Quantifying past forest cover and biomass changes in the Ecuadorian Amazon

Here, we developed and applied models to quantitatively reconstruct forest cover and biomass changes at three lakes in northwestern Amazonia over the past > 1500 yr. We used remotely sensed data and a modern dataset of 50 Amazonian lakes to develop generalized linear models that predict aboveground biomass, using phytolith morphotypes and forest cover as predictor variables. Also, we applied a published beta regression model to predict forest cover within 200 m of each lake, using Poaceae phytoliths. Charcoal and maize phytoliths were analysed to identify past land use. Results showed forest cover and biomass changes at our study sites ranged between 48-84% and 142-438 Mg ha^-1, respectively. Human occupation was discontinuous, with major changes in forest cover and biomass coinciding with periods of land use. Forest cover and biomass decreased notably after fire (at all sites) or cultivation events (Lakes Zancudococha, Kumpak^a). The timing and ecological impact of past land use were spatially and temporally variable. Our results suggest past human impact was small-scaled and heterogenous in northwestern Amazonia, with a significant impact of fire on forest cover and biomass changes.