At the February meeting of the California Water Commission, the commissioners began their 2022 State Water Project review by hearing a series of presentations on how the Department of Water Resources (DWR) is adapting its planning and operations to prepare for climate extremes.
The first presentation was by John Andrew, DWR’s Deputy Director for Climate Resilience, who provided context for the following presentations by briefly reviewing the Department’s Climate Action Plan. Next, state climatologist Michael Anderson detailed the work underway to improve the forecasting to account for the climate change impacts the state has been experiencing.
Covered in Thursday’s post, Andrew Schwarz, State Water Project Climate Action Advisor, discussed the State Water Project Delivery Capability Report and how it is being updated to include new information about climate change. Lastly, John Yarbrough, Assistant Deputy Director of the State Water Project, provided an overview of the 2022 drought response.
JOHN ANDREW: Overview of the Climate Action Plan (CAP) – DWR’s Comprehensive Response to Climate Change
The Climate Action Plan is the Department’s guide to addressing climate change in the programs, projects, and activities over which it has authority. The plan has three phases: A greenhouse gas reduction plan, climate change analysis guidance, and a vulnerability assessment and adaptation plan. The upcoming presentations are a part of the plan’s phase two and phase three components.
The phase one greenhouse gas emission reduction plan was adopted ten years ago by the Department. It was last updated in 2020. “Based on our 2020 emissions inventory, we’re actually 70% below our 1990 level of emissions,” said Mr. Andrew. “The benchmark in the state is Senate Bill 32, which requires the state as a whole to be 40% below 1990 level of emissions by 2030. So we’re fully a decade ahead and another 30 percentage points beyond what is required statewide.”
Phase two of the Climate Action Plan was to develop climate change analysis guidance to improve the consistency and scientific rigor of DWR’s approaches for analyzing the potential impacts of climate change while preserving both flexibility and efficiency. Mr. Andrew noted that this part of the Climate Action Plan is legitimately unique, noting that he has not seen this component included in the other climate action plans he has read.
“How we approach climate change and the methodologies and analytical approaches used are as important as reducing GHG emissions and adapting to climate change,” he said.
Phase three has two parts: The first part was the vulnerability assessment completed in 2019. The second part was an adaptation plan which was adopted in 2020.
The vulnerability assessment looked at five hazards: wildfire, extreme heat, sea level rise, long-term persistent hydrologic changes, and short-term extreme hydrologic changes. The assessment also looked at the impacts on habitat and ecosystem services and how the climate is changing the lands that the Department owns and manages. He noted this was also unique as most water utility climate action plans don’t necessarily focus on this either.
The adaptation plan pivoted from focusing on the hazards to identifying and implementing strategies for the staff in terms of extreme heat and emergency response impacts; loss of performance for the State Water Project; wildfire and watershed health for the Upper Feather River watershed, and additional stress on species and habitats on DWR lands.
For more information on DWR’s Climate Action Plan, click here.
MICHAEL ANDERSON: Forecast Improvement to Adapt to Climate Change
Michael Anderson, the state climatologist, discussed the work underway to integrate climate change into the Department’s water forecasts that support other department activities, including the State Water Project.
He began with a graphic of the statewide average mean temperature from 1896 to 2020. He noted that in the early 30s, there was one orange bar indicating an extreme; that year stood for decades as the warmest year and a real oddity. Then in the 1980s, there were similarly hot years a couple of times a decade. However, in the last decade, nearly every year has been above the mean temperature.
“So we’ve gone from an extreme event to episodic events, to now commonplace in terms of annual mean temperature,” said Mr. Anderson. “Watching how the extremes go in an event scenario and how those events build up into an annual value is what we’re going to talk about here.”
Water Year 2021 – New Extremes and Consequences
The picture is a satellite shot of California from October 6, 2020. It was an interesting time for California; the clouds up north are an atmospheric river making its way south; the swirling clouds off the coast of Baja is an eastern Pacific hurricane. The ‘fuzziness’ around California is the smoke from the wildfires that year.
“Those three events were set to collide, potentially bringing a lot of precipitation to California,” said Mr. Anderson. “However, no precipitation resulted from that confluence of events. So we’re really trying to figure out how these things form, what the potential of their interactions are, what the consequences are for water arrival to California, and then its translation from water falling to water running off.”
NOAA uses climate normals, which are 30 year periods where conditions are averaged over those 30 years. The graph below shows the temperature and precipitation for California for the 30 year periods, beginning in 1896 to the present day. The annual average temperature is shown on the x-axis; the annual accumulated precipitation statewide average is shown on the y-axis.
The diamond shapes are 1896 to 1925. The orange diamond is the average; the blue diamonds are the individual years that make up that average. The circles are the 30 years from 1961 to 1990; the squares are from 1991 to 2021.
Mr. Anderson pointed out that the diamond, the circle, and the square are all moving warmer. “It’s getting warmer in California from a climate perspective,” he said. “For precipitation, not a whole lot of movement; maybe a slight decrease. But the overall average state is about the same.”
He noted the diamonds are clustered around the mean (shown by the orange diamond) with a few extreme values, especially 1924, the single driest year in the observed record. The circles are more spread out from the mean (shown by the large circle). However, the squares are spread out much further from the mean.
“So if we’re looking at a mean value, each individual outcome is getting further away,” said Mr. Anderson. “We’re dealing with more extremes and more variability. This is how the narrative manifests itself when we look at the data.”
Water Year 2021 is not shown on the chart; it was the second driest overall, drier than 1977. “It’s down near record dryness, but it was three degrees warmer than those other events were in the history.”
The expectation is moving forward, it’s going to continue to get warmer. So how do we manifest this information as we pull it in and try and forecast?
The chart shows the percentiles for each region. Water Year 2021 was the second driest single year on record for statewide precipitation. The years 2020 and 2021, taken together, were drier than 1976-1977 and a lot warmer, setting a new record for a two-year drought. Water Year 2021 was also the second warmest year for the statewide mean temperature; only 2015 was hotter, and 2015 was the year without a snowpack.
It was also the driest and warmest spring on record (April, May, June). Unfortunately, that’s also the snowmelt period, when some precipitation comes in to help move the snowmelt into the streams and the reservoirs, but that didn’t happen.
On April 1, the snowpack was at 60% of average, but it was distributed differently across the landscape. The primary storm of the year occurred in January; it was a cold storm with a lot of low elevation snow. So rather than the amount of snow increasing with elevation, there was actually more snow at the lower elevations.
“How does that manifest itself when you mix that snowpack with a warm and dry year?” said Mr. Anderson. “The snowpack disappears a lot sooner, the runoff is less productive, and it’s not following the patterns that were in history. Remember, this one year sits outside as an extreme relative to even the modern normal in 1991-2020. So truly an interesting year, and things get more interesting.”
Mr. Anderson then presented a slide showing the Northern Sierra 8-station index, the San Joaquin 5-station index, and the Tulare Basin 6-station index on the left; on the right are the corresponding precipitation deficits from Water Year 2012 to the present.
The graph starts in October of 2011; 2011 was a year with an abundant snowpack and one of the shortest snow-free seasons in the Sierra. So with that as the starting point, the graph shows what happened over the past decade in terms of accumulating the surpluses and deficits in the precipitation index value.
“With all three indices, we’ve had more dry years than wet, and even those really wet years are not offsetting the accumulated dryness of the dry years,” said Mr. Anderson. “Underlying these individual episodic droughts is this notion that over this decade that a drying that is occurring. So the question is, is this an episodic event? Is this a story still to evolve?”
Forecast Improvement Efforts
The climate projections are uncertain for precipitation outcomes, but extreme events are expected to become more extreme with the possibility of more dryness in between. So given the events of last year, the Department is working to adapt faster and adopt emerging technologies to improve the collection of the data.
The previous chart showed index values, which use a single value to represent a broad area, but that can be challenging when the patterns across that area change. So there is a need to move towards more spatially explicit methods of collecting data. Those methods can be a lot more expensive, but there are emerging technologies that DWR is working with research partners to develop.
The second part is to move towards physical-based climate-informed runoff forecasting models. That takes time, but the Department is working with research partners to develop those models.
There are two implementation periods:
- Near-term: Over the next 12 months, what actions can improve forecasting this year?
- Longer-term: Efforts that take up to three years to fully develop and train staff to use them.
There are three strategies:
- Improving the data being collected
- Work to improve the forecast models
- Working with partners on emerging technologies and folding those into the framework
Forecast improvements underway
Mr. Anderson said there are several efforts underway to improve forecasts:
- Updating Hydrologic Averages from 50-yr average to 30-yr average to better reflect most recent years; each station with observed data needs to be updated, so it’s about 400 times that the averages have to be updated. There is a staff of four people who are working on this.
- Update precipitation and snow median increments based on new averages and reset the distribution based on the new centers being considered.
- Improve automation of daily precipitation data collection, full natural flow calculations, and quality control process; there is limited staff, so developing automation to help is important. Also, improving the ‘temporal fidelity’ of the data; oftentimes, the historic data is monthly, but at the daily scale, there is more variability. So they are working on getting daily data; however, as the temporal fidelity increases, the data becomes noisier, so having the right quality control elements to ensure good quality data is important.
- New methodology to evaluate and improve 90% and 10% exceedance forecasts, the balance of the distribution, and where expectations are being set.
- Develop new statistical models based on updated data: Historically, they have used multi-linear regression. But there are new methods, including machine learning, that they are working with partners on.
Forecast Improvement Projects: 0 to 12 Months
New models using machine learning and artificial intelligence models incorporate new variables, such as soil data from the USGS Basin Characterization Model, which looks at the soil column and how that soil column interacts with the atmosphere. They are also looking at the watershed level and how the climatic water deficit can be incorporated into the model, as these elements will prevent water from materializing as runoff.
“We used that as a time series for the model,” said Mr. Anderson. “The model uses the data set from 1896 up to present. We updated each month, working with the model developers to get the new addition to the time series. We use that as a pseudo observed variable. It’s a modeled variable, but it gives us traction into something that we don’t have other data sets readily at hand to pull into. So we’re working with that.”
They are also working to improve the fidelity of the full natural flow data. They are also considering pulling in more May snow data to get a data point in the snowmelt process and distribute it between April, May, June, and July.
They are working to separate the precipitation parameters previously lumped together, giving them monthly values and improving the fidelity of the data they are working with. In addition, they are working with Scripps and Microsoft, looking at several artificial intelligence models and the relationships in these complex datasets to determine if those relationships have more power in forecasting.
Forecast Improvement Projects: 12 to 36 Months
Longer-term, they are working with the airborne remote sensing of snow data. The Department participated in a research project with NASA’s Jet Propulsion Lab from 2013 to 2019, which used airborne LIDAR-based snow measurements and soil data measurements to consider how spatially explicit snow measurement can reduce the uncertainty.
“In other words, instead of using just the manual snow measurements of snow pillows, which gives you an index value of the snowpack, there is an explicit measuring of the snowpack using LiDAR, then working with that much, much larger data set, and feeding that into a model that can actually show how the snow is distributed on the mountainside,” Mr. Anderson explained.
They are pulling in more weather and climate forecast information into the modeling process. Historically, the median conditions were an anchor point, but they are looking to incorporate the improvements in weather forecasting as well as the sub-seasonal to seasonal forecasting.
“We have a lot of partner collaboration in the process that will help as we go, building models that can make use of this data, improving those forecasts, and then pulling it together so that the information presented in Bulletin 120 is moving towards more spatially explicit in terms of information provided, climate-informed, and using physically-based modeling to get us there. And we aim to be there in about three years.”
He noted that their partners and staff are working hard to execute the new technologies along with their existing work to show the difference between the existing methods and the newer methods; this means staff is doing the forecast twice.
California Water Watch
The Department has developed a new website, California Water Watch, which pulls in data from multiple sources and displays it for the public.
“Right now, it is the image of the data that we’re getting from partners,” Mr. Anderson said. “We have precipitation and temperature that’s coming to us from Oregon State’s prism datasets. We have the reservoir data from CDEC. The streamflow is from the US Geological Survey’s surface water and streamflow monitoring network. And groundwater data from the Sustainable Groundwater Management Office. So you don’t have to bounce around for information. It’s in one spot.”
The LIDAR-based snow data is not integrated into the website yet as they are still working on building capability to curate the LIDAR-based snow data. As the staff gets more adept at it, they can also work on making better use of the gridded data in all their forecasting processes, he said.
Water Year 2022 Highlights
In October of 2021 (the first month in Water Year 2022), Northern California experienced one of the largest atmospheric rivers to make landfall in the 21st century. It was category five AR; a scale has been developed for atmospheric rivers similar to hurricanes, with category five being the largest. The precipitation fell on record dry conditions and soils; it re-wetted the soils, but that moisture did not last. Mr. Anderson noted it was a single storm event, not a well-spaced and wet month.
November was dry, with less than 50% of the average for precipitation. Then December was exceptionally wet, with record snowfall for December at the central Sierra snow lab near Donner summit, but he noted that it wasn’t statewide. The October through December statewide precipitation was 155% of average – the 15th wettest.
It was a great start to the water year, but similar to 2013, the precipitation stopped in the new year; it was the second driest January in 127 years. With February staying dry, there will likely be a new record January-February pair drier than 2013.
With climate change, we’re moving into a world where we’re going to have to rely on understanding and making the best use of each storm as it comes.
“We need to move to real-time water management where, at the same time, you are trying to minimize the hazard of that large runoff event, you’re also trying to maximize the benefit,” said Mr. Anderson.
“We have the core of research, working with our partners to really get us that emerging technology and information that’s going to feed the observations,” he said. “Observations feed both forecasts and decision support, the California Water Watch being that curation of the data so you can look at it and see where things are. The observations feed into the forecast model, so you have the more spatially explicit climate-informed forecasts, then you can have the best observation and a great forecast. But if you aren’t digesting it in the information that can be used for subsequent decisions, such as the allocation process, you haven’t quite made it there yet.”
“So you have the third step of really making sure we’re developing that decision support so that we’re getting the information from the forecast to where it can be used in the subsequent analysis and decision making. So there’s a lot of work to go on but a continued investment. I think we’re on the right track to better adapt to the warming world we are living in.”
For more information on the state climatologist and hydrologic forecasting, click here.
QUESTIONS & ANSWERS
QUESTION: Commissioner Gallagher noted that Mr. Anderson mentioned he was switching from a 50-year to a 30-year period; what is the smallest range that could be used? How do you determine that range?
Michael Anderson: “Historically, we used the 50-year average to try and smooth out the decadal-scale variability that you see in the record. We’re finding out now that part of that variability may be the transition to a different state. So we’re moving back to align with our friends at NOAA with the 30-year protocol. That allows us to capture more of what has happened in this last decade with this latest normal update. The smaller you go, you get down into sample size challenges that you may or may not be reflecting all the outcomes, so the idea with a statistical forecast is you’re trying to forecast the whole distribution, not the exact outcome. The challenge there is then when you use that forecast is understanding where the observations are relative to the distribution. And the real challenge comes when the observations lay outside the distribution.”
“So the Department is having to consider much more in their allocation process and their position analyses beyond presuming that Mother Nature sticks inside the historical distribution. That assumption we’ve identified as one that needs to be examined further. So I think the 30 years gives you that sense of the whole distribution and then adding on that additional analysis of what happens and the potential to move outside the distribution.”
QUESTION: Commissioner Gallagher asked if the climate change projects are being incorporated into the allocations at this time?
Michael Anderson: “We still use historical distributions, the most recent 30 years, anchoring what is reflective of what’s going on now. What’s going on now versus what’s going to happen in mid-century versus what’s going to happen at the end of the century are very different items. Trying to figure out how that information might inform that secondary analysis of that transitory state, being outside the historical distribution, is something to work on, not what we’re trying to catch up to right now. We’re trying to catch up to how we best reflect a distribution of outcomes based on current conditions. How do we transition to more spatially explicit data that can capture the spatial variability, instead of an index value that covers the American up through above Shasta Reservoir, looking at the spatial variability … looking at the elevation gradients, and the way those are changing, that takes a while to develop capacity. We’re aiming for three years.”
QUESTION: Commissioner Makler asked since we’re so reliant on snowpack for storage, has there been much discussion or in nexus with forest management and the absorption of the runoff? Is that predictable? Or can you draw some causation with the state of the forest and the aftermath of the various fires?
Michael Anderson: “With the aftermath of the fires, we’re seeing landscape-level change. In fact, two-thirds of the Feather River watershed has burned in the last three years. So the watershed they’re working with this year is a very different watershed than they were working with back in 2013. So those fires have been that motivator that helps open that dialogue with our partner agencies, really talking about the relationship between land cover forest health and how that information works in terms of water management.
“One of the great things about airborne snow observatory and the LIDAR-based snow measurements is the LIDAR becomes fantastic data when you get that baseline. It can also be used to understand the forest structure. So a snow-free flight gives us the baseline information, which becomes fantastic information for the forestry folks to look at and understand forest canopy structure and other things.”
“In the future, we’re hoping that data then becomes helpful to us as we get those physically based models to be able to get to a better understanding of how the landscape is using water, and how that’s changing as we move into this hit and miss cycle of water arriving in the state. And understanding what of it is impacting the surface, what’s impacting that subsurface that still drains to the streams, and what the impacts are of that more resident water in the watershed that slowly dries out as you get into these multi-year droughts. Then on the flip side, how does that recover?”
“As we get better at articulating that process, we’ll get better at being able to forecast that water availability. So the forecasts we provide to the water project, they’ll have better information to make their decisions. That’s what we’re aiming for. We’ll see if we get there.”
