Albedo Trend Analyses in Atlantic Forest Biome Areas

The albedo is an important variable that controls the balance of radiation and energy of the atmosphere, so changes in land cover cause alterations in albedo values, influencing changes in climate behavior at different scales. The goal in this work was to investigate the possible occurrence and causes associated with surface albedo trends within the Atlantic Forest biome (São Francisco de Paula, state of Rio Grande do Sul, Brazil), during the last thirty years (1987-2017), evaluating the impacts of the forest cover structure on albedo trends. The study included images of the TM/Landsat 5 and OLI/Landsat 8 sensors over the period 1987 to 2017. The surface albedo was obtained from the SEBAL algorithm, which includes in its variables the reflectance values of each band, reflected solar radiation and atmospheric transmissivity. The trend analysis was performed by the Mann-Kendall test verifying the existence of significant trends over 30 years. Subsequently, the influence of vegetation greenness on the trend presented by the albedo surface was evaluated. Approximately 92% of the pixels with significant tendency are associated with the decreasing tendency of the albedo. The downward trend was observed with the change from the field to the forest cover, while increasing trends were influenced by the change in forest cover, such as the suppression of individuals from the upper forest canopy. The forest populations in areas of the Mata Atlântica biome had a large participation in the energy balance, which exposed a reduction of approximately 60% of the surface albedo with its implantation, showing its importance for reducing the emission of energy to the atmosphere. The spatial pattern of the trend distribution of the surface albedo is related to the concentration and vigor of the arboreal vegetation.


Introduction
Issues related to climate change have been a constant theme of international discussions over the last decades in order to establish measures aimed at reducing the emission of pollutant gases as well as finding mitigating measures for this problem (Souza & Azevedo, 2012;Global Climate Change, 2018).In this context, forests are relevant due to their potential for storage and capture of pollutants from the atmosphere, and particularly, for the atmosphere carbon storage and sequestration (Nowak et al., 2013;Ni et al., 2016;Planque et al., 2017).In this sense, the arboreal vegetation is particularly a bond between the flow of the terrestrial surface and the atmosphere.Thus, Davin and Noblet-Ducoudré (2010) consider that changes in vegetation cover can interfere in the cycle of flow changes and, consequently, influence the climate, and a variable closely related to the atmospheric radiation balance corresponds to the albedo (Lukes et al., 2016).
For this reason, the surface albedo is one of the most relevant climatic variables, and changes in this property modify the radiation and energy balance of the surface, which can be detected by monitoring this environmental variable (Silva et al., 2016).Planque et al. (2017) points out that the albedo presented by the forest cover represents an important variable to track changes in vegetation, so that changes in vegetation greenery have been shown to influence the albedo values of the surface (Tian et al., 2014).Thus, the Normalized Difference Vegetation Index (NDVI), obtained from the ratio between the reflectance difference in the near infrared band and the red band, evaluates the spectral variability and changes in the vegetation growth.This index is related to the amount and condition of green vegetation (Matos et al., 2015).
The regular observation of a phenomenon provides the reconstruction of the historical context of the land cover evolution on the landscape scale from the time continuity (Zhai et al., 2014).Thus, the orbital images provide information about the terrestrial surface in local, regional and global scale, evidencing possible changes in the variable of interest.The use of time series to estimate the albedo from MODIS or Landsat images can capture the dynamics of the Earth's surface with high similarity with field data collected from the tower installation (Wang et al., 2017).Monitoring of vegetation with remote sensor systems over long periods is fundamental to gain a better understanding of processes related to vegetation change (Yin et al., 2012).
Associated with the information available in time series, trend analysis allows the knowledge of the behavior of a given phenomenon over time.This analysis is often used to identify significant variations in the series of environmental variables such as precipitation (Menezes & Fernandes, 2016;Durães et al., 2016), temperature (Salviano et al., 2016;Wanderley et al., 2016), evapotranspiration (Alencar et al., 2011), vegetation coverage (Neeti & Eastman, 2011;Lukes et al., 2016;Planque et al., 2017), among others.
For Alencar et al. (2014), the trend analysis refers to both continuous and systematic change observed in a time series, which describes the degree of increase or decrease of data over a period.Likewise, Some'e et al. (2012) points out that the presence of trends in time series can demonstrate the behavior of observed data in the light of an environmental phenomenon.For the characterization of tendencies of environmental phenomena, a non-parametric approach is often used, because of their ability to determine how much the slope coefficient of the adjusted line differs significantly from zero (Wagner et al., 2013).
In this sense, the Atlantic Forest of Brazil is one of the richest ecosystems in the world, being composed of forest formations and associated ecosystems (MMA, 2017), contributing significantly to the capture of pollutants from the atmosphere.However, it corresponds to one of the Brazilian biomes with the highest degradation rates, remaining only 12.5% of its original area, when fragments of more than 3 hectares were recorded (SOS Mata Atlântica, 2015) being primordial its monitoring through the years.Aiming at preserving these ecosystems, priority areas for biodiversity conservation were created, in which part of the National Forest (FLONA) of São Francisco de Paula (São Francisco de Paula, state of Rio Grande do Sul, Brazil) is covered by the Atlantic Forest Biosphere Reserve as a Core Area, being considered a region of "high" to "high priority" for conservation (MMA, 2002).FLONA is a conservation unit for sustainable use, which is characterized as an area of native forest cover associated with forest plantations and native field.
Forest plantations have been proposed as a strategy to mitigate climate change, and climate benefits are generally evaluated in terms of carbon sequestration potential, ignoring biophysical processes (Nabuurs et al., 2007).However, the impacts of biophysical variables such as albedo are crucial for analyzing the responses of land cover land use, land use change and forestry (LUCF) on climate (Planque et al., 2017), especially for tropical forests such as Atlantic forest.
Therefore, the goal of this study was to investigate the possible occurrence and causes associated with surface albedo trends within the Atlantic Forest biome during the last thirty years , evaluating the impacts of the forest cover structure on albedo trends.

Spectral Data
The study area corresponds to the FLONA of São Francisco de Paula, which is located between the coordinates 29º27′29.91″at 29º23′20.96″south latitude and 50º24′53, 47″ at 50º22′39.01″ of West longitude.The FLONA belongs to the State of Rio Grande do Sul, in the South of Brazil.The area was monitored over 30 years and the summer period from 1987 to 2017 was observed annually.The spectral data used corresponded to TM/Landsat 5 and OLI/Landsat 8 images, which have a spatial resolution of 30 m.
Based on the TM/Landsat 5 and OLI/Landsat 8 images, the surface albedo was estimated by Equation 1.The albedo was calculated following the Surface Energy Balance Algorithm for Land (SEBAL) model, proposed by Bastiaanssen (1998), in which the atmospheric albedo (α atm ) can be obtained by means of a radiative transfer model, ranging from 0.025 to 0.04 (Allen et al., 2002).The SEBAL model allows to quantify the energy balance using satellite data as an input, being the value of 0.03 recommended for the SEBAL model (Liberato, 2011;Silva et al., 2016).Transmissivity for clear sky conditions, is described by Allen et al. (2002) as shown in Equation 2.
Where, α is the surface albedo; α toa is the albedo at the top of the atmosphere (planetary); α atm corresponds to the atmospheric albedo; τ 2 oc is the atmospheric transmittance.
Where, P o is the local atmospheric pressure (kPa); K t is the turbidity coefficient of air (1.0 is for clean air and 0.5 is for extremely cloudy or polluted air); Z is the zenith angle; W is water precipitation (mm): W = 0.14e a P o + 2.1; e a corresponds to the atmospheric water vapor partial pressure (kPa).

Albedo's Relation to the Vegetation Greenery
In order to relate the vegetation's greenery to the variation of the surface albedo values, the vegetation index NDVI was used.Vegetation indexes explore the spectral properties of vegetation, especially in the red and near infrared regions.The dense vegetation reflects little in the red region due to the absorption of solar radiation by the foliar pigments, while the reflectance in the infrared region is higher due to the scattering of the radiation by the internal structure of the plant cells.Thus, developed by Rouse et al. (1973), the NDVI (Normalized Difference Vegetation Index) encompasses both regions, presenting a range of -1 to +1.This vegetation index is widely known and used for the monitoring of vegetation, since it is related to the vegetative vigor of the same.
Green surfaces have a NDVI between 0 and 1 and water and cloud are usually less than zero.

Trend Analysis
In order to identify the albedo trends on the time series the Mann-Kendall statistical test was employed (Kendall, 1975;Mann, 1945), statistical tool suggested by the World Meteorological Organization (WMO) for the analysis of environmental variables over the time, such as albedo.The Mann-Kendall trend test was applied in the time series on the Landsat images, observing a period of 30 years and considering a 5% significance-level.
The Mann-Kendall test statistic considers the null hypothesis (H 0 ) when the data come from a population in which the random variables are independent and identically distributed.On the other hand, the alternative hypothesis (H 1 ) represents the existence of a monotonic tendency (Vilanova, 2014).
Subsequently, it was carried out the punctual analysis in areas that presented tendency when evaluated the surface albedo.The sample included 40 units (pixels) with increasing and decreasing trends.In the same way, the NDVI values were observed in these points in order to relate the two variables.For these samples was used the analysis of the standardized statistic of Mann-Kendall.

Tilt Test of Sen
The Mann-Kendall test does not provide the magnitude of the detected trends.Thus, the slope estimator proposed by Sen (1968) was used in combination with the trend analysis indicating the data slope degree of either increasing or decreasing albedo trends.The trend analysis was developed in programming language R version 3.4.3(R Development Core Team, 2017).However, Qgis software version 2.16.3 was used to cut the area of study and conversion of the projections of the images.

Results and Discussion
After the albedo analysis over 30 years the trend test was applied, looking for patterns in the albedo behavior that occurred during the observed period.For each pixel, the Mann-Kendall test was performed considering the summer series from 1987 to 2017.The Figure 1 illustrated the points at which the null hypothesis of no trend in the series was rejected at the significance level of 5% for the albedo of surface.
Significant albedo changes are indicated in blue and red, corresponding to decreasing and increasing trends, respectively.The adjustment of a simple linear model to the annual surface albedo values showed a predominance of non-significant variations (p > 0.05) for the FLONA of São Francisco de Paula.
The study 3,597 pixe pixels, 3,3 294 pixels showed a d In a study the Modis 94% of th precipitatio trends.

Table 1 .
A