Combining a national-scale grid-based hydrological model with an initial-condition climate ensemble: broad-scale analyses of changes in river flows

Abstract Climate change alters the hydrological cycle, with changes in river flows and water availability affecting both people and the environment. The typical way to investigate impacts on flows is to drive a hydrological model with climate model ensemble data, where the ensemble is formed by either perturbing the parameters of a single model or using multiple models. But the presence of natural climate variability in each climate model run makes interpretation of differences difficult. Here, a 40-member single-model initial-condition ensemble (SMILE) is used, where each run only varies due to its initialisation. The ensemble members can be considered to represent the range of possibilities from natural climate variability, without being confounded by other differences. A grid-based hydrological model is driven by the climate SMILE, producing 1km grids of monthly mean flows and soil moisture, annual maximum flows, and 7-day annual minimum flows across Great Britain. Broad-scale analyses show clear changes in seasonal mean, high and low flows, which are less noisy when derived from the pooled ensemble than from any individual ensemble member. In particular, the pooled ensemble enables a robust cluster analysis, grouping river pixels across the country by the flow changes they exhibit. Analysis of extreme high and low flows particularly benefits from the far greater sample within the pooled ensemble and seems more robust when then based on empirical distributions rather than fitted extreme value distributions. Better estimation of flow changes from hydrological models driven by SMILEs enables more robust adaptation planning.