An official website of the United States government
Abstract 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. Significance StatementOur 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.
more »
« less
Part II: Lessons Learned from Predicting Wildfire Occurrence for CONUS Using Deep Learning and Fire Weather Variables
Abstract 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.
more »
« less
- Award ID(s):
- 2019758
- PAR ID:
- 10531201
- 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
-
-
Abstract Wind turbine reliability monitoring and prediction are crucial for early failure detection, proactive maintenance, performance optimization, and cost reduction, especially as many utility-scale turbines near the midpoint or end of their operational lifespan. We proposed a wind turbine subsystem reliability prediction model to facilitate week-ahead forecasts of the occurrence, duration, and type of potential failures or probability of downtime for major turbine components, including the blade, hub, gearbox, generator, inverter, electrical, and control subsystems. Specifically, we developed an Instance-Normalization Decomposition Linear (IN-DLinear) model, grounded in deep time-series modelling theory. The distribution shifts in turbine state features across the training, validation, and test datasets, as well as across various time scales, were effectively mitigated with IN. The long-term inertia in turbine state features was addressed by decomposing the input time-series data to effectively capture seasonality. The efficacy of IN-DLinear is systematically evaluated using 10-year field measurements from a 2.5-MW wind turbine. IN-DLinear exhibited superior performance, reducing prediction errors by 13%∼30% compared to mean value judgment and other deep time-series models, including Seq2Seq, Transformer, and Autoformer.more » « less
-
Abstract Task‐incremental learning (Task‐IL) aims to enable an intelligent agent to continuously accumulate knowledge from new learning tasks without catastrophically forgetting what it has learned in the past. It has drawn increasing attention in recent years, with many algorithms being proposed to mitigate neural network forgetting. However, none of the existing strategies is able to completely eliminate the issues. Moreover, explaining and fully understanding what knowledge and how it is being forgotten during the incremental learning process still remains under‐explored. In this paper, we propose KnowledgeDrift, a visual analytics framework, to interpret the network forgetting with three objectives: (1) to identify when the network fails to memorize the past knowledge, (2) to visualize what information has been forgotten, and (3) to diagnose how knowledge attained in the new model interferes with the one learned in the past. Our analytical framework first identifies the occurrence of forgetting by tracking the task performance under the incremental learning process and then provides in‐depth inspections of drifted information via various levels of data granularity. KnowledgeDrift allows analysts and model developers to enhance their understanding of network forgetting and compare the performance of different incremental learning algorithms. Three case studies are conducted in the paper to further provide insights and guidance for users to effectively diagnose catastrophic forgetting over time.more » « less
-
Abstract Event logs, comprising data on the occurrence of different types of events and associated times, are commonly collected during the operation of modern industrial machines and systems. It is widely believed that the rich information embedded in event logs can be used to predict the occurrence of critical events. In this paper, we propose a recurrent neural network model using time‐to‐event data from event logs not only to predict the time of the occurrence of a target event of interest, but also to interpret, from the trained model, significant events leading to the target event. To improve the performance of our model, sampling techniques and methods dealing with the censored data are utilized. The proposed model is tested on both simulated data and real‐world datasets. Through these comparison studies, we show that the deep learning approach can often achieve better prediction performance than the traditional statistical model, such as, the Cox proportional hazard model. The real‐world case study also shows that the model interpretation algorithm proposed in this work can reveal the underlying physical relationship among events.more » « less
-
Satellite sensors have been widely used for precipitation retrieval, and a number of precipitation retrieval algorithms have been developed using observations from various satellite sensors. The current operational rainfall rate quantitative precipitation estimate (RRQPE) product from the geostationary operational environmental satellite (GOES) offers full disk rainfall rate estimates based on the observations from the advanced baseline imager (ABI) aboard the GOES-R series. However, accurate precipitation retrieval using satellite sensors is still challenging due to the limitations on spatio-temporal sampling of the satellite sensors and/or the uncertainty associated with the applied parametric retrieval algorithms. In this article, we propose a deep learning framework for precipitation retrieval using the combined observations from the ABI and geostationary lightning mapper (GLM) on the GOES-R series to improve the current operational RRQPE product. Particularly, the proposed deep learning framework is composed of two deep convolutional neural networks (CNNs) that are designed for precipitation detection and quantification. The cloud-top brightness temperature from multiple ABI channels and the lightning flash rate from the GLM measurement are used as inputs to the deep learning framework. To train the designed CNNs, the precipitation product multiradar multi-sensor (MRMS) system from the National Oceanic and Atmospheric Administration (NOAA) is used as target labels to optimize the network parameters. The experimental results show that the precipitation retrieval performance of the proposed framework is superior to the currently operational GOES RRQPE product in the selected study domain, and the performance is dramatically enhanced after incorporating the lightning data into the deep learning model. Using the independent MRMS product as a reference, the deep learning model can reduce the retrieval uncertainty in the operational RRQPE product by at least 31% in terms of the mean squared error and normalized mean absolute error, and the improvement is more significant in moderate to heavy rain regions. Therefore, the proposed deep learning framework can potentially serve as an alternative approach for GOES precipitation retrievals.more » « less
