Wildfires consumed more than 3.5 million hectares in the United States in 2018, and federal fire suppression costs topped US$3 billion. These fires destroyed more than 18,000 residences and caused the deaths of at least 85 people. Wildfire damages like these are not unique to the United States; they are a threat in many nations. Researchers from across the globe and across multiple scientific disciplines are working to improve fire danger rating systems to help protect natural resources and human health and safety.
One new concept emerging as a valuable contribution to this effort is the integration of soil moisture information as a predictor of wildfire probability. Soil moisture, particularly within the zone where plant roots reside, is a key link between weather conditions, such as precipitation and temperature, and the characteristics of the live vegetative “fuel bed,” which include fuel moisture and fuel loads (weight of fuel per unit area). These dynamic vegetation characteristics, which strongly influence wildfire probability, can be challenging to model and monitor at relevant spatial and temporal scales using field data. Optical remote sensing of fuel moisture also presents challenges; for example, remote sensing models that predict live-fuel moisture contain relatively large margins of error that vary by vegetation type.
Soil moisture monitoring capabilities, in contrast, have been steadily growing because of the development of in situ networks and dedicated satellites. Researchers have intuitively understood the relationships between soil moisture, fuel conditions, and wildfire occurrence for a long time. However, the increasing availability of soil moisture information is creating significant opportunities to quantify these relationships and incorporate them into new or existing weather-based fuel moisture models and fire danger rating systems.
Scientists and Practitioners in Dialogue
To explore these opportunities and envision next steps, 34 researchers and fire management practitioners participated in a 1-day workshop held in April in association with the 6th International Fire Behavior and Fuels Conference, hosted by the International Association of Wildland Fire. The workshop’s aims were to identify the needs and interests of the fire management community and to explore how soil moisture information could be used in wildfire modeling, risk assessment, planning, and decision support tools.
Invited speakers discussed the current trends and status of wildfire occurrence, impacts, and response; fire modeling and fire danger ratings; soil moisture monitoring and modeling; and linkages between soil moisture, fuel bed characteristics, and wildfires. Small group discussions identified practitioners’ needs related to fire management decision-making at local, regional, and national scales.
Raising Soil Moisture Awareness
These discussions revealed that fire management decision-makers (approximately one third of the attendees) are not generally seeking soil moisture information. They often are unaware of recent research indicating the utility of such information in fire danger prediction. Instead, these decision-makers often consult decades-old drought indices (e.g., the Keetch-Byram drought index). These indices were designed to roughly mimic soil moisture dynamics, even though research has shown that direct soil moisture observations are more effective in fire danger assessment.
Research has also shown, and stakeholders intuitively understand, that soil moisture represents the land’s integrated “memory” of recent weather conditions and could therefore provide valuable early warnings of rising fire danger. To facilitate greater use of soil moisture information in fire management, workshop participants identified a need to increase the quantity and quality of available soil moisture information at relevant spatiotemporal scales and to determine how soil moisture can be integrated effectively into existing fire danger rating systems. Furthermore, stakeholders expressed a need to consider both statistical and mechanistic modeling approaches to develop a better understanding of the relationships between soil moisture and fuel moisture conditions.
A Case Study in Oklahoma
Oklahoma is among the top 10 U.S. states in wildfire risk and has one of the world’s oldest soil moisture monitoring networks. It therefore provides a natural laboratory for exploring relationships between soil moisture and fuel moisture. PhenoCam images (a series of time-lapse digital images from a fixed location) from a tallgrass prairie site in north central Oklahoma (Figure 1) show the stark contrast in vegetation conditions occurring during 2012, a drought year (Figure 1, left), and 2013, a year of near-average precipitation (Figure 1, right). The year 2012 was characterized by low soil moisture, shown in the graph in Figure 1 as the fraction of available water capacity (FAW). This condition led to fuel having a low moisture content, which created conditions of extreme fire danger. Several large wildfires ignited near the PhenoCam site in the days after the image on the left was captured, one of which—the Freedom Hills Fire—burned 24,000 hectares and destroyed more than 300 homes.
To improve advance warning of these types of wildfires and predict safe conditions for prescribed burning, Oklahoma’s fire danger information system, OK-FIRE, now includes soil moisture data as a metric available to the state’s fire management personnel (e.g., Figure 2).
Workshop participants noted that although this encouraging progress in Oklahoma illustrates the feasibility of using soil moisture information in fire danger ratings, key challenges remain. One of the foremost challenges is to overcome the limited number of existing in situ monitoring systems and improve the spatiotemporal resolution of satellite-based soil moisture monitoring systems.
Current Limitations and Promising Solutions
In situ measurements continue to be the “gold standard” for quantifying soil moisture for many applications, but monitoring stations are costly to install and maintain. As a result, many regions still lack in situ measurements, so estimates of soil moisture conditions must be interpolated, simulated, or derived from remote sensing estimates for unsampled areas, many of which are fire-prone landscapes. Operational simulation systems such as NASA’s North American Land Data Assimilation System (NLDAS) provide soil moisture estimates for central North America, but the 4-day lag between when the data are recorded and when estimates are made available and the spatial resolution of the data (one eighth degree) are not optimal for fire danger monitoring.
Advances in satellite remote sensing for soil moisture, such as the Soil Moisture Active Passive (SMAP) mission, suggest a bright future for global soil moisture monitoring; however, there are some important limitations to using SMAP data for wildfire danger prediction. For example, the record of soil moisture is too short for use in statistical modeling of wildfire probability. Also, remote sensing observes only the surface soil, not the root zone, and the spatial resolution is too coarse for localized analysis.
Researchers are actively pursuing ways to overcome the deficiencies in these current sources of soil moisture information. The coordinated National Soil Moisture Network, for example, is an ongoing initiative to combine data from a diverse collection of state and federal in situ networks in the United States to create a standardized national soil moisture data product. Likewise, organizers of this year’s workshop are part of an interdisciplinary team working to develop a near-real-time, high-resolution soil moisture simulation model that is based on measured soil moisture and suitable for fire danger monitoring at regional or national scales.
In addition, a new data product called SoilMERGE (SMERGE) provides retrospective root zone soil moisture estimates across the continental United States. This record stretches back to 1979, a period that is adequate for statistical modeling of wildfire probability and size. SMERGE is produced by merging root zone soil moisture simulations from Noah land surface model version 3.2 within NLDAS phase 2 with remotely sensed surface soil moisture retrievals from the multisensor European Space Agency Climate Change Initiative. This merging of two different methods harnesses and combines their complementary strengths to create a unified set of root zone soil moisture (0- to 41-centimeter depth) estimates, available on a daily time step, across multiple decades (1979–2018), for the continental United States at a one-eighth-degree scale. In situ soil moisture measurements are then used to validate SMERGE estimates. SMERGE has not been applied to wildfire probability mapping yet, but such improvements in the availability of long-term soil moisture data are key steps forward for analyzing the relationships with wildfire.
The Many Applications of Soil Moisture Data
The workshop participants identified specific areas in need of further research efforts, some of which are currently under way. Field research is being conducted in diverse environments to better understand the relationships between soil moisture and fuel moisture conditions. Researchers are exploring both statistical and mechanistic modeling approaches to overcome data scarcity in measured soil moisture, with the goal of integrating these data into existing fire danger rating systems. As researchers and fire management professionals continue to learn about soil moisture–wildfire relationships, momentum is building for enhanced spatial and temporal root zone soil moisture information and for inclusion of that information in regional and national fire danger ratings systems.
In addition to improving wildfire danger forecasting, soil moisture data can potentially support other important aspects of wildfire management and decision-making. Applications for this data can include use in planning and implementing prescribed fires under optimal soil and fuel moisture conditions to ensure maximum success, as well as in predicting fire behavior during resource benefit (managed) wildfires. Soil moisture data could also be used to identify postfire windows for restoration seeding. Similarly, soil moisture could help predict the spread of invasive plants that readily exploit early-season soil moisture (e.g., cheatgrass, red brome, and buffelgrass in the western United States) and increase fire hazard at the wildland-urban interface by altering fire regimes.
Soil moisture information is vital for numerous other applications that intertwine social well-being and natural resource management. Often, scientific advancements emerge by solving problems that were not the original target of investigation. Improved soil moisture information could generate a wide range of societal benefits through such applications as drought monitoring for agricultural applications, groundwater recharge estimation, ground saturation flooding, and seasonal streamflow forecasting. Root zone soil moisture products can also link with satellite remote sensing products to provide valuable interpretations of ecological phenomena related to fire (e.g., tree die-offs, beetle outbreaks).
More details of the soil moisture and wildfire workshop are available, including expanded takeaways, speaker presentations, and a video summarizing the workshop. A similar workshop will be organized in 2021 in Stillwater, Okla., to continue conversations about integrating soil moisture and wildfire information.
We thank Mikel Robinson and the International Association of Wildland Fire for their assistance with organizing the workshop and all participants for their contributions. Financial support was provided by U.S. Department of the Interior South Central Climate Adaptation Science Center grant G18AC00278, with partial support for the workshop provided by the U.S. Department of Agriculture’s Northern Plains Climate Hub through award NNX16AH30G from the NASA Climate Indicator and Data Products for the National Climate Assessments program. We thank our many collaborators, including site principal investigators and technicians, for their efforts in support of PhenoCam. The development of PhenoCam has been funded by the Northeastern States Research Cooperative, the National Science Foundation’s Macrosystems Biology program (awards EF-1065029 and EF-1702697), and the Department of Energy’s Regional and Global Climate Modeling program (award DE-SC0016011). We acknowledge additional support from the U.S. National Park Service’s Inventory and Monitoring Program and the USA National Phenology Network (grant G10AP00129 from the U.S. Geological Survey) and from the USA National Phenology Network and the North Central Climate Science Center (cooperative agreement G16AC00224 from the U.S. Geological Survey). Additional funding, through the National Science Foundation’s Long-Term Ecological Research (LTER) program, has supported research at Harvard Forest (DEB-1237491) and Bartlett Experimental Forest (DEB-1114804). We also thank the U.S. Forest Service’s Air Resource Management program and the National Park Service Air Resources program for contributing their camera imagery to the PhenoCam archive.
Matthew R. Levi ([email protected]), University of Georgia, Athens; Erik S. Krueger, Oklahoma State University, Stillwater; Grant J. Snitker, University of Georgia, Athens; Tyson Ochsner, Oklahoma State University, Stillwater; Miguel L. Villarreal, U.S. Geological Survey, Menlo Park, Calif.; Emile H. Elias, Agricultural Research Service, U.S. Department of Agriculture (USDA), Las Cruces, N.M.; and Dannele E. Peck, Agricultural Research Service, USDA, Fort Collins, Colo.