Estimate of Long-Term Water Availability for a Dammed Watershed in Texas Using a Markov Chain Monte Carlo Method with Paleo Drought and Trend Consideration

Abstract:

How much water supply can come from a reservoir watershed in Texas in 2030, 2040, ......, 2080? This is a key question that must be answered in Texas’ State Water Plan. Current rules require regional water planning groups to use reservoir firm yield, which is maximum annual diversion from a reservoir at 100% reliability during a repeat of the drought-of-record (DOR). Firm yield is simulated using the Texas Commission on Environmental Quality (TCEQ) Water Availability Model (WAM) Run3 as an estimate of future water availability. Most of the TCEQ WAMs use hydrologic input data extending from the 1940s to recent years. The 1950s drought, or 2010s drought in some areas of the state, is considered as the DOR for planning purposes. However, there could be droughts worse than the DOR in future planning decades (2030–2080). There is a significant downtrend in stream flows in western Texas (Harwell, et al, 2020), implying non-stationarity in future hydrology. Basing on a single drought-of-record from the observational record, in the face of hydrological non-stationarity, means that future potential risk is not accounted for. This study proposes a methodology for how water planning could account for droughts potentially worse than the 1950s or 2010s DOR. We take the Lake Meredith in the Canadian River basin as an example. The Cook et al (2004) gridded (2.5° x 2.5°) reconstructed Palmer Drought Severity Index dataset (1400–2003) is employed to determine critical droughts (14 years with 600-year recurrence interval), exceeding the 2010s DOR (10 years). A Markov Chain Monte Carlo method is applied to generate a synthetic hydrologic dataset (10,000 series) for the future period (from 2019 to 2080) that accounts for the trend in naturalized flow in the Lake Meredith watershed. This is based on a classification of annual naturalized flow (1948–2018) into dry (lower 25%), wet (upper 25%) and average (25-75%), and the identification of annual lag-1 transitional probabilities. Using the 10,000 hydrologic input series, 10,000 firm yields for each planning decade are simulated using the TCEQ WAM Run3. Results indicate that future water availability for Lake Meredith watershed is about 12,000 ac-ft/yr in 2030, gradually reducing to about 750 ac-ft/yr in 2080, based on a criterion of 90% exceedance probability at the new benchmark DOR of 14 years.