An assessment of the MOD17A2 gross primary production product in the Caatinga biome, Brazil

Studying the dynamics of gross primary production (GPP) in seasonally dry tropical forests is of fundamental importance to understand the carbon dioxide (CO2) balance in this ecosystem, helping mitigate its potential impacts at the regional and global levels. Thus, the objective of this work was to evaluate the accuracy of GPP estimated via remote sensing in the Caatinga biome. A set of observed data retrieved from micrometeorological towers equipped with eddy covariance systems were used to validate remote sensing data. The set was measured in a preserved Caatinga fragment. Remotely sensed GPP data was retrieved from the MOD17A2 version 6.0 product of the Moderate Resolution Imaging Spectroradiometer (MODIS) sensor onboard the Terra satellite. The validation of MOD17A2 GPP estimates was carried out through the comparison with micrometeorological data measured in situ. In the Caatinga site the comparison between the two GPP data types showed a moderate correlation with Pearson’s correlation coefficient (r) = 0.65 and coefficient of determination (R2) = 0.43 and the product performed better in representing GPP in the Caatinga during the dry season. Results showed that although the MOD17A2 product represents the annual behaviour of GPP, the algorithm could be improved in order to provide GPP information that is more similar to surface measured data over these land covers.