OpenET is a collaborative effort to develop an online platform for mapping evapotranspiration (ET) at the scale of individual fields. OpenET was built to fill an important data gap in water management across the western U.S. OpenET uses best available science and publicly available data to increase access to satellite-based ET and consumptive water use information for farmers and water managers.
ET, short for evapotranspiration, is the process by which water is transferred from the land to the atmosphere. It includes both evaporation from soil and transpiration from plant leaves. ET is a core driver of the Earth’s water cycle, returning water to the atmosphere to fall again as precipitation.
For irrigated agriculture, ET is a measure of the water used to grow food and is the biggest share of water use in most arid environments around the world. In most agricultural regions, net ET (total ET less precipitation that contributes to ET) is also a measure of consumptive use of water that is diverted or pumped from surface and groundwater supplies.
The word “open” in OpenET stands primarily for the collaborative, open, and transparent development of the platform. OpenET brings together many of the leading scientists and developers behind satellite-based estimation of ET onto one team, and is making public the models, assumptions, and inputs behind the estimates. The OpenET team aims to expand science partnerships through open collaborations with other teams and leaders in the ET community. To ensure accuracy and transparency, the OpenET team completed the largest and most comprehensive ET accuracy assessment and intercomparison study to date. High level results are available here.
It is also a core objective of OpenET to provide open access to ET data for farmers, practitioners, and water managers alike. At a minimum, any visitor to the OpenET site will be able to view field-scale estimates of monthly ET data across the western U.S. for the last five years. Users will also be able to download limited quantities of data directly from the site at no cost.
Consumptive water use refers to all of the water within a system that cannot be recovered or reused, and includes water that is consumed by plants or humans, evaporated, or contaminated.
Sustainable water management is one of the most challenging issues of our time, especially in the arid western U.S. Maximizing the benefits of water supplies requires careful measurement of availability and use. However, one important information gap was compounding this challenge: the lack of consistent consumptive water use data. Before OpenET, access to this data was limited and expensive, keeping it out of the hands of most water users and decision-makers. OpenET fills this huge data gap in order to support sustainable water management and innovation in water conservation.
The OpenET collaborative is led by NASA, the Desert Research Institute (DRI), and Environmental Defense Fund (EDF), with in-kind support from Google Earth Engine. The technical team brings expertise in satellite-based estimation of ET, cloud computing, and user-driven website design. It includes approximately 30 researchers and practitioners from NASA, DRI, U.S. Department of Agriculture (USDA), U.S. Geological Survey (USGS), University of Nebraska, University of Idaho, University of Wisconsin, University of Maryland, California State University Monterey Bay, University of Montana, Google Earth Engine, and web development firm HabitatSeven.
Since its inception, OpenET has been dedicated to user-driven design to ensure it provides benefits to agriculture and supports sustainable land and water management practices. The OpenET community includes partnerships with more than 45 stakeholder entities and organizations. This includes growers and agricultural groups making irrigation scheduling and other decisions at the field scale, water district managers building water accounting and trading platforms, and state and federal agencies making drought and water budget assessments at large scales. The insights learned through these partnerships are integral to the success of OpenET, including defining user requirements, providing feedback on the website design, and testing the beta versions of the user interface.
The models used in OpenET ingest hundreds of terabytes of satellite and meteorological data, which is computationally expensive. Google Earth Engine provides a common computing platform where all of the satellite-based ET models used in OpenET can be implemented and run using consistent inputs. This allows the scientific community to collaborate on software development and produce the most accurate and consistent ET data possible. In addition, Google Earth Engine allows the OpenET team to store, rapidly compare, and analyze results from the ET models, enabling expedited progress on evaluation of model results and identification of opportunities to improve the ET estimates from both individual models and the full ensemble of models. OpenET uses Google Earth Engine to efficiently produce daily and monthly ET images across the western U.S. and to calculate data summaries for millions of agricultural fields or other regions of interest. These data are then exported to the OpenET web-based platform for data visualization and distribution. The Google Earth Engine team is supporting implementation of the ET models and providing the cloud computing resources as an in-kind contribution to the effort.
Providing free and easily accessible ET data to the public is a core objective of the OpenET project. At present, everything on the Data Explorer is free to view, and retrieval of limited volumes of data is also free. Users seeking large-scale access to OpenET data will be able to acquire it through an Application Programming Interface (API). Revenue generated will help to support the OpenET non-profit and fund continuing research and development of OpenET data services. Additional pricing details for access to the API for retrieval of large volumes of data will be announced when the API is launched in early 2022. Historical data will also be added to the public data catalog on Earth Engine and will be freely available.
Geographic and Temporal Scale
OpenET covers 17 western U.S. states: Arizona, California, Colorado, Idaho, Kansas, Montana, Nebraska, Nevada, New Mexico, North Dakota, Oklahoma, Oregon, South Dakota, Texas, Utah, Washington, and Wyoming. Over time, the intent is to expand OpenET to include other states in the U.S. and other regions across the globe.
OpenET is produced at a spatial resolution of 30m x 30m (0.22 acres). OpenET data can also be aggregated for individual fields or other locations of interest by calculating the average data value for the field at each timestep.
OpenET currently provides data at monthly and yearly timesteps. Daily data will be added by the end of 2021. However, since production and storage of daily datasets can be expensive, it is likely that daily data will only be available for irrigated agricultural regions for recent time periods.
Our goal is to provide monthly ET data across the western U.S. with a data latency of six weeks or less. For daily data, we recognize the importance of providing data within two days of satellite overpass for irrigation management applications, and we are working hard to come as close as possible to achieving this target for daily data products.
Publicly available field boundary datasets were collected from academic research teams, state and federal agencies, and the 2008 USDA Common Land Unit (CLU) database. In many cases, these datasets have been modified to remove redundant polygons, as well as small slivers and large polygons associated with grazing on non-irrigated rangelands and shrublands. Priority was given to more recent field boundary datasets produced by state and local agencies, and the modified CLU data were used only in cases where no state or local level datasets are publicly available.
The crop type information available on this site is from publicly available datasets, and does not affect the accuracy of the ensemble ET value, or the five models that use thermal infrared data as an input for calculation of ET (ALEXI/DisALEXI, eeMETRIC, geeSEBAL, PT-JPL, SSEBop). The sixth model, SIMS, does use crop type information to separate annual and perennial crops, and to adjust model parameters for stomatal resistance and maximum crop height for perennial crops. However, errors in crop type classification for annual crops don’t affect the SIMS ET calculations, and the crop-specific adjustments for perennial crops usually have a small influence on SIMS ET relative to other inputs. In cases where crop type errors do have an effect on SIMS ET data (e.g., misclassification of a perennial crop as an annual crop), SIMS is typically flagged as an outlier and removed from the ensemble prior to calculation of the ensemble ET value.
Crop type information available through the Data Explorer is not generated by OpenET. Crop type information is derived from publicly available datasets, which often lag 6-12 months or more behind the current growing season, and thus may not always accurately reflect the current crop. Crop type information is provided to assist with interpretation of ET values. Outside of California, all crop type information is derived from the USDA Cropland Data Layer. Crop type information for California for 2016 and 2018 is based on the CA Department of Water Resources (DWR) Land Cover database developed by LandIQ. Data for California for all other years are from the USDA Cropland Data Layer; this data has lower classification accuracies for many of the specialty crops grown in California.
OpenET will launch an API and custom reporting tools in 2022. The API was developed so that data from OpenET can be easily integrated into other data systems. API and custom reporting tools will allow users to query the data using their own field boundaries or areas of interest, and create custom reports for specific time periods of interest. The API will support machine to machine data retrieval to enable automated and operational use of the data within irrigation scheduling tools, hydrologic models, water accounting or trading platforms, and other farm, ranch, and water management software. Data available through the custom interface and API will go back to 2016 initially, with more years being available by the end of 2022. The ultimate goal is to develop ET data going back as far as 1984.
Applications of the Data
Potential applications of reliable and widely available ET data at the field scale include:
- Development of water budgets and innovative management programs that promote adequate water supplies for agriculture, people, and ecosystems.
- Support for groundwater management programs that require consistent, accurate ET data for monitoring historical and current consumptive water use.
- Support for water trading programs .
- Support for ET-based irrigation practices that maximize “crop per drop” and reduce costs for fertilizer and water.
OpenET is not intended to be a new irrigation scheduling tool. Our goal is to provide transparent, consistent, and easily accessible ET data only. OpenET does not take the next steps of integrating that data with soil moisture, nutrient content, and other information for making decisions or recommendations about when to turn an irrigation system on or off. However, OpenET will make the ET estimates used by such tools much more cost-effective to retrieve, and in many cases more accurate.
OpenET is not intended to replace meters where they are desired or preferred. However, there are two reasons that OpenET may be a complementary and/or alternative source of information:
- Meters are expensive, both to purchase and to maintain, and do not necessarily lead to more accurate estimates of water use. OpenET can provide a cost-effective alternative to meters for tracking changes in water use over time at a geographic scale that is not possible with solely on-the-ground approaches to measurement. In addition, OpenET can provide historic data as far back as 1984, which can be valuable in locations where meters have only been installed recently.
- Meters measure the volume of water pumped or diverted, but they do not measure the amount of applied water that is actually lost to the atmosphere and consumed from the water supply. Some water that is pumped or diverted may re-enter the local water supply as runoff or groundwater recharge. Water managers and growers will often need both sources of data – applied water as measured by meters and estimates of ET – to best understand their water budgets over time.
When evaluating ET rates, it is important to consider the many factors that may lead to variations across fields and regions, even for the same vegetation or crop type. These include soil type and texture, salinity, ground cover, the age of an orchard or vineyard, irrigation system type, distribution uniformity, production goals, fertilizer, pest and pathogen management considerations, and other factors.
The relationship between ET and crop water requirements can be complex. For many crops, ET is a key measure of the minimum amount of water that must be replaced during the growing season through irrigation or precipitation to ensure a well-watered crop with maximum yields. OpenET provides data that can help growers track the amount of water that is transferred from the root zone to the atmosphere through ET and manage irrigation to maintain sufficient soil moisture levels. However, in addition to accounting for ET, growers and irrigators must also budget water for crop planting and germination, salinity management, frost protection, cover crops, variability in the soil type, the distribution uniformity of the irrigation system, the timing of availability of water for irrigation, and other factors. As a result, in many situations the total amount of water diverted or pumped and applied for irrigation must equal or exceed ET, even in highly efficient irrigation systems. Conversely, for some crops, deficit irrigation for key periods during crop development can also be important to maximizing crop quality or crop yields. And in some regions, deep rooted perennial crops can also access stored soil moisture or shallow groundwater deep in the root zone, resulting in high yields even when the total amount of water applied for irrigation is well below total ET over the growing season.
Access to ET data can empower local communities to find their own best paths to water security and sustainability. OpenET levels the playing field, providing small farmers, ranchers, and water managers equal access to ET data so they can:
- Coordinate within a basin to facilitate locally developed water management strategies.
- Shape policymaking with more reliable and timely data that is available to local growers and communities at the same time that it is available to regulators and other decision-makers.
- Cost effectively track and demonstrate the degree to which changes in irrigation technology or management practices contribute to water conservation goals, and share results as different approaches are tested within a community.
- Test flexible management solutions like water trading or incentive-based conservation programs.
ET data can play an important role in quantifying reductions in consumptive water use from water conservation efforts. Managers can use ET data to measure the water savings of different conservation activities and evaluate their effectiveness, not only in terms of water use but also in terms of financial costs. This ultimately allows managers to identify conservation practices best suited for their region and to match incentives with both costs and actual reductions in water use.
OpenET is expected to have a range of applications beyond irrigated agriculture. For example, current use cases are exploring how OpenET can be used by rangeland, forest and watershed managers to help monitor how vegetation management, prescribed fire, and wildfire affects ET and other water budget components.
Ensemble of Models
OpenET provides data from multiple models that are used to calculate ET and also provides a single ET value, or “ensemble value,” from those models for each location. The models currently included are listed in the table on the right. Each model has its own strengths and limitations for different geographies, crops, and conditions.
Some models were developed for global-scale applications and others focus on local-scale irrigation management. Several models leverage gridded meteorological reference ET data for time integration between Landsat satellite overpass dates, and other models utilize coarser resolution data from geostationary satellites.
Currently, a key challenge for practitioners working with ET data is that there are multiple methods and sources of information, leading to confusion or hesitation about selecting the “correct” number. By working with an ensemble of ET models, the OpenET team can identify consistent biases within different models and work to explain and resolve those differences. We included all models in an intercomparison study to determine which provide the highest accuracy for different land cover types, regions, and seasons. We used the results from the intercomparison study to calculate a single value from the ensemble of models, taking advantage of the ensemble of ET values while first identifying and removing outliers.
Current Ensemble of ET Models
|Model Acronym||Model Name||Primary References|
|ALEXI/DisALEXI||Atmosphere-Land Exchange Inverse model/ALEXI disaggregation||Anderson et al., 2007; Anderson et al., 2018|
Evapotranspiration at high
Resolution with Internalized
|Allen et al., 2005; Allen et al., 2007|
|PT-JPL||Priestley-Taylor Jet Propulsion Laboratory||Fisher et al., 2008|
|geeSEBAL||Google Earth Engine Surface Energy Balance Algorithm for Land||Bastiaanssen et al., 1998; Teixeira et al., 2009; Laipelt et al., 2021|
|SIMS||Satellite Irrigation Management Support||Melton et al., 2012; Pereira et al., 2020|
|SSEBop||Operational Simplified Surface Energy Balance||Senay et al., 2014; Senay et al., 2018|
Another advantage of using an ensemble of models is that we expect to see a range of ET values across the ensemble. This can provide a useful measure of uncertainty of the remotely sensed ET data where no ground-based measurements exist. In addition, the range across the ensemble allows OpenET users to evaluate individual model estimates of ET against the range of values from OpenET.
For OpenET, a simple yet robust approach was chosen where the single ensemble ET estimate is computed at each time step as the simple arithmetic average after outlier ET estimates are removed. Outlier ET estimates are removed using the Median Absolute Deviation (MAD) method initially developed by Carl Friedrich Gauss, and more recently rediscovered and popularized. The MAD is a measure of scale, or spread of the data, based on the median of the absolute deviations from the median of the distribution. It has been described as “the single most useful ancillary estimate of scale” since it overcomes many limitations of more common standard deviation and interquartile approaches for identifying outliers. A refinement was added to the MAD outlier detection approach to account for the small size of the OpenET ensemble of models. Rather than exclude all models that may be flagged as outliers, a minimum of four models was always retained to calculate the single ensemble value. This approach still consistently eliminates outliers, while also taking advantage of an ensemble of models to improve the accuracy of ET estimates.
From close inspection of the OpenET ensemble average, median, and individual model ET values across space and time, along with comparisons to ground-based data from flux towers and weighing lysimeters, it is clear that each model can produce erroneous ET estimates, and that these errors can be random or systematic. In calculating the ensemble, the OpenET team applied the Median Absolute Deviation (MAD) approach to identify and eliminate model outliers at the pixel level and at monthly time steps. Based on the OpenET team’s experience, and results of the intercomparison and accuracy assessment to date, the ensemble average value appears to provide the most reliable and stable estimate of ET for expansive regions with well-watered crops, and for many natural land cover types. Examples include most of California’s Central Valley and Delta, and most agricultural regions in the Midwest. However, from the limited number of cropland in-situ flux stations located in arid and semi-arid environments, it is evident that some models have a systematic low bias for smaller agricultural areas in arid regions, and the MAD outlier filtering approach does not filter outliers as desired due to the large range in model estimates, resulting in a low bias for the ensemble ET value. These areas are often indicated by a wide range of ET values across the ensemble of ET models for the majority of fields within a region. Over the coming months, the OpenET team will continue to conduct additional research in these more challenging settings and develop a Best Practices Manual that will provide more region and application specific guidance. The ensemble value is likely to evolve in the coming year as the team conducts additional research and designs more region-specific approaches for calculation of the ensemble ET value. We strongly encourage users to rely upon their knowledge of local conditions in applying the ensemble ET value, or selecting a single model or subset of models for use in their application. When the Best Practices Manual is complete, it will be made prominently available on the OpenET website.
It is important to keep in mind that the data from OpenET do not represent a direct measurement of ET but rather an estimate, or calculation, of ET based on the measurement of other variables in the environment. Our goal is to be as transparent as possible about the quality of satellite images and other data sets, methodologies, and equations used in the calculations. The results from the intercomparison study will be used to inform the overall accuracy for the data from OpenET.
Satellite-based ET Data
OpenET provides satellite-based estimates of the total amount of water that is transferred from the land surface to the atmosphere through the process of evapotranspiration (ET). This is also referred to as ‘actual ET’, since it represents an estimate of the actual amount of ET that occurred. OpenET also provides access to data for grass reference ET (ETo), which is the evapotranspiration from a uniform grass surface that is well watered. ETo is calculated from weather data and widely used as an input to irrigation scheduling, and is included as an important benchmark to evaluate and interpret the satellite-based ET. However, it is the actual ET, estimated using satellite-driven models, that quantifies the amount of water that is removed from the water system through ET.
Direct measurements of ET in the natural environment, such as from weighing lysimeters, are costly, difficult to obtain, and therefore usually limited to a few research sites across the western US. ET data from OpenET do not represent direct measurements of ET; rather they are estimates based on satellite, meteorological, soil, and vegetation datasets that are used within state-of-the-art ET models. While they are generally accurate and consistent, these data and models have limitations, particularly during times of consistent and dense cloud cover, and within complex topography.
The majority of the models included in the OpenET platform are based on the energy balance approach. This approach relies on satellite measurements of spectral reflectance (the amount of light energy reflected off the land surface measured in specific wavelengths) and surface temperature combined with ground-based measurements of other key variables that govern the surface energy exchange. The only component of the surface energy balance that is not directly or indirectly measured by satellites and weather stations is ET, leaving it as the “residual” variable to solve for by applying a model or algorithm such as the ones available through the OpenET platform. For a brief overview of Landsat surface reflectance and temperature measurements and how these measurements are used for estimating ET, please see this NASA video: https://youtu.be/c6OmfYEgzA0.
The discrepancy between the various energy balance approaches primarily relates to different methods for correcting for atmospheric instability, interpolation between satellite images, and pre-processing of satellite and meteorological inputs to the models. A key strength of satellite-driven energy balance approaches is that they provide a consistent and accurate estimate of actual ET from each field and have been shown to be almost as accurate as ground-based data in many studies. One key limitation for many of these models is that, traditionally, they have typically required supervised operation and calibration by an expert to achieve maximum accuracy. The OpenET team has worked hard to overcome this limitation and has now fully automated these approaches.
SIMS, or the Satellite Irrigation Management Support system, relies instead on a reflectance-based approach. SIMS converts the normalized difference vegetation index (NDVI), a satellite-based measure of the density and health of vegetation, to crop fractional cover, the fraction of the ground that is covered by vegetation. It then combines this with estimated canopy height and a stomatal control factor (to account for physiological differences between crops) to calculate a crop coefficient for every 0.22-acre pixel in each satellite scene. Combining crop-coefficient data with California Irrigation Management Information System (CIMIS) data, gridMET gridded meteorological data, or other ground-based estimates of reference ET facilitates daily estimation of actual ET following approaches recommended by the Food and Agriculture Organization of the United Nations (FAO). Strengths of this approach include estimation of irrigation requirements under well-watered conditions, full automation, and the ability to integrate data from multiple satellites to increase the number of available observations. Limitations include a limited ability to detect short-term or intermittent deficit irrigation or stressed rainfed crop conditions.
OpenET builds upon decades of investment by NASA, USGS, the National Oceanic and Atmospheric Administration (NOAA) and the European Space Agency (ESA) to develop, launch, and operate a constellation of Earth-observing satellites, and to establish the ground data systems required to capture, process, store, and distribute satellite data. All of the models in the OpenET ensemble leverage data from Landsat satellites to produce field-scale ET estimates, which is the primary satellite dataset used by the OpenET platform. However, multiple models implemented within the OpenET framework also integrate data from other satellites, including GOES, Sentinel-2, Suomi NPP, Terra, and Aqua, to produce ET data at a range of spatial and temporal scales.
OpenET has conducted one of the largest intercomparison and accuracy assessments of field-scale, satellite-based ET models. Satellite-derived OpenET data were compared against in-situ ET estimates collected by 136 flux stations throughout the continental U.S., and data records from four weighing lysimeters.
The flux stations use eddy covariance methods to measure the exchanges of carbon dioxide, water vapor, and energy between the land surface terrestrial ecosystems and the atmosphere. These stations are important because they provide ground-based, or in-situ ET datasets for comparison against a very specific set of locations with known land use types and characteristics. These data are derived directly from micrometeorological measurements used to calculate water and energy fluxes from the land surface. The 139 ground-based flux stations used in the OpenET analysis include Ameriflux sites and additional sites located within agricultural fields operated by collaborators at USDA, USGS, University of California Cooperative Extension, and other university partners. We developed an extensive, automated screening process to identify any outliers or gaps in the instrument measurement record, and we have reviewed the sites using aerial imagery as well as wind speed and direction data from the towers to ensure that the location and ‘foot print’ of the station is representative of the surrounding field or land cover type.
For croplands, the ensemble performed as well as or better than any individual model across most accuracy metrics, with a mean absolute error (MAE) for the growing season of 13.2% (80.3 mm), and a MAE value of 16.6% (15.6 mm) at monthly time steps. The mean bias error is less than 4% for both the growing season and monthly averages, indicating that many of the errors are random errors, and the overall bias in the OpenET ensemble values are minimal for croplands.
However, from the limited number of cropland in-situ flux stations located in very arid environments, it is evident that some models have a systematic low bias for smaller agricultural areas in arid regions, and the MAD outlier filtering approach does not filter outliers as desired due to the large range is model estimates. This may result in a low bias in the ensemble average. These areas are often indicated by fields with a wide range of ET estimates across the ensemble of ET models.
For natural land cover types, there is a bigger range in the accuracy metrics, but values for slope and bias errors are still reasonable for all land cover types. We did see a positive bias in evergreen and mixed forests, highlighting some key areas for future research and continued refinement. Additional information and technical details are available in the OpenET Intercomparison and Accuracy Assessment Report.
An important part of the mission for the OpenET non-profit is to continue advancing the underlying science. Our goal is to continue to improve the ensemble and individual model accuracies over time.
Development of the OpenET platform is supported by the S. D. Bechtel, Jr. Foundation, Gordon and Betty Moore Foundation, The Walton Family Foundation, The Keith Campbell Foundation for the Environment, Lyda Hill Philanthropies, The Laney and Pasha Thornton Foundation, the Water Funder Initiative, Desert Research Institute Maki Endowment, Windward Fund, the North, Central, and South Delta Water Agencies, the NASA Western Water Applications Office and the NASA Applied Sciences Program. In-kind support is provided by partners in the agricultural and water management communities, Google Earth Engine, and the Water Funder Initiative. OpenET will need to secure additional funding to cover the transition of OpenET to a non profit organization and to allow the team sufficient time to implement funding models, licensing contracts, and other aspects critical to supporting the platform long term.
One goal of OpenET is to continually provide the best available science-based estimates of ET. This requires a team of researchers and programmers to maintain and update the code as the science and underlying input data improves and as cloud computing and the Google Earth Engine platform evolve. We also expect ongoing computing and data charges and administrative and management needs associated with initiating and maintaining contracts and coordinating efforts between OpenET’s scientific community and its users. OpenET also anticipates the need to provide user support and the development of training resources for the OpenET user community.