The potential for large declines in annual streamflow in the western United States due to
climate change presents one of the most significant water management challenges in the United
States. For example, major negotiations regarding water allocation on the Colorado River in
2020 will begin just as new analyses from CMIP6 will be arriving. A physical, process-based
evaluation of CMIP6 model projections is essential to better inform how best to use this
information in water management planning and decision-making in the Colorado River basin as
well as throughout the western United States.
We propose to reduce uncertainties in streamflow projections over the snow-fed West by
using a new class of CMIP6 models, suitably assessed and vetted. Our preliminary work using
AMIP approaches indicates a substantial improvement in climatological runoff production over
complex terrain of the West in models of high spatial resolution, having physically realistic land
surface treatments. We hypothesize that such models, identified via physically-based runoff
process diagnostics, will likewise exhibit improved realism of their runoff sensitivity to
meteorological forcing changes.
The goal to reduce uncertainties in streamflow projections has two core elements. First,
we will characterize uncertainty in future streamflow from CMIP6 models through the use of key
metrics relevant for runoff production including a “Budyko” framework that links water and
energy budgets with climatic factors. Major steps include:
• Validate physics of runoff production in individual CMIP6 models through a process-
based Budyko evaluation that validates the joint statistics of aridity versus runoff production
relative to observations.
• Determine runoff sensitivity to historical meteorological drivers (e.g. temperature and
precipitation variability) across CMIP6 models in the context of both spatial scale and
physical process dependencies.
• Characterize uncertainty in model projections, including the role of GCM horizontal
resolution on model process fidelity using outputs from standard resolution models
and HighResMIP as part of CMIP6.
The second element will use the above hydroclimate characterization of models to constrain the
uncertainty in CMIP6 projections of runoff change. We will cull and/or weight models using a
direct approach and a sensitivity-based approach in order to generate physically conditioned
ensembles of projections. We will focus in detail on the Upper Colorado River Basin–the source
of over 80% of the flow in the Colorado river—to develop metrics and validate our approach.
We will then extend the most pertinent analyses to provide an assessment for the entire Western
We address Priority Area B of the competition by using process-based evaluation metrics
of historic simulations to better characterize uncertainty in hydrologic projections in CMIP6
models. We will also address Priority Area A through using constraints based the evaluation
metrics to create culled and/or weighted ensembles of CMIP6 GCM projections for use in the
water resources sector. Our proposed project supports NOAA’s long-term climate research goal
of addressing challenges in the area of “climate effects on water resources,” as stated in section
1.A of the call for proposals.
Climate Risk Area: Water Resources