Date of Award

Spring 1-1-2013

Document Type


Degree Name

Master of Science (MS)

First Advisor

Edith Zagona

Second Advisor

Balaji Rajagopalan

Third Advisor

Kevin Werner


Climate informed ensemble water supply forecasts are continually gaining popularity as enhanced methods improve skill and tools become available to incorporate these forecasts. This research builds on this idea, using an enhanced Ensemble Streamflow Prediction (ESP) method to create streamflow forecast ensembles. The drawback of ESP is that it uses historical weather sequences to generate ensembles. To address this, previous work adapted a stochastic weather generator that uses K-nearest neighbor resampling approach to incorporate probabilistic seasonal climate forecasts, more specifically winter precipitation forecasts in the San Juan River Basin. This enables generation of a rich variety of weather sequences that are consistent with large scale climate forecasts. We continue to use this flexible approach and incorporate spring temperature forecasts to attempt to better capture the timing of snowmelt runoff.

We then set up a framework to evaluate the streamflow forecasts using the US Bureau of Reclamation's probabilistic RiverWare based Mid-Term Operations Model (MTOM). The streamflow forecast ensembles become input to the San Juan River Basin (JSRB) portion of the MTOM, from which we analyze operational performance metrics to evaluate the enhanced streamflow forecasts methods. The management objectives in the basin include water supply for irrigation, tribal water rights, environmental flows, and flood control. The spring streamflow ensembles were issued at four different lead times on the first of each month from January - April, and are incorporated into the MTOM for the period 2002-2010. Ensembles of operational metrics for the SJRB such as Navajo Reservoir releases, end of water year storage, environmental flows and water supply for irrigation were computed and their skills evaluated against variables obtained in a baseline simulation using historical streamflow.