Evaporative loss from reservoirs is increasing in demand, and not just by the atmosphere

Joshua B. Fisher
Science Lead, Hydrosat
Associate Professor, Chapman University
Project Scientist, UCLA

With Matthew Dohlen, Gregory Halverson, Jacob Collison, Christopher Pearson, and Justin Huntington

I was recently in a room of State water managers introducing them to some of the data and tools we’ve been developing for satellite-based evapotranspiration (ET), part of a NASA Applied Sciences Water Resources project that I lead. They were impressed, commenting with feel-good phrases such as “innovative”, “game changer”, “extremely useful”, “will revolutionize our work”, and “bad ass”.

Most of the focus was on agriculture, drought, and wildfire. We can provide crop stress detection and irrigation recommendations through cutting-edge tools from Hydrosat’s IrriWatch platform. Our Evaporative Stress Index (ESI) provides better resolution than existing drought tools, allowing the State Climatologist and team to make decisions on drought categorization with greater clarity and connection to the people and ecosystems beneath otherwise coarse US Drought Monitor maps. The ESI also provides forest stress maps, which, in conjunction with ET, improve predictive power of wildfire spread and intensity.

Of the 50 slides I presented, I included n=1 on our reservoir evaporation capabilities (Figure 1). This one slide was disproportionately well-received. Apparently, the State uses a hand-drawn map from the 1970s for open water evaporation. And they use it a lot. Any of my hesitations or insecurities about my data uncertainties immediately evaporated (but my bad puns will never disappear); I know that whatever I’m producing is going to be a big improvement. There is strong demand for data on evaporative loss from reservoirs and lakes for water management. They need that data more than ever as water resources become increasingly variable and on a collision course with rising demand for water both by expanding populations and atmospheric aridity.

Figure 1. New capabilities in operational high resolution reservoir evaporation reveal fine scale patterns.

State water managers can sometimes move water from one reservoir to another to reduce loss from evaporation. Half the water loss from a reservoir may be from evaporation alone, on par with human abstraction. In sum those should be roughly equal to total water inputs. But, with leakage and other water uses such as river health maintenance, if the outputs exceed inputs, then the reservoir shrinks (Figure 2). Hence, knowledge of evaporative loss is crucial for making these management decisions.

Figure 2. When outputs exceed inputs.

In the absence of sophisticated reservoir evaporation measurements, one could make a guess based on pan evaporation. Or do a quick model calculation with a nearby weather station. And this is what people have done for decades. Then do your best to calibrate because both these approaches are woefully inadequate.

The processes controlling terrestrial water body evaporation are dynamic in space and time. They’re similar to those that control land and plant ET — radiation, humidity, wind, and air and surface temperature — but with different sensitivities. Plants and land introduce additional surface, stomatal, and aerodynamic resistances, as well as variable access to water. Open water, on the other hand, has much deeper radiation-absorbing and energy-moving characteristics that may result in evaporation in an entirely different place than where that energy was absorbed, as heat is circulated throughout the water body. These characteristics are strongly determined by the physical structure of the water holding landform and underlying bathymetry (e.g., you feel warm water when you start walking into the ocean from the beach that rapidly turns cold when you get into deeper water). Open water may be more sensitive to wind events than are physically/structurally buffered forests. Ultimately, radiation continues to be the dominant driver of both land and open water evaporation at large space and time scales; but, the process shifts to atmospherically-controlled at short time scales. For these reasons, simple estimates of open water evaporation can easily be in error.

However, measurement of open water evaporation is very challenging. While forest eddy flux towers are difficult to set up in their own right, at least they exist on relatively stable ground, unlike water eddy flux instrumentation, bobbing around, and inevitably the prime spot for birds and other animals to hang out and chill (netflux and chill?). Rigorous eddy covariance assumptions about stability and fetch are often ill-constrained, potentially contaminated by land advection, and these point measurements are supposed to represent the entire water body. Monitoring networks such as the US Bureau of Reclamation’s Open Water Evaporation Network (OWEN), the Great Lakes Evaporation Network (GLEN), the Global Lake Ecological Observatory Network (GLEON), and the Western Reservoir Evaporation Network (WREN), among others have done admirable and invaluable job tackling these challenges to provide these critical in situ data.

Remote sensing can overcome the spatial representation problem, especially given high spatial resolution measurements of surface temperature. Such resolution can identify not only spatial variability in evaporation, but also dynamically changing surface area related to water height, total volumetric water change, and changes in bathymetric impacts on evaporation, particularly relevant to sinuous and finely shaped water bodies (Figure 3). But, terrestrial water bodies have been masked out in remotely sensed ET datasets primarily due to the fact that it’s just a different retrieval approach from land/plant ET, and most of us remote sensing ET scientists have been interested in primarily land/plant science and applications.

Figure 3. Some of these reservoirs are really squiggly (aka fine-scale heterogeneity necessitating high resolution observations).

So, responding to demand from water managers and scientists interested in reservoir, lake, and inland sea evaporation, we implemented a retrieval approach for terrestrial open water evaporation into our operational data production software. We added it to the ECOSTRESS and OpenET operational data production software systems. We collected in situ open water evaporation measurements across the world (Figure 4) in among the largest open water evaporation validations to date (Fisher et al., 2023. Remotely sensed terrestrial open water evaporation. nature.com/articles/s41598–023–34921–2). We found that the retrieval does a pretty good job — not as good as we do for land/plant ET, but there may be more error in the open water in situ measurements than in land flux measurements. For comparison, we also ran a suite of machine learning models, but found that they did not necessarily do much better than the process-based retrieval approach. Much of the uncertainty at the instantaneous level was due to acute high wind events; this sensitivity minimizes, however, when moving to daily temporal integration. Still, the analysis gives us a baseline of error that practitioners can decide is tolerable or not for their applications, and it gives us some room to improve.

Figure 4. Among the largest open water validations allowed us to quantify error in the remote sensing data, showing that the remote sensing does a pretty good job though not perfect, but useable (Fisher et al.., 2023).

The future of remotely sensed terrestrial open water evaporation is promising with new missions emerging that increase the spatial resolution and frequency of surface temperature measurements. The Landsat record continues to be supported with regular launches to replace aging satellites. ECOSTRESS will stay on the ISS pretty much close to the end of the ISS life. SBG, TRISHNA, and LSTM will eventually provide consistent, high quality, and well-calibrated data. Finally, Hydrosat will provide the highest spatiotemporal surface temperature measurements at 50 m, daily, multiple times per day. Other commercial companies such as ConstelIR will continue to add TIR measurements. Additionally, synergies with radar measurements from SWOT enable new monitoring capabilities of changes in reservoir and lake water height levels. Coupling these measurements with reanalyses and in situ networks, along with our operational systems such as ECOSTRESS and OpenET, expand the instantaneous remote sensing measurements throughout the day. Taken together, our ability to operationally monitor open water evaporation from millions of water bodies is key to ensuring water management and analysis of changes in climate and hydrological cycling into the future. And hopefully this is a game changer, extremely useful, and overall generally bad ass.