Abstract

The influence of forest cover on dry-season stream flow in the humid tropics has been of enduring interest to eco-hydrologists and policy makers. We used data from eight nested gauging stations in the Terraba river basin (4774 km²) of Costa Rica to demonstrate how a Bayesian regression approach could be applied to test hypotheses on “infiltration-evapotranspiration trade-offs”, between the “sponge effect” and evapotranspiration demands of forests on dry-season flows. We summarized three competing hypotheses as informative prior distributions for regression slopes as 1) positive, 2) zero, and 3) negative effect of forest cover, and combined them with dry-season flow data (response), and annual rainfall and per cent forest cover (as covariates) to obtain posterior estimates. The time series (1962–2002) included a deforestation phase until the 1980s and subsequent forest recovery from the 1990s. The sign, magnitude, and uncertainty of posterior slopes showed that forest cover effects on dry-season flow were generally positive or neutral, influenced by both forest loss and recovery. Forest cover effect on dry-season flow for the entire basin was strongly positive. Four sub-catchments showed weakly positive and two showed neutral effects, and only one indicated a weakly negative effect. Mean effect size of forest cover diminished with increase in basin size. Our results predict a gain of 3.3 mm in dry-season flow for every 1% increase of forest cover, corresponding to 48 km² of basin area. Enhanced infiltration and wet-season groundwater recharge dominating evapotranspiration can be attributed partly to deep soils and cloud forests. We demonstrate how informative prior distributions can represent different hypotheses in Bayesian regression models offers advantages in investigating complex hydrological processes. The rigorous evidence for positive and neutral effects of forest cover on dry-season flow, reported by our study, potentially contributes to ecohydrology-based management of tropical landscapes undergoing rapid change.