Mapping Physical Vulnerability to Extreme Precipitation on River Basins with Water, Agriculture, and Energy Resources Infrastructure

Infrastructure’s (e.g. agriculture, water resources, energy generation and ecosystem services) vulnerability and exposure to floods evidence how physical and socioeconomic factors affect preparedness, resiliency and adaptive capacity of strategic infrastructure across spatiotemporal scales and political boundaries. Our goal is to propose a diagnostic and prognostic mapping of physical vulnerabilities to extreme precipitation across geospatial contexts relevant for agriculture, water, and energy resource management at basin scale. Statistical modeling of flood inundation represents an alternative to producing a “hydrologic-hydraulic downscaling” of precipitation’s complexity into diagnostic and prognostic maps of flood inundation areas. The spatial distribution, intensity, and frequency of extreme precipitation illustrate hydroclimatic areas of extreme precipitation but poorly identify how prone these areas are to such hydrometeorological extreme event. Traditionally, engineering planning has use return periods to estimate the likelihood of occurrence of a particular precipitation event. Certainly this stationary process relays on the historical record of observations and requires further investigation to implement a non-stationary approach for forecasting and prediction purposes. Determining the return periods of extreme precipitation and their spatial distribution may offer an alternative to gradually introduce non-stationary concepts into planning and design of future infrastructure. Our objective is to create flood vulnerability maps from the integration of statistical modeling of the return period of precipitation altering temporal frameworks and using historical field records. We consider the two statistical steps to study extreme precipitation: (1) the use of statistical distribution functions and (2) the estimation of percentiles (estimated from a selected distribution function). Multiple statistical distribution functions have been applied to assess probabilistic estimates of precipitation occurrence according to the scope of the study and the availability of data (e.g. Kappa, Generalized Pareto, generalized logistic, exponential, lognormal, and Pearson). However, modeling extreme precipitation through such distributions is theoretically inappropriate. Thus, our use of the Generalized Extreme Value distribution (GEV) applied to maxima of a variable using a block technique is also recommended by the World Meteorological Organization (WMO). The precipitation data used is a 64-year (1950-2013), daily, gridded product available at 1/6^th degree resolution from southern Canada to southern Mexico. The experiments to be performed include: a 64- and two 32-year samples using daily data on within monthly timespans to select precipitation maxima and fit those values into the GEV. The database includes more 300,000 gridcells and each gridcell will be analyzed to obtain the parameters of form, shape and location, as well as return periods of 10, 20 and 50 years. Mapping of basins with key infrastructure on hydropower generation (Columbia and Grijalva River Basins in the US-Canada and Mexico, respectively), agriculture (Sacramento and San Joaquin River Basin, in the US) and water resources (Missouri River Basin) will be selected. Aggregated values and variability in precipitation return periods at basin scale will be used to identify areas physically vulnerable to flood. Ultimately we will show changes in vulnerability by month, basin, sector, timespan, as well as the computational effort invested in the process.