Modelling spatial and temporal variations of annual suspended sediment yields from small agricultural catchments

Estimates of sediment yield are important for ecological and geomorphological assessment of fluvial systems and for assessment of soil erosion within a catchment. Many regulatory frameworks, such as the Convention for the Protection of the Marine Environment of the North-East Atlantic, derived from the Oslo and Paris Commissions (OSPAR) require reporting of annual sediment fluxes. While they may be measured in large rivers, sediment flux is rarely measured in smaller rivers. Measurements of sediment transport at a national scale can be also challenging and therefore, sediment yield models are often utilised by water resource managers for the predictions of sediment yields in the ungauged catchments. Regression based models, calibrated to field measurements, can offer an advantage over complex and computational models due to their simplicity, easy access to input data and due to the additional insights into factors controlling sediment export in the study sites. While traditionally calibrated to long-term average values of sediment yields such predictions cannot represent temporal variations. This study addresses this issue in a novel way by taking account of the variation from year to year in hydrological variables in the developed models (using annual mean runoff, annual mean flow, flows exceeded in five percentage of the time (Q5) and seasonal rainfall estimated separately for each year of observations). Other parameters included in the models represent spatial differences influenced by factors such as soil properties (% poorly drained soils and % peaty soils), land-use (% pasture or % arable lands), channel slope (S1085) and drainage network properties (drainage density). Catchment descriptors together with year-specific hydrological variables can explain both spatial differences and inter-annual variability of suspended sediment yields. The methodology is demonstrated by deriving equations from Irish data-sets (compiled in this study) with the best model efficiency of 0.84 and best model fit of adjusted R² of 0.82. Presented approach shows the potential for regression based models to model contemporary suspended sediment yields in small river systems.