2.3 Assessing Flow and Rainfall Data
2.3.3 Rainfall Record Sources
There are many organisations within the UK that record rainfall, for a variety of purposes. Water utilities require rainfall data for resource estimation and allocation, the Environment Agency require rainfall data for water resource assessment and flood warning and the Met Office require rainfall data to assess the performance of numerical weather prediction models (among many other reasons).
There is one major source of freely available raw archived rainfall data (for research) within the UK. This is the British Atmospheric Data Centre (BADC) and it hosts gauged rainfall information from utilities, regulators and organisations such as the Met Office, as well as some private records. It has built in facilities for querying and extracting raw time series and the records it holds are considerable.
For the researcher or scientist interested in analyses using country –wide rainfall data, the use of raw BADC data requires careful consideration. If a large number of records are required, the download time and volume can be considerable. After this, a significant amount of data checking and assessment of quality is required. Issues such as the double counting of rain days or mixing hourly and daily data require that the raw data are subject to extensive quality control procedures. Finally, there is the issue of gauge location, and how many independently extracted records can be used consistently for a study. Individual rain gauge measurements are subject to a variety of errors such as outsplash (rain entering the gauge after splashing off adjacent ground), wind induced under-catch and snowmelt estimation problems. For a more detailed discussion on rain gauge measurement problems see Strangeways (2004). These problems tend to be specific to individual gauges. Regarding the impact on this study, it is important to bear these measurement problems in mind, however, they are difficult to consistently account for in a quantitative manner.
14 The need for consistent rainfall data for large parts of the country for both research and commercial purposes has led to the development of long term time series and gridded rainfall data products. Gridded data for the UK are produced mainly by the Met Office and sold as a commercial product to companies for application in areas such as agriculture, hydrology, ecology and forestry. Gridded datasets are also increasingly used in research for climate model inputs, model validation and trend analysis.
Perry and Hollis (2005) describe the production of a monthly gridded dataset for a range of climatic variables, including precipitation. Their use of techniques such as geographically weighted regression within a GIS allows for the interpolation of climatic variables. For rainfall, there was typically one station for every 7 x 7 km grid cell, however cover was not consistent. Areas such as the Scottish Highlands tend to suffer from sparse coverage of rainfall collection due to the low levels of habitation there, whereas the South of England tends to have better coverage. This station coverage is reflected in the accuracy of the final gridded data product, where areas of low station coverage tend to have higher errors in the gridded data set and vice versa for those areas of high station coverage.
The motivation for the development of long-term time series such as that of Alexander and Jones (2001) has been rooted in the need to put recent climate change into a longer term context. It also allows for trend analysis on a consistent dataset, something which using a series of individual gauges does not easily permit. Alexander and Jones’s work considered both spatial and temporal aspects in the data comparison, and they did this by creating time- series for different regions, identified for England and Wales by Wigley et al. (1984) and extended to Scotland by Gregory et al. (1991). Their focus on using the resultant dataset for an analysis of extremes is noteworthy as it provides some basis for the use of derived products as opposed to raw data for this type of work. However, as these time-series represent a region, they may not be suitable for application over a catchment where the rainfall regime may be considerably different.
The Met. Office has produced a 5 km gridded interpolated daily rainfall data set covering the time period 1958-2002 for the UK. This data set is not
15 freely available and is only licensed for use under certain conditions. The method used to construct this data set has not been explicitly published, although Smith (2010) provides some evidence that the method may be similar to that used to produce the monthly data as reported by Perry and Hollis (2005). Smith also undertook investigations into the use of the 5km gridded data and concluded that while concerns may exist over the lack of relevant information on its construction, the 5 km data set provides an accurate representation of extreme rainfall events. In Smiths study, the 5 km resolution was considered sufficiently small so that individual grid cells could be considered as pseudo- stations.
Fowler et al. (2005) use the 5 km gridded dataset in work that assessed regional climate model output for its ability to reproduce extremes. Their assessment used the RMED values calculated from the 5 km grid and suggests that the gridded data may have potential for use in large scale studies of extremes. There is a considerable advantage in having access to such a consistent data set over raw station data as it allows consistent temporal and spatial comparisons to be made. It is for these reasons that this type of gridded dataset is particularly useful to a study such as this.