Google Earth Engine for analysis of land cover change and streamflow trends in the Ribeirão da Caveira River Basin, Brazil from 1988-2019

Abstract

Cloud computing platforms for processing satellite imagery will boost our understanding of relationships between land use/cover, precipitation, and streamflow, thereby providing crucial information for water management policies. In this article, Google Earth Engine (GEE) was used to process and assess the impact of land use and land cover change (LULC), forest biomass, and precipitation on streamflow of the Ribeirão da Caveira River Basin (RCRB), a basin located in the Brazilian semi-arid region, from 1988–2019. Land use land cover maps comprised six classes: forest, natural non-forest formation, agriculture, pasture, water bodies, and bare soil. In addition, the following spectral indices were calculated: normalized difference vegetation index (NDVI), modified soil-adjusted vegetation index (MSAVI), and bare soil index (BSI). NDVI was used to estimate forest biomass. The data were fitted to multiple linear regression models using streamflow trends as a target variable and principal component analysis was used to further interpret the data. The analyses revealed agriculture and forest classes had the largest extension within RCRB. Changes in forest biomass had no apparent effect on streamflow. Furthermore, the results showed both precipitation and bare soil areas were the most important factors affecting streamflow, and best-fit models showed moderate predictive power.