Monday, 23 January 2012: 11:45 AM
The ICI-WARM Non-Proprietary Regional Frequency Analysis Tool Using the Method of L-Moments
Room 238 (New Orleans Convention Center )
Manuscript
(943.6 kB)
Poster PDF (2.7 MB)
When it comes to setting policy, making water use plans, designing infrastructure, and making operational water management decisions, it is important to be informed concerning the frequency, severity, and duration of droughts (and floods) in the region. Common questions posed by water planners, managers, and decision-makers include: “How rare is the current drought?”, “How likely is it that the drought will end in X months?”, “How large a drought should we plan for?”, and “How rare is the drought of record?”. These questions can be answered by performing a frequency analysis of rainfall over a particular region. Most tools that have been developed for this purpose require a license or are developed specifically for a particular group's interests. The International Center for Integrated Water Resources Management (ICI-WaRM) is developing a non-proprietary and comprehensive Regional Frequency Analysis (RFA) tool to be used to answer the questions above. This RFA tool uses the method of L-(Linear-)moments, which allows estimation of the shape of the rainfall frequency distribution for an entire homogeneous region. L-moments are analogous to regular moments (mean, standard deviation, skewness and kurtosis) except they are calculated using equations representing linear combinations of the sample or population data elements. The resulting statistics are referred to as L-location (mean), L-scale, L-CV (Coefficient of Variability), L-skewness, and L-kurtosis. The advantage of L-moments over regular moments is that using linear statistics makes the procedure more robust with respect to outliers. L-moments can also characterize a wider range of frequency distributions and are less subject to bias in their calculations. Another advantage of this tool is that it estimates site L-moments by first estimating the L-moments and the associated frequency analysis curves for multiple homogeneous regions that each contain multiple sites. Advantages of performing a regional frequency analysis as opposed to an at-site analysis is that it allows the inclusion of sites with short record lengths, few rain gage stations, several missing data, or stations with a higher variability due to the presence of mountain ranges and multiple precipitation sources. In addition, it is difficult to find frequency distributions that fit an individual site's rainfall data well. A regional analysis that combines the data from all sites can produce a closer fit. An example of the advantages of RFA over at-site analysis is given in Fig. 1. The first graph attempts to fit the Generalized Extreme Value (GEV) frequency distribution to one site's rainfall growth curve. A growth curve expresses rainfall intensities in terms of their non-exceedance probabilities. Instead of plotting non-exceedance probabilities on the x-axis, though, a “Gumbel reduced variate” is used in order to expand the extreme event portion of the graph, which is of most interest. In addition to GEV, the software can attempt to fit 13 other frequency distributions. In this case, the GEV results in the closest fit, but it is observed that the GEV deviates from the data for more extreme events. The location of the 100-year event on the graph is shown for reference purposes. Substantial improvement is seen in the second graph when attempting to fit the GEV to the growth curve of the homogeneous region that contains the site depicted in the first graph. The scales of the y-axes are different due the normalization of the rainfall data within the region, but the fits are still comparable. Once the region growth curve is established, it is multiplied by a site's average annual or monthly rainfall, whichever was used in the analysis, in order to produce the at-site growth curve. The end products of the ICI-WaRM Regional Frequency Analysis tool include estimations of rainfall intensity for individual sites and rainfall probably maps for a particular region for any rainfall frequency, duration, and starting month. When looking at multiple maps of extreme low-rainfall events, the collection of maps is called a Drought Atlas. In addition, a gridded text file can also be outputted for additional spatial analysis in such software programs as ArcGIS. The fact that this software will be free facilitates capacity building through training experts from other nations in its use. This training will be focused in South and Central America as a step towards creating one drought atlas for the entire Latin American continental region. A preliminary map of this drought atlas will be shown. Other future applications include a regional analysis of precipitation frequencies in Rwanda and as a mechanism to compare the effects of climate variability and change. The ICI-WaRM tool can facilitate a simple climate change induced analysis by looking at rainfall data over a region for a range of years using different starting years. Trends over the entire period can then be identified. In terms of climate variability, a database of indices from various types of meteorological events such as the El Nino/Southern Oscillation, the North Atlantic Oscillation, the Pacific Decadal Oscillation, and others has been included. Using this database, only rainfall data from a particular phase of any one of these phenomena can be analyzed and compared to the opposite phase, giving an indication of the effect of that phenomenon on climate variability in terms of rainfall.
Supplementary URL: