This paper illustrates the lessons learned as we applied the U-Net3+ deep learning model to the task of building an operational model for predicting wildfire occurrence for the contiguous United States (CONUS) in the 1–10-day range. Through the lens of model performance, we explore the reasons for performance improvements made possible by the model. Lessons include the importance of labeling, the impact of information loss in input variables, and the role of operational considerations in the modeling process. This work offers lessons learned for other interdisciplinary researchers working at the intersection of deep learning and fire occurrence prediction with an eye toward operationalization.
Part I: Improving Wildfire Occurrence Prediction for CONUS Using Deep Learning and Fire Weather Variables
The purpose of this research is to build an operational model for predicting wildfire occurrence for the contiguous United States (CONUS) in the 1–10-day range using the U-Net 3+ machine learning model. This paper illustrates the range of model performance resulting from choices made in the modeling process, such as how labels are defined for the model and how input variables are codified for the model. By combining the capabilities of the U-Net 3+ model with a neighborhood loss function, fractions skill score (FSS), we can quantify model success by predictions made both in and around the location of the original fire occurrence label. The model is trained on weather, weather-derived fuel, and topography observational inputs and labels representing fire occurrence. Observational weather, weather-derived fuel, and topography data are sourced from the gridded surface meteorological (gridMET) dataset, a daily, CONUS-wide, high-spatial-resolution dataset of surface meteorological variables. Fire occurrence labels are sourced from the U.S. Department of Agriculture’s Fire Program Analysis Fire-Occurrence Database (FPA-FOD), which contains spatial wildfire occurrence data for CONUS, combining data sourced from the reporting systems of federal, state, and local organizations. By exploring the many aspects of the modeling process with the added context of model performance, this work builds understanding around the use of deep learning to predict fire occurrence in CONUS.
Our work seeks to explore the limits to which deep learning can predict wildfire occurrence in CONUS with the ultimate goal of providing decision support to those allocating fire resources during high fire seasons. By exploring with what accuracy and lead time we can provide insights to these persons, we hope to reduce loss of life, reduce damage to property, and improve future event preparedness. We compare two models, one trained on all fires in the continental United States and the other on only large lightning fires. We found that a model trained on all fires produced a higher probability of fire.
- Award ID(s):
- 2019758
- PAR ID:
- 10531202
- Publisher / Repository:
- American Meteorological Society
- Date Published:
- Journal Name:
- Artificial Intelligence for the Earth Systems
- Volume:
- 3
- Issue:
- 3
- ISSN:
- 2769-7525
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
more » « less
Abstract -
Viegas, Domingos Xavier (Ed.)In this paper, we present an integrated wildland fire forecasting system based on combining a high resolution, multi-scale weather forecasting model, with a semi-empirical fire spread model and a prognostic dead fuel moisture model. The fire-released heat and moisture impact local meteorology which in turn drives the fire propagation and the dead fuel moisture. The prognostic dead fuel moisture model renders the diurnal and spatial fuel moisture variability. The local wind and the fuel moisture variation drive the fire propagation over the landscape. The sub-kilometer model resolution enables detailed representation of complex terrain and small-scale variability in surface properties. The fuel moisture model assimilates surface observations of the 10h fuel moisture from Remote Automated Weather Stations (RAWS) and generates spatial fuel moisture maps used for the fire spread computations. The dead fuel moisture is traced in three different fuel classes (1h, 10h and 100h fuel), which are integrated at any given location based on the local fuel description, to provide the total dead fuel moisture content at the fire-model grid, of a typical resolution of tens of meters. The fire simulations are initialized by a web-based control system allowing a user to define the fire anywhere in CONUS as well as basic simulation properties, such as simulation length, resolution, and type of meteorological forcing for any time meteorological products are available to initialize the weather model. The data is downloaded automatically, and the system monitors execution on a cluster. The simulation results are processed while the model is running and displayed as animations on a dedicated visualization portal.more » « less
-
more » « less
Abstract Satellite‐based Fire radiative power (FRP) retrievals are used to track wildfire activity but are sometimes not possible or have large uncertainties. Here, we show that weather radar products including composite and base reflectivity and equivalent rainfall integrated in the vicinity of the fires show strong correlation with hourly FRP for multiple fires during 2019–2020. Correlation decreases when radar beams are blocked by topography and when there is significant ground clutter (GC) and anomalous propagation (AP). GC/AP can be effectively removed using a machine learning classifier trained with radar retrieved correlation coefficient, velocity, and spectrum width. We find a power‐law best describes the relationship between radar products and FRP for multiple fires combined (0.67–0.76
R 2). Radar‐based FRP estimates can be used to fill gaps in satellite FRP created by cloud cover and show great potential to overcome satellite FRP biases occurring during extreme fire events. -
The size and frequency of wildland fires in the western United States have dramatically increased in recent years. On high-fire-risk days, a small fire ignition can rapidly grow and become out of control. Early detection of fire ignitions from initial smoke can assist the response to such fires before they become difficult to manage. Past deep learning approaches for wildfire smoke detection have suffered from small or unreliable datasets that make it difficult to extrapolate performance to real-world scenarios. In this work, we present the Fire Ignition Library (FIgLib), a publicly available dataset of nearly 25,000 labeled wildfire smoke images as seen from fixed-view cameras deployed in Southern California. We also introduce SmokeyNet, a novel deep learning architecture using spatiotemporal information from camera imagery for real-time wildfire smoke detection. When trained on the FIgLib dataset, SmokeyNet outperforms comparable baselines and rivals human performance. We hope that the availability of the FIgLib dataset and the SmokeyNet architecture will inspire further research into deep learning methods for wildfire smoke detection, leading to automated notification systems that reduce the time to wildfire response.more » « less
-
Abstract Patterns of energy and available moisture can vary over small (<1 km) distances in mountainous terrain. Information on fuel and soil moisture conditions that resolves this variation could help to inform fire and drought management decisions. Here, we describe the development of TOPOFIRE, a web-based mapping system designed to provide finely resolved information on soil water balance, drought, and wildfire danger information for the contiguous United States. We developed 8-arc-second-resolution (~250 meter) daily historical, near real-time, and 4-day forecast radiation, temperature, humidity, and snow water equivalent data and used these grids to calculate a suite of drought and wildfire danger indices. Large differences in shortwave radiation and surface air temperature with aspect contribute to greater snow accumulation and delays in melt timing on north-facing slopes, delaying fuel conditioning on shaded slopes. These datasets will help advance our understanding of the role of topography in wildland fire spread and ecological effects. Integration with national programs like the Wildland Fire Assessment System, the Wildland Fire Decision Support System, and drought early warning systems could support more proactive management of wildland fires and refine the characterization of drought in mountainous regions of the United States.more » « less
