Roaring River levees overtop at Grizzly Island in Solano County, Calif, located in the Sacramento San-Joaquin Delta, January 13, 2017. Florence Low / California Department of Water Resources

DECISION SCALING: A risk management approach in Delta water supply vulnerability

Roaring River levees overtop at Grizzly Island in Solano County in January, 2017. (Photo by DWR/Florence Low)

Planning for future conditions has always involved uncertainty, but the additional uncertainty imposed by a changing climate means new approaches are needed for exploring system vulnerabilities and potential adaptation responses. At the 2019 State of the Estuary conference, Andrew Schwarz, Supervising Engineer for the Delta Stewardship Council, gave this presentation describing a relatively new approach to handling uncertainty in climate change impact assessments and how it can inform water resources management.

The approach is called Decision Scaling, and what’s novel about this approach is that it fits into a risk management approach to decision making which is more familiar to planners, resource managers, and engineers that have dealt with uncertainty from other aspects of the system.

We have some decision making frameworks for dealing with uncertainty,” he said.  “We’ve always had uncertainty about population growth, about land use change, and about other aspects of our environment that are changing that we don’t know how things are going to unfold in the future and so this approach fits more into that line of thinking.”


The traditional approach to assessing climate change impacts begins with the selecting scenarios from global climate model projections and downscaling them.

For water supply, those scenarios are typically run through a hydrologic model to look at streamflow projections.  Those streamflow projections might then be run through an operations model to get conditional performance projections, such as if this happens in the future, then we’ll be able to supply this much water or we’ll have this much water in our reservoirs.

Often, there are an array of scenarios, such as a worst case, middle of the road, and best case scenario.  The result usually is decision paralysis, he said.  “We have to do something, but we don’t make the wrong decision because then we’ll have stranded assets and our decision will be terrible.  So put yourself in the decision makers situation of spending billions of dollars to adapt, these things fall prey to other priorities if we don’t have a clear path forward.”


The World Bank did an evaluation after years of trying to invest in climate resilient projects around the world, looking at how they had been making decisions, and they came up and the review was not very positive about using these top-down approaches.  The top-down approach is useful for setting context, but less useful for making informed investment decisions; it has relatively low value added for many applications.  To address that, the World Bank is starting to employ more bottom-up approaches that fit into a bottom-up line of thinking.

Bottom-up approaches such as Decision Scaling start with trying to understand the system first, so in the case of a water supply system, we would use a system and operations model and do a vulnerability assessment; then as a later or a last step in the process, we bring in the global climate model information, he said.

We can use a broader ensemble of tools from the Global Climate Model to try and understand how likely we might be in circumstances that really threaten our system,” he said.  “It really feeds directly into more adaptive planning process where you can evaluate different assessments in the future more readily and quickly.”


For this presentation, Mr. Schwarz will be drawing on an example for work that was done at Department of Water Resources prior to his coming to the Delta Stewardship Council.  The project was a collaboration between the Department of Water Resources and the Hydrosystems Research Group at the University of Massachusetts Amherst.  The work led to DWR’s Climate Change vulnerability assessment along with a detailed report.

This was an amazing example of the coproduction of science for policy implementation, which is something that is really important,” he said.

The approach he would be describing is called Decision Scaling.  “Decision scaling is a weird name, but if you think of downscaling as starting with the climate and downscaling, this is starting with the decision and scaling from the decision,” he said.

The first step is to set a critical threshold of performance; this will be different depending on the system you are working with.  For a flood protection system, it might be a 200-year level of protection, and if a storm that goes over, then the system fails.  For a species in an ecosystem, it might be a minimum population level that if it goes below that, the species cannot reproduce and continue.

For DWR’s climate vulnerability assessment, Mr. Schwarz was focused on State Water Project delivery reliability, and the critical threshold was set at the historical level of reliability of the system versus it getting worse or perhaps getting better in the future.

The next step is to identify the key uncertainties.  There are a lot of uncertainties in different systems, such as temperature increases, average precipitation change, sea level, rise, population growth, land use change – all these different things that can be a source of uncertainties in the system.  For DWR’s study, they looked at temperature increases, average precipitation changes, and sea level rise.

Mr. Schwarz then presented the resulting graphic, showing the percent change in average annual State Water Project deliveries.  On the vertical axis is temperature change; on the horizontal axis is change in precipitation over time.  The 0,0 point is the current conditions or the amount that the State Water Project can deliver today.  The black line is the critical threshold, which is the combination of temperature and precipitation or the level of change that can occur and maintain the historic level of water deliveries.  Worse than historical is shown in the brown/orange colors; blue colors indicator greater than historical.

This surface is agnostic with respect to time; it doesn’t matter when these impacts occur,” he said.  “If they occur, this is how the system responds.  This is a stress test; it is a two-dimensional stress test or sensitivity analysis to these two aspects of change.  There’s also sea level rise embedded in here, snow melt embedded in there, and loss of snowpack is embedded in there, and so that just gives you the sense of vulnerability of the system.”

The next question is how likely are we to end up at the far left hand side, far right hand side, or somewhere along that black line? The black dots or triangles on the graph (below, left) are different outputs of global climate models for the State Water Project watershed area at 2050, which show where the global climate models are projecting for the future.  He noted that it’s very easy to swap in and out different information at this step if new climate models come along.

Then with some fairly simple math, that ensemble of outputs from different global climate models can be compressed using a probability density function to create a cloud (above, right) of uncertainty about where climate change will take future conditions.

The warming is undeniable among all the models,” Mr. Schwarz said.  “There’s some uncertainty in how much warming and how fast it’s going to occur but it’s going to get warmer, and by the 2030s or so, there is virtually no probability left that we will not have seen warming.  But the precipitation story is much more uncertain.  It’s all over the map.

Adding these two pieces of information together, they can plot the cloud over the top and get the conditional probability of where things might end up.

What we’re seeing is that it’s possible that we might get an outcome that leads to something better, but it’s pretty darn unlikely versus all the probability over here that things are going to be worse,” he said.  “Now if you’re trying to make a decision on whether to act or maybe it might not be that bad, this probably leads you to believe that probabilistically, you had better act because there’s a much better chance that things are going to be bad or even really bad.”

They performed the same analysis with several other different performance metrics such as storage and net Delta outflow (below, left).  He acknowledged Net Delta outflow is one of the more complicated metrics, but for all the other metrics, the conclusion is clear.

It’s nearly certain that we’re going to lose performance of the system with climate change,” he said.  “There is a very small sliver of getting better, but it’s diminishingly small.

Another way to look at this is the annual condition (above, right).  The blue line shows current conditions, and there already is uncertainty in the delivery reliability of the State Water Project on an annual basis even without whatever amount of climate change we have already seen today, he said; some years there are high deliveries, some years there are low deliveries.  The red curve is at 2050 with no adaptation, and the shift to the left is the loss of performance across all the uncertainty and climate change projections, rating them from how likely they are to occur based on the probability density function.

The gray in the background on the graph to the right are the current conditions, and the red line shows what would happen at 2050 if we do nothing. The purple line is a hypothetical adaptation strategy that could potentially push things to a better position.  The slumping down is really a loss of certainty.

With any adaptation strategy that we might see in the future, we are not going to be able to get back on this curve because our view of the future is less certain; it’s always going to be less certain than our view of the present,” he said.  “There will be things that we don’t know about, so what we should be trying to do is just look for adaptation strategies that can move this curve to the right.”

Now it’s still less certain, and there’s still some bad years over here that are much more likely than our historical gray in the background,” he continued.  “But it’s certainly on average pushes us in the direction that we would want to go, so this maybe would be a strategy we would want to consider in addition to some other strategies that are going to hopefully push these years back to the right so we don’t have those bad drought years.


The climate models and top-down approaches have been more useful for setting context.  They are still excellent tools for exploring possible ranges of outcomes and establishing the need to look deeper at these problems and to understand what’s going on with our system, he said.  They can be very effective for communicating with the public because they are more deterministic, the public oftentimes doesn’t understand probabilities and risk in the same way professionals that deal with this every day do.  This approach is less modeling intensive and it can evaluate a lot of different stressors simultaneously.

But in terms of informing investment and policy choices, bottom up stress testing really is a better tool to use, he said.

We can articulate existing and climate driven risks more explicitly so we can understand that we live in a system that has inherent risks,” he said.  “We drive our cars every day, we know we have risks with that.  We’ve all agreed to pay a little bit more to have seatbelts in our cars because we know that’s a cheap investment that makes things safer.  But how many of you are driving cars that don’t have all the new eyesight assist or stopping technologies and all the different things that you can buy that could make your car safer and make every trip safer?  You’re making a policy decision in your own mind that that investment that the additional money is not worth the additional safety that it’s going to allow it or you can’t afford it because you have other needs.  We as a society have to do that.

There are risks that we’re going to have to absorb, and policy makers have to make the decision of what to invest in and what residual risks will go unaddressed.  “I think this approach really focuses on the decision and the investments that folks can make, and that probabilistic evaluation for adaptation strategies really allows them to make that tradeoff between how much to invest and what we get for it,” he said.  “It fits into a risk management framework and we can really explore the system along those different axes of each stressor so then we can look at how fast is temperature changing; if the system is really sensitive to temperature, we can monitor temperature change and say, this is happening faster than we thought it was, we had better make more rapid decisions.”

Mr. Schwarz closed by noting that at the Delta Stewardship Council, they are currently working on a vulnerability assessment taking a hybrid approach with these two tools to really employ them for different purposes and for communicating to different audiences.  They are evaluating Delta flooding using a critical threshold of when the water surface goes over the top of the levees, looking at sea level rise and changes in riverine flow.


Daily emailsSign up for daily email service and you’ll never miss a post!

Sign up for daily emails and get all the Notebook’s aggregated and original water news content delivered to your email box by 9AM. Breaking news alerts, too. Sign me up!

Print Friendly, PDF & Email