). The soil is deep and sedimentary, with a sandy texture (more than 90% sand) and low water-holding capacity (Calama et al., 2019). Pinares forest has a continental Mediterranean climate which, according to the Köppen classification, is Bsk (cold semiarid). The annual rainfall is 400 mm, the annual mean temperature 11 °C (Moreno-Fernández, Montes, Sánchez-González, Gordo, & Cañellas, 2018) and the potential annual evapotranspiration 1,100 mm (Vicente-Serrano et al., 2014). In the area, the confluence of the rivers Pisuerga and Duero takes place.
2.2 Satellite Imagery and Meteorological data
Actual evapotranspiration was estimated for the two experimental areas by using the imagery of Landsat 5 and Landsat 8. Landsat 5 captured images from March 1984 to January 2012 on six 30-m visible bands and a 120-m thermal band. Landsat 8 was launched in February 2013 and is still operating. Landsat 8 OLI images have eight 30-m bands, two 100-m thermal bands and a 15-m panchromatic band. In this study, all the available cloudless images between 1995 and 2019 were used (Figure 2): 37 images of the Caatinga forest and 74 of the Pinares forest, totalling 111 images. The imaging time was 12:00 UTC (path 217 and row 65) for Caatinga, and 11:00 UTC (path 202 and row 31) for Pinares.
The Caatinga forest usually is excessively cloudy during the rainy season, so that practically no image was cloudless. During the rainy months, were used images with up to 10% cloud coverage, but the specific areas covered by clouds were disregarded.
Shuttle Radar Topography Mission (SRTM) images were employed to generate the Digital Elevation Model (DEM). Both Landsat and SRTM data were provided by the United States Geological Survey (USGS) (https://earthexplorer.usgs.gov/).
Daily and hourly field data of wind speed, temperature, relative humidity, solar radiation, air pressure and precipitation were obtained from the National Institute of Meteorology (INMET) and the Ceará Foundation of Meteorology and Water Resources (FUNCEME) for the Brazilian forest, and from the State Meteorological Agency (AEMET) for the Spanish forest. Table 1 shows the stations and periods of available data.
In the Caatinga forest, meteorological data were obtained from one of the four stations close to the area, depending on data availability on the analysis day (Figure 1): the BEA station is located inside the experimental area and provided 27% of the data; IBAMA, 8%; Campos Sales, 62%; and Tauá, 3%. In the case of the Pinares forest, data concerning air temperature, wind speed and relative humidity were used from the Villanubla station, which is located outside the city of Valladolid and suffers less influence of urbanization effects. All other data were provided by the Valladolid station, which is closest to the area.
2.3 Surface Energy Balance Algorithm for Land (SEBAL) model
The SEBAL model, applied in this work for having been used with success in many scientific research, calculates the components of the surface energy balance (Liaqat & Choi, 2015), so that the ET for each image pixel can be estimated using Eq. 1 (Bastiaanssen, Menenti, Feddes, & Holtslag, 1998):