Produtividade primária bruta na Amazônia legal: relação com variáveis meteorológicas e validação do produto mod17A2

Abstract

The Gross Primary Productivity (GPP) is the rate at the atmospheric CO2 is converted by photosynthetic activity in organic substances and is a measure of the total carbon fixed by the ecosystem. In the context of climate change, the focus for the fixation of carbon by forest ecosystems is the basis to mitigate emissions of anthropogenic CO2. There are several methods for the quantification of carbon stored in the vegetation, using data obtained by micrometeorological towers or by remote sensing. Among these methods, stands out the eddy covariance technique, that is much used to quantify the CO2 flux in many forest ecosystems. However, this technique has a high cost, limited operation and covers a small extent. Accordingly, the methods that employ remote sensing have the advantage of providing terrestrial primary productivity estimates for large areas where the methods at the field level are not feasible. This work has as main objective to evaluate the dynamics of the GPP in the Brazilian Legal Amazon over land use and land cover and weather variables. For understanding the relationship of the eddy covariance GPP with meteorological data, rainfall data from TRMM 3B43 product and environmental variables of flux towers were used. Data from TRMM satellite were validated with data from seven conventional weather stations of Amazonas state. The comparison was based on the Average Error (AE), Root Mean Square Error (RMSE), linear correlation coefficient (r) and Wilmott agreement index (d). We also used the Spearman correlation coefficient and a regression tree model to assess the relationship between flux tower GPP and environmental variables. To evaluate GPP estimates from two versions of the MOD17A2 product, derived from MODIS sensor data, these have been validated with surface data from seven LBA flux towers, of which four are in forest cover, one in transition forest and two in disturbed areas. The statistical analyzes were performed with R software, version 3.1.0. The results of the validation of TRMM 3B43 were positive, with high linear correlation (r = 0.83), high level of agreement (d = 0.85) and an adequate RMSE (59.77 mm), showing that this product can be used as an alternative source of quality data. Regarding the influence of the land use and land cover in GPP, it was found that the disturbed areas had lower productivity compared to the conserved areas and had their water balance affected, presenting higher values of the Bowen ratio. The GPP seasonality was predominantly related to radiation at the top of the atmosphere in forest areas of the equatorial Amazon. In areas more distant of the Ecuador, the GPP was influenced by radiation at the top of the atmosphere and also by rainfall and VPD, indicating limitation both by radiation and by water availability. MOD17A2 had no satisfactory agreement between in estimating GPP compared to the method of eddy covariance, underestimating productivity for most locations studied. The seasonality derived from the MODIS algorithm was only similar to the seasonality of GPP by eddy covariance method for non-equatorial locations. The areas in Equatorial Amazon exhibit distinct MOD17A2 GPP seasonal pattern of that verified by data from the micrometeorological towers. Given this, it is necessary to improve the MOD17A2 algorithm to enable it to estimate GPP depending on the different vegetation responses to drought and radiation. This improved understanding may help to produce better estimates of GPP in Amazon and the use of remote sensing in conjunction with the surface data can contribute to generate an overview of GPP in this biome.