Estimation of Energy Flux and Biomass in Pasture Areas through Remote Sensing Techniques

Pasture production is estimated through remote sensing techniques with the aid of models and algorithms. The application with no need for extensive field measurements is one of the advantages of the Surface Energy Balance Algorithm for Land (SEBAL). The objective of this work was to estimate energy fluxes and, subsequently, pasture biomass with the aid of remote sensing techniques. The study area is located on the Experimental Farm of Embrapa Beef Cattle, municipality of Campo Grande, State of Mato Grosso do Sul, Brazil. For the implementation of the SEBAL and estimation of energy fluxes and biomass of the pasture areas, meteorological data and Landsat 5 TM image were used. It was found that the technique has the potential to be applied to indicate the forage availability and to support decision-making in the planning and management of the extensive production of beef and milk cattle, with economic and environmental sustainability of pasture areas. Keywords— geotechnology, livestock, SEBAL, sustainability, rural planning.


I. INTRODUCTION
The growing demand for food further reinforces the importance of increasing production with economic and environmental sustainability. Vilela et al. [1] have reported that intensifying and enhancing the efficiency of production systems may help to harmonize these interests. Hence, the sustainable use of pastures is a strategic issue, since most pasture areas are show some indication of degradation. The recovery of these areas may reduce the pressure for the opening of new farming and livestock frontiers, as well as contributing to diminishing the emission of greenhouse gases [2]. According to Dias-Filho [3], the support capacity would be the most flexible indicator to quantify the degradation of a given pasture. Currently, the only official parameter that is related to support capacity is the pasture stocking rate. This parameter has been presented by DIEESE [4], and it is the result of the division of the number of animals by the area occupied by pastures of a given geographical unit. However, Dias-Filho [5] warns that a priori, it is not possible to guarantee the degradation condition of pasture only by evaluating its instantaneous support capacity (maximum number of animals supported by pasture, with no harm to pasture and to the animal). Studies such as that carried out by Grigera et al. [6] highlight the potential of remote sensing techniques as a tool to assist in the implementation of systems for monitoring pasture production that makes it possible to identify, for example, the areas that require the adoption of acceptable management practices. Thus, models and algorithms are used to estimate pasture production using remote sensing techniques. One of these algorithms is the SEBAL (Surface Energy Balance Algorithm for Land) that was developed by Bastiaanssen et al. [7,8]). One of the advantages of SEBAL is the flexibility in its structure so that other models can be coupled [9], facilitating the applications in studies carried out at a local and regional scale, with no need for extensive field measurements (Andrade et al. [10]). In order to estimate the above-ground biomass, Bastiaanssen and Ali [11] and Samarasinghe [12], among others, obtained good results by coupling the model of biomass accumulation proposed by Monteith [13] in the SEBAL associated with the model of efficiency use of the radiation that was structured by Field et al. [14]. As a result, the objective of this study was to estimate energy fluxes and, therefore, pasture biomass through the application of remote sensing techniques.

II. MATERIAL AND METHODS
The study area is located in the Experimental Farm of Embrapa Beef Cattle, municipality of Campo Grande, state of Mato Grosso do Sul, Brazil ( Figure 1). According to the climatic classification of Köppen, the region is situated in a transition zone between humid temperate climate with hot summer (Cfa) and tropical climate with dry winter season (Aw). The average annual temperature is 22.8°C, and the average annual rainfall is around 1,500 mm. The months with the lowest rainfall are June, July, and August. The soil in the study area is Red Latosol Dystrophic class whose characteristic is a clayey texture, acid pH, low base saturation and high aluminum concentration [15]. The algorithm SEBAL was applied by using the data to estimate the components of the energy fluxes from pixel by pixel. In this case, the latent heat flux component was estimated as a residue of the other components of the energy balance, according to equation [7]: Where LE is the latent heat flux, Rn is the radiation balance, H is the sensible heat flux, and G is the soil heat flux, all in W m -2 . The radiation balance (Rn) was the first variable of the energy balance equation to be obtained. In this case, the equation suggested by Allen et al. [9] was applied: Where, Rs↓ is the short-wave incident radiation (W m -2 ), RL↓ is the longwave radiation emitted by the atmosphere in the direction of the surface (W m -2 ), RL↑ is the longwave radiation emitted in the direction of the atmosphere (W m -2 ) and ɛo is the emissivity of the surface (dimensionless), α is the surface albedo (dimensionless). Allen et al. [9] show in detail the procedures involved in obtaining Rn. After the calculation of Rn, the empirical equation suggested by Bastiaanssen [16] was used to estimate the soil heat flux (G, W m -2 ), given by: Where, Ts is the surface temperature (°C), α is the surface albedo (dimensionless), NDVI is the Normalized Difference Vegetation Index and Rn is the radiation balance. In order to correct the soil heat flux values for water bodies (NDVI < 0), G = 0.3Rn was considered [17].
Once the value of G was obtained, a new series of steps was started to obtain the sensible heat flux (H). For the estimation of H, the process started by considering the neutral atmosphere condition. Firstly, the expressions suggested by Allen et al. [9] to obtain the initial roughness parameter (zom initial), the initial friction velocity (u*initial), the wind velocity at a height (z) of 100 m (blending height), where it is assumed that the surface roughness effects are negligible) and the initial aerodynamic drag (rah initial). The roughness parameter (zom) obtained as a function of SAVI (Soil Adjusted Vegetation Index) was used in the following steps. SEBAL uses two pixels termed "anchor pixels" to set boundary conditions for the energy balance. These pixels are denominated "hot" and "cold" and located in the study area. The "cold" pixel can be selected on a wellirrigated crop surface that completely covers the soil with the vegetation or the surface of a pond. In this case, the "cold" pixel was selected on the surface of a lake. The air temperature near the surface and the surface temperature are considered equal for that pixel. Thus, zero value was assumed for the sensible heat flux (H) and the maximum latent heat flux was determined. The "hot" pixel was selected in a dry farming field, with soil exposed or without vegetation, where the zero value was considered for the latent heat flux (LE) and thus, maximum sensible heat flux was obtained [9,18]. The correlation coefficients a and b were used with the aid of the anchor pixels to obtain dT in each pixel. Because in the cold pixel dT = 0, that is, there is a system with two equations and two unknowns, which enabled the calculation of a and b and then the initial sensible heat flux (Hinitial) was obtained. The next step was to consider the condition of atmospheric stability, making corrections in the values of H in the iterative process. Therefore, the Monin-Obukhov similarity theory (L, in m) was applied to know the stability condition of the atmosphere, that is, whether it is unstable (L <0), s table (L> 0) or neutral (L = 0). In sequence, through the formulations suggested by Allen et al. [9]), it was possible to obtain the values of the stability corrections for the transport of momentum (ψm) and sensible heat (ψh) and the friction velocity (u*) was estimated considering the atmospheric condition [9,18]: By using the corrected value of u*, the aerodynamic drag (rah) corrected for the stability conditions of the atmosphere was obtained: After that, the temperature difference function (dT) was recalculated by repeating the procedures mentioned above until stability in the successive values of dT and rah for the hot pixel was observed. Finally, the latent heat flux (LE) was obtained as a residue of the classical equation of the energy balance. The components of the energy balance were used to estimate the evaporative fraction (λ) [ [14]. By the evaporative fraction (λ, [19]) it was possible to estimate the efficiency of the use of radiation (ɛf) by: Where, ɛf * is the maximum efficiency of the use of the radiation, equal to 2.5 g MJ -1 [20,21]; and T1 and T2 are temperature scales. Afterward, the above-ground plant biomass was obtained through the model proposed by Monteith [14] and applied by Bastiaanssen and Ali [11] and Teixeira et al. [22]: Where Bio is the plant biomass in the period t (kg m -2 ). Figure 2A shows the estimated albedo map for the pasture areas of the Experimental Farm of Embrapa Beef Cattle. It is observed that the albedo was higher than 0.35 in pixels of the image highlighted in orange and red, probably in places with soil exposure, dry vegetation or post-grazing period (residual grass). On the other hand, albedo values in green shades, ranging from 0.21 to 0.28, are located in areas of wet (darker) soils or better vegetation vigor. Andrade et al. [23] obtained, for cultivated pasture area, an average albedo value of 0.225. However, in places of soil exposure or dry vegetation, albedo values between 0.30 and 0.40 were observed. In a study of Moura et al. [24], in pasture areas in the Amazon region, values of average albedo hours of 0.197 and 0.204 were observed for the rainy and dry seasons, respectively.

International Journal of Advanced Engineering Research and Science (IJAERS)
[

Fig. 2. Visualization of the maps generated from the application of the SEBAL algorithm and TM -Landsat 5 to estimate: (A) albedo (α, dimensionless), (B) radiation balance (Rn, W m -2 ), (C) soil heat flux (G,W W m -2 ), (D) sensible heat flux (H, W W m -2 ), (E) latent heat flux (LE, W W m -2 ) and (F) evaporative fraction (λ, dimensionless) of the pasture areas of the Experimental Farm of Embrapa Beef Cattle, Campo Grande, state of Mato Grosso do Sul, Brazil.
The radiation balance (Rn) of pasture areas, in most of the area, varied between 610 and 650 W m -2 ( Figure 2B). However, significant spots of Rn ≤ 600 W m -2 were found in pasture areas with a predominance of exposed soil or water deficit and values of Rn > 660 W m -2 in areas of dense vegetation and with no water restrictions. In a study by Santos et al. [25] in the Amazonian biome using MODIS images and SEBAL algorithm combined with data of micrometeorological tower in the pasture area, obtained Rn values for the geographic location of the tower varying from 475.10 W m -2 to 599.44 W m -2 .
In relation to the soil (G, Figure 2C), sensible (H, Figure  2D) and latent (LE, Figure 2E) soil heat fluxes, it was found that pasture areas with water deficit or soil exposure showed, as expected, higher components of G and H and lower for LE. However, an inverse condition was observed when analyzing these components of energy fluxes in pasture areas with dense vegetation or with no water restrictions.  Figure 2F shows the evaporative fraction (λ, dimensionless) estimated for pasture areas. It is observed that the greater evaporative fraction of pasture areas is represented in shades of blue (λ > 0.70). According to Figures 3A and 3B, vegetation indices (NDVI ≥ 0.5 and IAF ≥ 2.0 m 2 m -2 ) indicate good vegetation development in these areas. For pasture areas , Rubert et al. [28] observed an annual average value of the evaporative fraction of 0.71, with minimum and maximum values of 0.54 and 0.89, respectively. Figure 3C shows the map of the biomass availability estimate (Bio, kg/ha) of the pasture areas after the application of the SEBAL algorithm and TM -Landsat 5 image. It is observed that on pastures with the predominance of exposed soil (brown class) the Bio was below 1000 kg ha -1 . On the other hand, in some pasture areas, represented on the map in shades of blue, the Bio was above 7000 kg ha -1 . Barbosa et al. [29] obtained a biomass availability of Panicum maximum cv. Mombaça of 7200 kg ha -1 in the summer and 2400 kg ha -1 in winter.
Carnevalli [30] analyzed the availability of this forage in pre-and post-grazing submitted to combinations of grazing intensities and frequencies in rotational stocking observed variation between 4300 and 8900 kg ha -1 in pre-   Figure 4 shows the comparison of the average biomass availability data observed by Barbosa [15] together with the average data estimated with the aid of SEBAL and TM -Landsat 5 images for experimental pasture area in 15 pickets with Tanzania grass in 6 treatments (25/100; 25/95; 50/100; 50/95 and 50/90) in intensity combinations (residues from 25 to 50 cm) and frequencies (90, 95 and 100% light intercepted by the canopy) of monitored defoliation based on the predetermined condition of the canopy structure over the evaluation period. The average value observed for the 15 pickets with Tanzania grass was 4009 kg ha -1 . However, the average value was estimated at 3689 kg ha -1 , with an average difference between the observed and estimated values of 320 kg ha -1 (8.68%) ( Figure 4). Figure 5 shows that the estimated average values had a coefficient of determination (R 2 ) of 0.7108. In this case, the adjustment between the observed and estimated average data is considered reasonable.  It should be emphasized that, in complementary studies, it is interesting to use high spatial resolution images (such as Geoeye, Rapideye, Worldview, among others) for estimating the biomass and thus to evaluate the adjustment between the observed and the estimated data. However, since these images do not have a thermal band for SEBAL application, the SAFER (Surface Algorithm For Evapotranspiration Retrieving) algorithm appears as one of the alternatives [22]. It is worth mentioning that there is also the possibility of using Unmanned Aerial Vehicle (UAV), which can be boarded with cameras for the imaging of surface targets in visible (RGB), nearinfrared (NIR) and thermal bands, generating data in which the algorithms and models can be later applied to extract information of the targets of interest with very high spatial and temporal resolution, for example, in the estimation and monitoring of the biomass and h ydric conditions of pastures on a picket scale.

IV. CONCLUSION
In general, it is concluded that the application of remote sensing techniques associated with models and algorithms, appears as a relevant alternative for estimation of biophysical parameters and indicator of forage availability in different time and space scales. These geotechnical tools also stand out for the possibility of performing several types of monitoring that can assist the farmer in making decisions regarding the planning and management of the extensive production of beef cattle and milk, taking into account, for example, analyses of indicators related to the economic and environmental sustainability of pasture areas .