3.1 Weather and Climate Forecast Modelling
The terrestrial biosphere, atmosphere, and oceans are integrated in an Earth system characterised by variability of controlling parameters at a wide range of time and space scales. The variability and memory in this system are due to, among others, the cycling of water in different phases and storage compartments. Small fluctuations in initial forcing conditions may be amplified by feedback mechanisms. Although the land fraction of the Earth is relatively small (~30 percent), its distribution in large contiguous areas and its distinctive hydrothermal inertia cause significant variations in regional climatic systems.
There is evidence that on weather- and storm event- time scales, initial soil conditions can reinforce the development of precipitating weather systems. At the regional scale, soil moisture availability has substantial influence on elevated mixed layers and on associated "lids" on atmospheric instability that focus the release of convective instability and hence determine the distribution of the regional precipitation in time and space (Clark and Arritt, 1995). For example, using numerical mesoscale atmospheric models, the evolution of summertime weather systems in the Midwestern U.S. was found to be critically dependent on so-called "dryline" conditions where sharp gradients in soil moisture are present (Chang and Wetzel, 1991). Hydrological processes with long memory, such as soil moisture, may serve to integrate past atmospheric forcing and enhance prediction skills for regional climates (e.g., Fennessey and Shukla, 1999). - The presence of feedback mechanisms can enhance land-memory phenomena. Thus, if positive feedback mechanisms are present in the coupled land-atmosphere system, an initial anomaly can persist through reinforcement at both climate and weather time scales. Scott et al. (1997) and others demonstrated the importance of both soil-moisture reservoir size and the recycling of precipitation.
Thus, it is apparent that, in certain conditions, land memory in the form of the soil-moisture storage, perhaps reinforced by positive feedback mechanisms such as recycling of precipitation, has significant effects on atmospheric variability and predictability and can lead to greater persistence of weather and climate anomalies. Delineation of the conditions under which soil-moisture state is important to the evolution of weather and climate, coupled with ways of estimating the initial soil-moisture state based on in situ and satellite observations and the realistic simulation of the subsequent evolution of that soil-moisture state, are expected to improve skills in predictive models for weather and climate. Thus, hydrological observations available in a timely fashion can be expected to make a very important contribution to improving the accuracy of regional weather and climate predictions. They are important for medium term weather forecasts but the significance increases significantly with the length of the forecasting period. In addition to soil moisture, freshwater input into the oceans plays an important role in modulating thermohaline circulation and thereby affecting climate variability at longer time scales. It should be noted that observation requirements for short-range weather forecasts to seasonal forecasts are already being considered by the Task Force on the redesign of the global observing systems (CBS-OPAG).
3.2 Climate Diagnostics and Change Detection
Prof. Phil Jones discussed the need for long measurements of climate and hydrologic data for studies of climate variability and change. Gridded fields are easiest to work with in most of these studies, although point measurements are necessary is some applications. New et al. (1999, 2000) have developed fields for many climate fields and it is essential to develop these further and extend them to some hydrological variables such as discharge and runoff, for both climate variability and change studies and climate model validation.
Most studies of changes in climate variability and in climate change detection have focused on temperature, because this field can be well represented by the available network. The temperature field exhibits relatively high correlation decay lengths, allowing grid-box scale (5o x 5o) averages to be reliably developed. Both future impacts and those of past events are, however, much more dependent upon changes in precipitation. Changes are not just vital for hydrology, but are also much more important than temperature for many other sectors, for example agriculture and forestry. Studies of large-scale changes in precipitation are hampered by the need to gain access to considerably more precipitation data than is conventionally available. A similar case can also be made for runoff data. Climate change detection studies need to be undertaken on a global scale, and both the available networks of precipitation and runoff data are inadequate. At present, the best that can be achieved are studies on regional and catchment scales.
3.3 Water Availability Modelling
Dr Döll presented data requirements for the global modelling of water availability and use, the aim of which to derive scenarios of the future water situation in river basins. She listed the following types of data which, too a large extent, relate to the human interference with the natural water cycle and are necessary for an integrated water assessment:
Gridded data (spatial resolution 0.5° or higher):
Meteorological data, in particular precipitation (preferably daily, and including correction factors for measurement errors) and extent/depth of snow cover
Land cover
Point data:
River discharge directly upstream and downstream of (large) lakes and reservoirs
Evapotranspiration
Soil moisture
Water quality parameters (in particular nutrients)
Naturalised flows
Location, volume-area function and management of reservoirs
Location and other data of hydropower plants
Location and other data of wastewater treatment plants
Area (and volume) of lakes and wetlands (time series)
Polygon data (for administrative units):
Withdrawals and consumptive water use differentiated according to source (groundwater, surface water) and user sector (agriculture, households, manufacturing, cooling etc.), similar to water use data for all the counties of the USA (collected by United States Geological Survey)
Water quality: source emissions of relevant substances.
Preferably, the data should be global and represent a long time series. They should include information on how the data set was derived, and on the uncertainty of the data.
Dr Fröhlich presented the methodology and results of a case study for a GIS-related balance of water availability and water demands in large river basins by using published generally accessible data and information. The core of that balance calculation is the location- and time-related comparison of available resources with sectoral water demands in the river basin, while the underlying methodology is an in-depth balancing by means of a long-term water management model on the basis of the Monte-Carlo technique. It allows to take into account diverse water uses and demands in their temporal and spatial variability. Thus, demand functions may be adapted individually, and qualitative or economic parameters, interactions with groundwater or flow-times in the river system may be considered. The monthly balancing step makes it possible to evaluate the satisfaction of demands both in the annual averages and in the variations during the year. The outputs of the balancing procedure may be exceedance probabilities of events at any point along the river course, duration of events, mean values and mean minima and maxima of monthly streamflow. Among others, the study yielded a method that is applicable on large river basins, provided that plausible and reliable data can be obtained that describe the anthropogenic impacts on the hydrological system.
3.4 Precipitation Assessment
Precipitation links the global water and energy cycle, and simultaneously belongs to the scientific disciplines of meteorology and climatology as well as to hydrology.
The variable of interest is first the precipitation depth (measured in mm = ltrs.m-2 of water fallen during a defined time-interval as hourly daily, weekly or monthly).
Other variables of interest are the precipitation rate or intensity during short time-intervals, number of days with precipitation per month or year, or the fraction of liquid, solid and mixed of the total precipitation depth.
3.4.1 Precipitation Data Sources:
a) Point data
Conventionally and during most time of the past, daily and monthly precipitation depth is locally but directly observed using gauges, called raingauges or precipitation gauges) at stations operated in national or local meteorological, climatological or hydrological networks. The number of raingauges being operated is estimated to be about 200,000 world-wide. Recording raingauges operated at some of the stations also deliver short-time precipitation rates or intensities. The systematic measuring error due to evaporation losses and wind drift is a problem to be considered.
An operational global exchange of a limited number of data (about 8,000 stations) is established within the World Weather Watch co-ordinated by the WMO. For climatology purposes, data from a subset of these stations (about 1,000) are monitored and collected within the GCOS Surface Network (GSN). Additional data are bilaterally exchanged between countries or provided by the originators to regional or global research projects, or have been collected by individual institutes from more or less official sources.
b) Gridded data
Gridded precipitation data represent the area-averaged precipitation depth for grid-boxes defined by metric or geographical co-ordinates. There are three different types of gridded data sources:
Area-mean respectively grid-box precipitation can be derived from locally observed point data using interpolation methods as kriging, objective analysis or empirical algorithms. The accuracy of the resulting area-means respectively the sampling error strongly depends on the number of data points (stations), but also on the grid or area size, the temporal resolution and the regional climate variability.
Remote-sensing techniques, being available since the seventies, as radar and satellites provide information on precipitation rates based on volume-integrated observed radiation using radiation-cloud-rainfall models or empirical relationships. Advantage of these data is the equal and complete area-coverage. But the indirectly gained results need to be adjusted to corresponding data derived from direct (in situ) observations. Time-series of satellite-based data are heterogeneous since the techniques have continuously been improved during the last 20 years and there is still a development long-term going on.
Numerical weather prediction models also deliver area-mean precipitation data corresponding to the spatial model resolution. But even in the NWP model re-analyses of ECMWF or NCEP the precipitation results are fully predicted and not based on analysed observed precipitation data and need be verified.
3.4.2 Requirements
The requirements for precipitation data depend on the purpose and way of application. For statistical analysis and climatological diagnosis of temporal variability and trends, observation-based climate change studies, long-term homogeneous time series are required, which are available from a smaller number of stations only. A full area-coverage is not given by the available data, but seems to be not necessary.
Studies on the global climate system and the global energy and water cycle, on global and regional interactions between atmosphere, ocean and land surface, on the mechanisms and structure of quasi-periodic climate variations (as El Nino), but also the verification of model results and validation of remotely sensed data require gridded precipitation data sets of best possible accuracy and of a suitable spatial-temporal resolution. Hydrological studies generally require area-mean precipitation data. Partitioning of solid and liquid precipitation is required for run-off modelling.
Specific requirements for the assessment of area mean precipitation (gridded data) based on raingauge observations are sufficient data density (about 10 stations per grid-box), performance of quality-control for the data used, assessment of systematic measuring errors and corresponding data correction, and delivery of the estimated total error on the individual grid-box with the product.
3.4.3 Limitations
The major constraint is data availability. Under real conditions a full global (land surface) coverage by gridded precipitation data of sufficient accuracy can be reached so far for the scales 1.0° by 1.0° spatial and monthly temporal resolution. For certain regions such as West and Central Europe, North America or Australia as well as for a number of smaller regions (individual countries), continuous daily analyses on a 0.5° or smaller grid could also be produced and made available for international programmes. Progress might be possible in future based on satellite observations.
3.5 Water Balance Modelling
Understanding the terrestrial energy and water cycle over a range of spatial and temporal scales is a primary focus of the Global Energy and Water Cycle Experiment (GEWEX) and its Continental Scale Experiments (CSEs) such as GCIP, BALTEX, GAME, LBA, MAGS, etc.). The development of coupled hydrological and atmospheric models and their verification over a number of catchments is an integral part of these projects. The transfer of methodologies to other continental areas and different climate regimes is anticipated in order to finally improve global climate models and regional scale weather forecast models. In these and similar studies, hydrological data form the basis for:
The development and validation of parameterisation schemes for surface and soil hydrological processes
The verification of the description of surface hydrological processes in climate and weather forecast models
The assimilation of hydrological variables in real time into weather forecast models
The real-time validation of the hydrological cycle in weather forecast models.
To meet the requirements of atmospheric models, hydrological data provided by a global network should be reported as point data with the highest possible temporal resolution. The spatial resolution of atmospheric models varies from some km to a few hundred km with a temporal resolution of minutes to hours.
A hydrological network should include real-time variables as well as long time series to allow the verification of operational weather forecast models and the investigation into climate variability and trends. A global network with an increased density of the observing stations over particular regional areas should be anticipated as the regionalisation of climate change predictions is of utmost importance.
The key variables to be included are precipitation -liquid and solid- (3 or 6 hourly), discharge (naturalised flow, i.e. streamflow corrected for manmade storage; daily), lakes and reservoir levels, soil moisture, snow water equivalent, snow cover area, snow depth, and evapotranspiration.
