An extreme value analysis of UK drought and projections of change in the future

Abstract

Extreme value analysis of drought indices was used to assess the likelihood of drought over the UK during the 20th century and potential future changes due to increased atmospheric greenhouse gasses. Precipitation indices were derived using 3, 6, 12 and 18 month accumulations, normalised by the 1961–1990 mean and standardized by the corresponding standard deviation. A soil moisture index, similarly standardized, was calculated from monthly soil moisture data. Output from an 11-member perturbed physics ensemble of the Hadley Centre regional climate model (HadRM3), forced at the boundaries by equivalent versions of the Hadley Centre global climate model (HadCM3), was evaluated against available observational-based reference data. Soil moisture reference data were derived from a distributed land surface scheme driven by meteorological reanalysis. Whilst the model ensemble underestimates the variability of the precipitation climatology in the reference data, it accurately replicates the extreme characteristics of dry months for the majority of the ensemble members. Differences in land surface characteristics and the representation of sub-grid scale heterogeneity means that there are significant differences in the representation of soil moisture between model and reference data. However, when HadRM3 is forced at the boundaries by ERA-40 reanalysis data the extreme characteristics of low soil moisture are replicated by the model ensemble indicating that using the regional climate model ensemble to downscale from the coarse resolution of HadCM3 is appropriate. Projections of drought for the 21st century were estimated by applying non-stationary extreme value theory to these monthly drought indices. All drought indices show an overall increase in drought in the future. However, the spread of values is considerable ranging from little change or a slight decrease to a significant increase depending on ensemble member and, to a smaller extent, location. The impact of these projections are put in the context of the notorious UK drought of the summer of 1976. This work provides preliminary steps towards a probabilistic assessment of changes in future drought.