Constraining the response of the hydrological cycle, land surface and regional weather to global climate change.

HYDRA will investigate the sensitivity to, and uncertainties in, rainfall, evaporation and river runoff to changes in land use and the carbon cycle by comparing models with observations from the last 50 years.

Analysis of the HYDRA experiment will focus on the influence of the land on climate. We want to know how sensitive rainfall, evaporation and river runoff are to changes in land use and the carbon cycle. This is the very first experiment where these land uncertainties will be explored.

In the experiment each model has a different map of land use combined with carbon cycle and atmospheric parameter perturbations. For each land grid cell in the model the percentage of needleleaf and broadleaf forest, grassland, shrubs, urban and bare soil are given. These percentages are based on an observed dataset. With these maps the below figure was made showing the decrease of forest cover between 1850 and 2000. With the experiment we want to investigate how important such changes are relative to other model uncertainties.

We hope to answer the question: what impact did pre-industrial land use change have on rainfall? How important are differences between land cover reconstructions for our understanding of hydrological change? To do this, models with time evolving land use from different datasets will be compared with models where the land use is kept constant over time.


The aim of this project is to examine much more exhaustively than has previously been possible the climate uncertainties due to uncertainties in how to model land surface effects. This new experiment will be the first combining atmospheric uncertainties with all aspects of land surface uncertainties, including model parameter perturbations and the spatial distribution of land use.


Many of the most important impacts of global warming will probably not be the actual changes in temperature, but the associated changes in hydrology. A warmer atmosphere can contain much more moisture, which we expect to make flash flooding more likely. But the global-mean rainfall is thermodynamically constrained to increase more slowly, so we expect a general tendency to increased drought as well. Yet we cannot constrain any of this well on the more local scales where it matters. This is due to the fact that the spatial scales of precipitation are much smaller than of temperature. Precipitation is much more dependent on smaller-scale circulation and topography. Therefore the agreement between different General Circulation Models (GCMs) is correspondingly poorer.

Within HYDRA this uncertainty in the hydrological cycle on both the global and local scales will be explored. In addition to perturbing the atmospheric model parameters as in previous experiments, the land surface parameters and spatial map of the land use will be perturbed. These maps describe where certain vegetation types occur, such as forest and grassland. This is significant, because the variation between available land use datasets derived from observations is large. Deviations between datasets are a result of different classification and aggregation methods. Only a few studies have explored the impact of these maps on the climate system by comparing a small number of models. With a large ensemble of models this can be explored in much more detail. The differences between ensemble members will be most noticeable on the regional scale.

Uncertainties in the hydrological cycle due to land surface parameterizations can be divided into uncertainties from the spatial distribution of vegetation and from the model parameter values. These parameters regulate processes such as plant carbon uptake by photosynthesis and how precipitation is separated into evapotranspiration to the atmosphere and river runoff. This parameter uncertainty has been studied before with a much smaller coupled model ensemble that could only be run on a supercomputer. From this ensemble it became clear that the climate uncertainties due to the land parameters are as large as from the atmospheric parameters.

The initial focus of this experiment will be on the impact of the time evolution of land cover on the hydrological cycle. What impact did pre-industrial land cover change have on the hydrological responses? How important are current differences between land cover reconstructions for our understanding of hydrological change? Models with a time evolving land use from different datasets will be compared with models where the land use is kept constant over time.

Experiment Design

HadAM3PM2 will be used, which has been set up especially for this experiment. This model is comparable to HadCM3, but there is no dynamic ocean (instead prescribed sea surface temperatures are used). The major difference with HadCM3 is the land surface model (MOSES2), which interacts more dynamically with the atmosphere. It contains a much more detailed coupled carbon and water cycle of the land surface, which means that for instance it is possible to study the effects of an increasing atmospheric CO2 concentration on the hydrological cycle.

For this experiment 50,000 different models will be run, which is only possible with your help. These consist of 10-year runs spanning the years 1959 to 2002, with an overlap of two years to account for the spin up of the moisture in the soil, which is an important variable in this experiment. The differences between the models are the atmospheric and land surface parameter values and the land cover maps.


HYDRA is part of the NERC CWC (Changing Water Cycle) program. The project involves scientists from CEH, University of Exeter, Met Office and the University of Oxford.