Why is it so hard to forecast things such as crop yield, disease outbreaks, or water quality on the scales that matter most to humans? Is something missing from our environmental models?
Environmental models have improved dramatically over the past decades as our understanding of environmental systems and how they interact has grown. And on a global scale or for longer timelines, many of our models work very well. For example, observations of rising global average temperatures over the last decade have been consistent with projections from many climate change models.
On smaller scales, however, our forecasts often break down. Our models accurately predict that rising temperatures will result in an increase in hurricane activity in the Atlantic over time. They are much less accurate in predicting how many hurricanes to expect in a given season or where an emerging hurricane will eventually make landfall. Similarly, the weather predictions from our local meteorologist can change substantially from day to day, making it hard to plan outdoor activities with confidence more than a day or two ahead.
Why is it so hard to predict next Tuesday's weather or this season's growing conditions? Part of the problem may be that we are missing something in our observations that is needed to inform better models. A new study out of the University of Wisconsin—nicknamed "CHEESEHEAD19," for Chequamegon Heterogeneous Ecosystem Energy-balance Study Enabled by a High-density Extensive Array of Detectors 2019—seeks to explore and resolve some of the inconsistencies between local observations and the data that drive our environmental models.
The study, led by University of Wisconsin–Madison Professor of Atmospheric and Oceanic Sciences Ankur Desai, is funded through a grant by the National Science Foundation. Stefan Metzger, a Battelle scientist and part of the NEON project team, is acting as a co-Principal Investigator and will be developing the algorithms needed to analyze CHEESEHEAD data across different spatial and temporal scales.
Bridging the Scale Gap Between Models and Observations
The CHEESEHEAD project seeks to improve environmental forecasting at the meso-scale—that is, at the local or regional level—by bridging the "scale gap" between environmental models and the observations collected on the ground.
Environmental models typically are built using data at scales of tens of square kilometers to hundreds of square kilometers. Data gathered on the ground through flux towers and other observation methods is much finer scale, on the order of centimeters, meters or hundreds of meters. The observations made at this scale are often different than would be predicted using environmental models built on larger data scales.
Project CHEESEHEAD will deploy 18 or 20 observation towers in a 10x10 square kilometer section of the Chequamegon-Nicolet National Forest in northern Wisconsin. The towers will gather detailed data about energy fluxes at the surface-atmosphere boundary layer, which will be supplemented by remote sensing data gathered by airborne remote sensing platforms. The grid of flux towers at the Chequamegon-Nicolet site could help resolve some of the discrepancies between models and local observation.
The 10 km x 10 km plot represents a scale often used for environmental modeling. Point observations taken from within a plot of this size can vary widely depending on exactly where they are taken. Is the area observed covered by thick vegetation or bare soil? Is it in the middle of a lake? Models typically only have a limited data set from within each 10 km x 10 km area that is meant to represent the whole. But because of wide variations in topography and ecosystem characteristics within each square, models built on this scale are highly likely to have inaccuracies built into them.
CHEESEHEAD will look at data collected in the study area across multiple spatial and temporal scales. Researchers will analyze the data to identify discrepancies and look for potential biases that may influence our current environmental models.
Exploring an Energy Imbalance
One specific hypothesis that the study will explore is that our environmental observations are missing something when it comes to measuring energy going into and out of environmental systems.
The first law of thermodynamics is inescapable: energy can neither be created nor destroyed. That means when energy goes into a system—such as an ecosystem—the same amount of energy must either come back out of the system or be retained somewhere inside it.
But when making observations of earth systems with flux towers, energy appears to be missing: more energy is measured coming into the system from the sun than is measured going out of the system through radiation or retained by living organisms and non-living elements such as soil, rock and water. This mismatch between expected and actual observations may point to a problem with how we are using these observations to benchmark our models that could explain why the forecasts we build from them are so often inaccurate.
The CHEESEHEAD project will use the finer scale data gathered using multiple towers to provide a much more detailed and accurate view of energy fluxes over the entire 10 km x 10 km grid. Analyzing these data will provide a clearer picture of how differences in vegetation, soil types, moisture, topography, bodies of water and other variables impact the flow of energy into and out of an ecosystem.
CHEESEHEAD and the NEON Project
While the study does not take place at a NEON site, NEON does have a nearby terrestrial field site, STEI that collects field observations and airborne data in Chequamegon-Nicolet Forest. The analysis will also draw on the expertise of the NEON project team. Battelle was awarded a subcontract under the NSF grant to assist with the experimental design and develop novel algorithms for analysis of the data. Stefan Metzger, Science Lead for the NEON Surface-Atmosphere Exchange group, is heading up these efforts.
The project may also utilize one of the NEON airborne observation platforms (AOPs) and up to two Mobile Deployment Platforms (MDPs) through the NEON Assignable Assets Program.
Data for the CHEESEHEAD project will be collected in the summer and fall of 2019. Analysis and algorithm development will be conducted over the next two years and is expected to be complete by the summer of 2021.
These data and algorithms will allow researchers to compare and analyze observations at different scales and answer critical questions that will inform future environmental studies and models. Some of these questions include:
- How many observation towers are needed to collect unbiased surface-atmosphere exchange data for a given area?
- How do data collected through remote sensing compare to observations collected by the towers?
- Can correlations be made between remote sensing data and tower observations so that remote sensing data can be used as a reasonable proxy in operational forecasting?
Answering these questions could enable researchers to detect and correct biases in current environmental models and make better predictions for shorter time scales and smaller areas. These types of predictions are invaluable for agricultural forecasting, disaster planning, weather prediction and other human-scale questions and challenges.