home models publications

---

Models

Rainsim

Rainsim is a model designed to make rainfall projections both for current conditions and taking account of possible future climate. Based on the Spatial Temporal Neyman Scott Rectangular Pulses (STNSRP) stochastic approach the model is an improvement on existing models and is a practical tool for predicting rainfall for small catchments ( < ~5000 km² ).

The STNSRP process

1. Storm origins arrive as a Poisson process in time;

2. each origin generates a set of circular raincells whose centres follow a spatial Poisson process and with exponentially distributed radii;

3. each raincell follows the storm origin after an exponentially distributed time interval;

4. raincell duration and intensity are each exponentially distributed and uniform throughout the disc and lifetime;

5. the rainfall intensity is the sum of active raincells scaled to account for orographically enhanced rainfall amounts. The random variables are independent and described with six parameters. These vary by calendar month, providing an annual cycle, and are fitted by numerical minimization of the difference between expected and required rainfall statistics. The parameterized model may then generate synthetic rainfall fields of arbitrary duration.

Improvements

Using the Dommel, Brenta and Gallego catchments, three calibration improvements have been made in version 3 of Rainsim making it a sigificant improvement on previous STNSRP models

• the robustness and efficiency of the fitting algorithm was improved through a better choice of numerical optimizing scheme;

• the objective function was changed, providing a better match to the required rainfall properties;

• fitting biases for proportion of dry days and hours statistics were modelled and corrected.

Improved modelling of mountain rainfall

An illustration of raincells simulated during a storm sampled at five locations, X, within a catchment (red line). The raingauges at greater elevation experience a greater frequency of rainfall events.

Simulated spatially varying rainfall occurrence compared with five Gallego raingauges (a separate colour for each).

The traditional STNSRP process can simulate spatial non-stationarity in rainfall intensity. A new model variant also allows the probability of simulated raincell occurrence to vary in space, improving simulation of rainfall strongly influenced by elevation. This approach successfully simulated spatial-temporal rainfall for the Gallego catchment in the Pyrenees (see Burton, A., Fowler, H.J., Kilsby, C.G., and O’Connell, P.E. The Non-Stationary Activation of Raincells (NSAR) model: a modified Neyman-Scott Rectangular Pulses model for spatially non-stationary rainfall occurrence. In revision for Water Resources Research).

EARWIG

The Environment Agency Rainfall and Weather Impacts Generator (EARWIG) is a user friendly application containing UKCIP02 scenarios (50km resolution), meteorological data (5km resolution) and stochastic downscaling to provide consistent synthetic daily rainfall, weather and PE time series for a control (1961-1990) and for future (2020s, 2050s and 2080s) scenarios. The downscaling combines CF and SF approaches to calibrate a single-site daily rainfall model and a conditioned auto-regressive weather generator. Stationary climate simulations then correspond to the selected decade.

UKCP09

The EARWIG approach has been extended and incorporated into the latest UK Climatic Projections, directly providing scenarios appropriate for hydrological investigations.
New features include:

• model ensembles derived from a range of GCM and RCM projections
• a range of decadal time slices
• probabilistic climate projections
• hourly time step
• 5km grid cells across the UK
• web access