Problem Description

The study explores the potential impacts of climate change upon flood frequency for the gauged, Lossie catchment in the northeast of Scotland, UK using the UKCIP02 climate change scenarios.

(Full article: "An application of the UKCIP02 climate change scenarios to flood estimation by continuous simulation for a gauged catchment in the northeast of Scotland, UK (with uncertainty)", by David Cameron, Journal of Hydrology, 2006)

Available data

UKCIP02 climate change scenarios: mean temperature and precipitation (monthly change at 50km spatial resolution for all four emissions scenarios for the 2080s time-slice).

14 years observed continuous hourly rainfall and flow data.

46 years observed annual maxmimum (AMAX) flood data.

Used Software

TOPMODEL, stochastic rainfall model, GLUE (all in FORTRAN).

TauDEM? digital terrain analysis software (for TOPMODEL's topographic index; operates in ESRI ArcMap?).

Methodology

The UKCIP02 scenarios datasets were used to modify a series of continuous two thousand year simulations with hourly timestep, created using a stochastic rainfall model and used to drive a rainfall-runoff model (TOPMODEL), to estimate the potential impact of climate change on flood frequency for the Lossie catchment.

Initially, TOPMODEL was run using the stochastic rainfall model under current climatic conditions (the stochastic rainfall model being based upon a 14 year observed hourly rainfall record). The model output was compared with the observed average hourly annual maximum (AMAX) flow. This approach was deemed to give a satisfactory simulation of current climate and flooding frequency. To estimate the potential effects of climate change, the stochastic rainfall model hourly time-series were altered based on the UKCIP02 climate change scenarios. Since it is not currently possible to assign probabilities to the UKCIP02 climate change scenarios, all four were considered. In addition, the “High Emissions” uncertainty margins were used to derive two additional climate change scenarios: the “H-Dry” scenario (the largest increases to temperature change and the largest decreases to rainfall available from the ‘‘High Emissions’’ scenario uncertainty margins) and the “H-Wet” scenario (the largest decreases to temperature change and the largest increases to rainfall available from the “High Emissions” scenario uncertainty margins).

In order to perturb the TOPMODEL/stochastic rainfall model simulations, estimated percentage changes to both rainfall and potential evapotranspiration (PET) were required from each climate change scenario. The perturbation to rainfall was achieved by uniformly applying the monthly percentage change to rainfall (for each climate change scenario) to all rainfall amounts within that month. Changes to PET are not directly available in UKCIP02, and were instead estimated from mean monthly temperature.

Upon completion of the climate change simulations, the Generalised Likelihood Uncertainty Estimation (GLUE) framework was used to weight TOPMODEL outputs, allowing cumulative distribution functions (cdfs) to be calculated for AMAX using the scaled likelihood weights and discharge estimates. The example below shows the 1 in 200 year return period flood simulated under current conditions, and for all six climate change scenarios.

Results

It was demonstrated that, while flood magnitude changed under all six of the climate change scenarios considered, the magnitude and direction of that change was dependent upon the choice of scenario. An overlap between the likelihood weighted uncertainty bounds estimated under the conditions of the current climate and those estimated under the four UKCIP02 scenarios and the ‘‘H-Dry’’ scenario was also observed. These findings highlight the need to consider multiple climate change scenarios and account for model uncertainties when estimating the possible effects of climate change upon flood frequency.