Dynamics of sediment discharge in relation to land-use and hydro-climatology in a humid tropical watershed in Costa Rica

Hydrology in humid tropical regions is often characterized by considerable natural variability and uncertainty. Hydrologic and land-use data from the Terraba basin in Costa Rica are used to analyze dynamics in sediment discharge processes during the period 1971–1993. Time-series of log transformed sediment concentration and flow are analyzed with a Bayesian dynamic linear regression model (DLM) to detect changes in the sediment–flow relationship over time. Annual time-series of estimated sediment discharge based on the DLMs were regressed against various hydro-climatic variables based on the Southern Oscillation Index (SOI), rainfall, stream flow and rainfall erosivity. Hydro-climatic variables such as rainfall, stream flow and rainfall erosivity were also regressed against SOI. The results from the regression models, the presence of trends in the DLM slope and other hydro-climatic parameters for each sub-basin were used to compare the hydro-climatic and sediment response characteristics of the sub-basins. Over 60% of the variability in hydro-climatic variables like rainfall, rainfall erosivity and stream flow is explained by SOI but less than 40% for sediment discharge measured at the basin mouth. The clustering of stations and sub-basins with respect to sediment discharge responses is different from that based on hydro-climatic characteristics. Stations or sub-basins corresponding to areas with large rates of deforestation and/or agricultural intensification and earthquake related soil disturbance tend to fall in one cluster. These sub-basins are characterized by an increase in the sediment–flow regression slope over time, and weaker relationship between sediment discharge and hydro-climatic variables such as SOI, rainfall and flow. In sub-basins undergoing rapid land-surface changes, variability in sediment supplies accounted for an estimated 50–90% of the variability in annual sediment discharge. The methods described in this paper can be used to analyze many existing hydrologic time-series to detect land-use effects on watershed hydrology.