Assessing uncertainties in hydrological response to climate at large scale

The construction of useful models for assessing hydrological response requires parameter estimation from observational data, predominantly precipitation, temperature and discharge time series. A lumped parameter model is applied to two basins with around three decades of such observations to elucidate the uncertainties associated with the simulation of discharge, and hence evaporative losses, at basin scale. Model performance is assessed over a range of historical conditions. This allows prescribed changes in air temperature and basin rainfall to be translated into effects on evapotranspiration and streamflow. The study provides both an indication of the level of uncertainties to be expected and a methodology for assessing response in other basins. Requirements for extending the work to continental scale are data-based and do not depend on major advances in scientific knowledge. WmODUCIION Appropriate models are required to assist with the assessment of macroscale hydrological impacts arising from fluctuations in climatic forcing variables and/or spatially extensive changes in land use. Hydrological models for separating the water balance, including inference of the evapotranspiration component, are also required to assist in studying the global redistribution of solar energy and especially for inclusion in global climate models. Rind et al. (1992) point out that the results from hydrological impact models depend crucially on their representation of the hydrological cycle, and argue that uncertainties in both climate and impact models limit confidence in current assessments. Detailed physically-based models for these purposes would be ideal; it would be convenient if spatially and temporally distributed inputs of rainfall, radiation and other climate variables could be used to drive models which, given spatial descriptions of vegetation, soils, terrain and other important physical catchment descriptors, yield as output the dynamic water (and energy) fluxes to the land and back to the atmosphere. But this requires a thorough understanding of key biological-hydrological interactions of the soil-vegetation-atmosphere system. The dynamics of the hydrological processes in this interactive system are only poorly understood (e.g. IGBP, 1991). Even at the hillslope scale there still exist many problems associated with the use of spatially- distributed models for predicting runoff and evapotranspiration losses from precipitation. These include the estimation and uncertainty problems which accompany