Abstract: A Data-based Model of Drought Occurrence in the Colorado River Basin
The waters of the Colorado River Basin (CRB) are allocated between the states of the Upper Basin (Colorado, Wyoming, Utah and New Mexico) and Lower Basin (Arizona, California and Nevada) on the basis of the total annual flow of Colorado River water that flows past a US Geologic Survey gage located at Lees Ferry on the border between Arizona and Utah. If the total annual amount falls below a threshold set by a Federal statute, the Upper Basin states must in principle make up the shortfall. For this and other purposes of planning a method for estimating likely future flows is highly desirable. In this talk I’ll sketch a method for making such an estimate using data from a database of annual naturalized annual flows maintained by the US Bureau of Reclamation. The method is in development at UA based on 3 main tools: Peaks Over Threshold (POT) analysis, Return Period Estimation (RPE), and Hidden Markov Models (HMM). The main technical challenge is to apply these methods to non-stationary-in-time records of annual streamflow at the Lees Ferry gage supplemented by records from other gages in the CRB. I’ll focus this talk on the integration of POT with RPE & will outline current thinking on how to drive HMM in the context of non-stationary dynamics.