Climate Engine maps and time series help scientists and managers see and study earth observation data

By Justin Huntington — Research Professor of Hydrology at the Desert Research Institute, and John Abatzoglou — Associate Professor of Geography, University of Idaho

The archives of satellite remote sensing and spatial climate data run into the millions of gigabytes. If you’re an environmental scientist, downloading, processing, and analyzing these data demands a huge amount of time that could otherwise be spent making scientific discoveries. Even if you need to access just a few pieces of data, you often have to download the entire dataset and write programs to process it — which becomes really cumbersome when the datasets are in different formats and you have to stitch multiple datasets together. Scientists don’t want to become part-time software engineers to study the environment. We’d rather just do our science.

Our team at Desert Research Institute and the University of Idaho launched the web application Climate Engine in 2017 largely because we were in that same boat: researchers trying to make sense of the petabytes of geospatial data on climate, drought, and natural resources in an efficient way. We knew there had to be a better way to process and visualize the data.

The better way turned out to be the Google Earth Engine platform. Using the Earth Engine APIs, we built Climate Engine so that programmers and non-programmers alike can quickly and easily process satellite and climate data, and create maps and time series viewable in a web browser. The data processing happens in the background, with Earth Engine doing the heavy lifting. Users can choose from multiple climate and remote sensing data type options, variables (such as precipitation, temperature, evapotranspiration, and various vegetation and drought indices), calculations (such as average, difference from average, or percent of average), analysis periods, and can visualize and download maps and time series of the resultant output in real-time.

Climate Engine web application, showing the land surface temperature anomaly from mid-February to mid-March, 2019.

If you’re trying to study how climate or natural resource management affects the land surface, nothing beats the combination of maps and time series for unpacking the data. When scientists and government officials use Climate Engine to visualize data that used to require programmers and costly consultants, they can make more efficient science based assessments and decisions by having both climate and satellite earth observation data readily processed for analysis and interpretation with a few clicks on a web-browser.

If you’re trying to study how climate or natural resource management affects the land surface, nothing beats the combination of maps and time series for unpacking the data.

Maps and time series in seconds, more power for scientists and decision makers

The National Oceanic and Atmospheric Administration’s National Integrated Drought Information System (NIDIS) is supporting Climate Engine for assessing regional and field scale satellite-based vegetation condition and climate maps to improve drought monitoring and early warning systems. With several types of regional scale data being used currently, the use of Climate Engine by U.S. Drought Monitor authors, regulatory agencies, and ranchers and farmers adds local scale information for assessing field scale impacts — such as vegetative drought (short-term soil moisture deficits) and hydrologic drought (long-term water deficits and low river flows). The vegetation index anomaly map below is a nice example of how relatively high spatial resolution (30 m) Landsat satellite imagery can be used to monitor farmlands and rangelands at the same time. In this case, the map shows opposite signs of drought in 2015 across northwestern Nevada, where the rangelands were greener than average due to above average spring and summer rain and lower than average evaporative demand (as can be easily seen by looking at the Standardized Precipitation Index and Evaporative Demand Drought Index for the area), but the farmlands were fallow due to below average mountain snowpack for many years in a row and little to no water available in the river for diversion and surface water irrigation.

Map of the Landsat Normalized Difference Vegetation Index (NDVI) anomaly for April 1 — October 31, 2015 relative to the Landsat-5, -7, and -8 long term average (1984–2018) for Lovelock, Nevada. The map shows fallowed and stressed alfalfa fields shown in red, and vigorous rangeland vegetation shown in green. Normally these alfalfa fields are well-watered and the surrounding rangelands are dry. However, due to the 2012–2015 hydrologic drought in the region and groundwater pumping upstream, the Lovelock Irrigation District received no irrigation water from the Humboldt River during the growing season of 2015. At the same time, spring and summer rains were well above average, resulting in abundant rangeland vegetation. This field scale map highlights the challenge that regional scale maps like the U.S. Drought Monitor have in depicting different signs of drought at the same time and in the same region (i.e. hydrologic vs. vegetative drought or lack thereof).

For many organizations, assessment of climate and vegetation data is critical to spurring action. The Bureau of Land Management’s Nevada office uses Climate Engine maps to better understand how vegetation greenness pairs with precipitation levels. For example, if an area has received ample precipitation but the vegetation is stressed, the cause isn’t lack of precipitation; it could be overgrazing by livestock. Climate Engine maps and time series are providing BLM staff new place-based information for assessing grazing permits. Because we pull in many data sets that span 30+ years and are spatially explicit, our maps and time series tell a rich story about what’s happening on the ground.

“Climate Engine provides specialists with the opportunity to essentially go back in time and see how our Western landscapes have changed due to changes in climate over the past few decades. It‘s able to quantitatively analyze years of data in a matter of a few seconds.”

— Sarah Peterson, Bureau of Land Management, Nevada State Office, Soil-Water-Air-Riparian Programs Lead

In Northern California’s Sierra Nevada mountains, scientists with the California Tahoe Conservancy, Tahoe Regional Planning Agency, and the U.S. Forest Service are using Climate Engine to track the greenness and wetness of the area’s meadows, which support many plant and wildlife species. One of the common questions that managers and conservation scientists ask is if meadow restoration actually results in an increase in vegetation vigor. Field based assessments of the efficacy of meadow restoration require many years of observations both prior to and following restoration, and can be impractical and prohibitively expensive to implement across spatial scales. Alternatively, high-spatial resolution remote sensing products can provide an effective means of providing answers. Let’s take a quick look using Landsat and climate data. The map below shows a portion of the Sierra Nevada meadows layer available in Climate Engine.

Sierra Nevada meadow polygons (black areas) shown with a median Landsat-8 false-color infrared map based on images acquired from June, July, and August of 2018.

Now let’s zoom in and pick meadow in South Lake Tahoe, California that’s had multiple restoration activities over the last 20 years with restoration activities beginning in the mid 1990s. The new creek bed sits about a meter higher than the old one, allowing for better distribution of water and more vegetation that reduces sediments and nutrients from flowing to the lake.

Trout and Cold Creek meadow complex located in South Lake Tahoe, California, selected to make a time series graph of the spatially averaged median Landsat NDVI for the period of July — September each year from 1984–2018 — shown as “Variable 1” in the user interface. Time series “Variable 2” (not shown) was selected to be the total October — September gridMET precipitation each year. These meadows have undergone multiple restoration projects over the years, beginning in the mid-1990s.

Once the datasets, calculations, and time series options are set and we click the green button, Earth Engine requests are made to return summary time series that are graphed with data download options. The time series of late-summer meadow NDVI (bars) and water year precipitation (line) illustrate that NDVI has been substantially higher post-restoration than pre-restoration. You can also see that the NDVI generally tracks with annual precipitation. The fact that low NDVI values are on the rise, even during multi-year drought, is a great sign that meadow restoration efforts are paying off.

Annual time series of July-September mean Landsat NDVI (bars) plotted with October-September gridMET precipitation totals. Data is made available for download as a list or multiple file options such as .csv or .xls.

Before agency scientists found Climate Engine, they were spending over $100,000 to extract Landsat vegetation data for the meadows. Not only is that an expensive way to get information, it’s slow, since it can take months to receive reports. Now these agencies and conservation groups are accessing these same data in Climate Engine to assess both meadow restoration and degradation in just a couple of minutes and can use that money for other projects (see this post for a degradation example).

Besides saving money and time, Climate Engine gives scientists more ownership over their research. Instead of ceding work to outside consultants, scientists can do the work of gauging meadow health internally. That means they can make more rapid and cost effective conclusions about which meadows are in good health, or at risk and may need restoration.

Perhaps one of the most powerful uses of Climate Engine centers around water resources and agriculture. The following are some testimonials given a few weeks after workshops we held:

“The historical remote sensing data and the tools in Climate Engine helped in a recent water right dispute by producing information showing the historical irrigation of a water right. Climate Engine analyzed massive amounts of remote sensing data in a few minutes that would have taken weeks without Climate Engine”

— William Kramber, Senior Remote Sensing Analyst, Idaho Dept. of Water Resources.

Contrast between irrigated and non-irrigated lands along the Snake River Snake near Burley, Idaho using Landsat-8 maximum NDVI for June, July, August of 2018.

“Climate Engine has proven to be a useful tool in assisting the Southern Nevada Water Authority (SNWA) with monitoring the agricultural land fallowing programs it administers in Southern Nevada. The analyses and data products available in Climate Engine not only enables SNWA to ground truth and monitor fallowed fields, but also to coordinate with local stakeholders to assure that both SNWA and parties still irrigating can cooperatively utilize their respective water rights in the most effective manner possible”

— Sean Collier, Hydrologist, Southern Nevada Water Authority

In a recent Nevada State Engineer’s Office water right ruling, consultants used Climate Engine vegetation greenness maps to show that someone who wanted to expand their irrigated acreage hadn’t actually watered their lands at all over a five-year-period — and therefore, should forfeit their water rights. The State Engineer’s Office agreed based on the clear results from Climate Engine and clearly defined law regarding forfeiture after 5 years of non-use.

A couple of years after launching Climate Engine, it’s great to see how people are using Climate Engine maps and time series in many different ways related to ecology, agriculture, water, and litigation. Data discovery and transparency is no doubt driving change and improving decision support- with hopefully much more to come.

Coming up: API and trend maps showing climate impact over time

This year, we hope to release a Climate Engine Application Programming Interface (API) to allow for machine to machine queries, and adding more useful maps by displaying trends in climate and vegetation over long periods of time, instead of just snapshots. People will be able to generate maps that show changes in vegetation affected by changing climate, land, or water management. With trend maps, scientists can offer more insights on how to manage resources and provide early warnings on climate’s impact on everything from agriculture to snowfall levels.

Climate Engine was originally funded by a Google Earth Engine Faculty Research Award in 2014, and has since been supported by NIDIS, U.S. Bureau of Land Management, NASA, U.S. Geological Survey, and the California Landscape Conservation Partnership — a great example of public-private partnership. For details on Climate Engine see our publication in the Bulletin of the American Meteorological Society.