Progress 07/01/23 to 02/29/24

Outputs
Target Audience:During Phase I, our research results have been discussed in detail with experts via outreach, planning collaboration for Phase II, and / or our participation in the DoD Wildland Fire Science Initiative (WFSI) prescribed (Rx) burn at Fort Stewart in February, 2024. These conversations included individuals affiliated with the Missoula Fire Lab, specifically Natalie Wagenbrenner, Jason Forthofer, Bryce Nordgren, Mark Finney, and Dan Jimenez. We also spoke with Joe O'Brien at the Athens Prescribed Fire Technology Hub. At the WFSI burn, we discussed these capabilities with multiple members of the Integrated Research Management Team (IRMT), including James Furman, Linda Chappel, Rick Anderson, and Sue Wilder. In addition, we spoke with several Incident Meteorologists (IMETs), including Larry Van Bussum, Scott Stearns, and Julie Malingowski. We spoke with Adam Watts, head of the FASMEE burn effort, as well. Lastly, we discussed these efforts with our planned Phase II collaborators at Michigan Tech Research Institute (MTRI) William Buller, Brian Wilson, and Joseph Paki. Looking forward, for our FireFGH Repeater Coverage Tool, our future outreach will focus on certified Communications Technicians (COMTs) and Communications Unit Leaders (COMLs), of which there are roughly 100-200 people (per personal communication with Bryce Nordgren, qualified COMT). For AIRWISE, our high-resolution wind uncertainty tool, our outreach will focus on individuals specifically focused on predicting wind or fire behavior. Wind and weather predictions are primarily handled by IMETs, while fire behavior predictions and associated decision making are handled by FBANs as well as Long Term Analysts (LTANs) and Strategic Operations Planner (SOPLs). There are roughly 100 IMETs in the U.S., and a combined total of roughly 400-500 FBANs, LTANs, and SOPLs (per personal communication with Jason Forthofer, whose research focuses include fire behavior). Furthermore, in total, there are roughly 5000 WindNinja users (per personal communication with Natalie Wagenbrenner, one of the creators and maintainers of WindNinja), an overlapping market to IMETs and FBANs, but nonetheless presenting thousands of additional potential customers. For both of these tools, although again our primary focus will be the above-stated groups, during Phase II, we will also begin outreach into several additional markets, including other forms of disaster relief, fire insurance for structures in the Wildland Urban Interface, and UAV use cases like defense, land surveys, and Search and Rescue. Changes/Problems:First and most notably, we originally proposed to combine the use of Kestrel Weather Meters with a TAK plugin to provide situational awareness via real-time data products (Tasks 1, 2, 3.3, 4.4, and 6.2) . To summarize at the highest level, we determined we simply were not the correct people to implement this. The clearest commentary in support of our pivot came from an email conversation with Bryce Nordgren (qualified ITSS and COMT) from the Missoula Fire Lab: [T]rying to construct a geographically distributed, reliable, high bandwidth data network in the woods on the fly is not something the Incident currently does, and it is sufficiently different that an entirely new position would likely need to be created in order to manage this infrastructure [emphasis added]. IT support for the camp is already understaffed. This could open a whole New World of hurt on an incident. - Bryce Nordgren, qualified ITSS and COMT, Missoula Fire Lab With the regulatory bar being that high and our lack of connections to push our desired TAK ideas, we believe progress and advocacy will be far more productive coming from groups like the Dingell Act Resource Tracking (DART) Group, who are already doing rigorous, TAK-based exercises for example at the Caldor Fire (https://storymaps.arcgis.com/stories/3ff970aa446d4e778a7b0ade3f2219ab) and the Tamarack Fire (https://storymaps.arcgis.com/stories/6933d86567ac4f93887bdd5468a85007). Furthermore, we believe Kestrel data integration into TAK could be accomplished far better by having Kestrel do the integration directly in their LiNK app (https://kestrelinstruments.com/link-connectivity). The second pivot was away from commercializing a fast-running, cloud-based WindNinja service (Task 4). Although we deployed a first-pass RESTful WindNinja service to Google Cloud (Tasks 4.1 and 4.3), we learned from conversations with Natalie Wagenbrenner and Jason Forthofer from the Missoula Fire Lab, the creators and maintainers of WindNinja, that they had already contemplated and decided against building / running such a service due to insufficient demand, particularly when weighed against the effort and maintenance costs required to stand up a scalable cloud architecture to achieve the customer-desired speed-ups. This Phase I development effort, however, was by no means a waste, as this infrastructure will be critical for our plans to train a WindNinja-resolution wind uncertainty model (Task 2.2) during Phase II, as we discuss in our Phase II proposal. We also plan to make improvements to WindNinja to make it possible to run WindNinja with real-time environmental data (Task 4.2) and Numerical Weather Prediction (NWP) models like High Resolution Rapid Refresh. Third, we decided against creating formal intervention recommendations (Task 5). Aside from a drastic down-scoping of possibilities here when we removed TAK-based real-time data from our plans, we've found through our Phase I customer discovery that this is an area best left to tunable notifications set by users. This solution is already being best pursued by the Fire Weather Alert System (FWAS), created and maintained by Natalie Wagenbrenner and Jason Forthofer from the Missoula Fire Lab. Rather than reinvent the wheel here, we instead are planning in Phase II to add a static repeater coverage product to FWAS (Task 3.3) to allow users to create their own alerts about repeater coverage via the FWAS framework, discussed further in our Phase II proposal. Fourth, in our Phase I work plan, we talked about Rx burns as merely a more controlled proving ground before bringing our products to wildfire scenarios. However, through our involvement in the WFSI Rx burn at Fort Stewart in Georgia in February, 2024, as well as having had multiple conversations with researchers from the Southern Research Station, we have now pivoted to targeting explicit support for situational awareness in and around Rx burns, a market that we anticipate will grow dramatically in the coming years as the United States deals with an enormous build-up of dead fuels after decades of suppression-first efforts. We will actively pursue this effort during Phase II via collaboration with Joe O'Brien (Task 6.3). What opportunities for training and professional development has the project provided? Nothing Reported How have the results been disseminated to communities of interest?During Phase I, our focus was on customer discovery, which led us to conversations most notably with the Missoula Fire Lab, the Athens Prescribed Fire Technology Hub, and individuals involved in the DoD Wildland Fire Science Initiative (WFSI) program. These connections were a function of previous networking done by GDA employees as well as introductions made on our behalf by members of the WFSI program during Phase I. More specifically, the focus of these conversations was to align our plans with market needs. We believe we achieved this alignment, with the necessary next steps written into our Phase II work plan. During Phase II, we plan to quickly create meaningful artifacts for the firefighting community, after which we will commence broader outreach with the USDA Research Stations, starting with our RMRS and SRS connections, and following up with other Research Stations as well as CalFIRE via our WFSI connections. Probably our greatest asset in these conversations was the kindness and generosity we repeatedly observed in the wildland fire management community. Our efforts to connect with experts have consistently been met with appreciation and assistance rather than defensiveness and isolation, so we want to commend the openness of RMRS, SRS, and WFSI in making the success of our Phase I effort possible. What do you plan to do during the next reporting period to accomplish the goals? Nothing Reported

Impacts
What was accomplished under these goals? Our Phase I goal was to provide proof-of-concept technical capabilities with respect to situational awareness, modeling, and decision making. In Phase I, we deployed a publicly-available MVP of our radio repeater coverage tool. The MVP allows users to create maps that show firefighters in the field where they can anticipate having, or more importantly, losing radio communication. This tool will be the basis of improved understanding of communications constraints in the field, improving firefighter safety. Second, we built our initial Advanced Integrated Resolution for WInd uncertainty Simulation and Evaluation (AIRWISE) Model. During Phase I customer discovery with several Incident Meteorologists and the maintainers of the high-resolution wind modeling tool WindNinja, we found a market need for a high-resolution wind uncertainty model. Building off of our flexible model-training pipeline from Phase I, we can easily retrain our model on WindNinja outputs to create a high-resolution model in Phase II. Our wind uncertainty tool will allow weather experts to better-anticipate unexpected changes, improving situational awareness and safety for firefighting crews around wildland fires. Tasks 1.1 Complete automated data ingest pipeline able to process real-time Kestrel Weather Meter data. And Task 2.1 Create scalar field environmental data products, and provide visualizations for current conditions over regions-of-interest. Tasks were not pursued further. See "Describe major changes" section below. Task 2.2 Implement data coverage / uncertainty assessment of environmental data products, and provide visualizations over regions-of-interest. We are no longer pursuing real-time data collection in and around the fire through TAK, for reasons discussed in detail in the "Describe major changes" section below, but our baseline AIRWISE model generates wind uncertainty data products. During Phase I, we trained our baseline AIRWISE model using gigabytes of weather station data along with years of collected High Resolution Rapid Refresh (HRRR) data, a 3km-resolution Numerical Weather Prediction (NWP) model. During customer discovery throughout Phase I, most notably with several Incident Meteorologists as well as the creators and maintainers of WindNinja, from the Missoula Fire Lab, we found a market need for a high-resolution uncertainty model. Using our containerized WindNinja service (Task 4), we can easily retrain our Phase I model on WindNinja outputs to create a high-resolution uncertainty model in Phase II. Task 2.3 Containerize services to interface with different data inputs and deploy on any system at arbitrary scale. Per our pivot away from TAK, discussed in the "Describe major changes" section below, our only result in Task 2 was AIRWISE. Per our AIRWISE customer discovery discussions, we chose not to containerize and deploy our HRRR-resolution model, focusing instead on further outreach for further customer-specific needs for the planned WindNinja-resolution model. Task 2.4 Provide Incident Command access to real-time environmental and uncertainty data products via a RESTful endpoint. See discussion for Task 2.3. Task 2.5 Integrate situational awareness summaries from the RESTful endpoint into ATAK platform (or other high-priority end-user platform). This task was not pursued further, for reasons discussed in detail in the "Describe major changes" section below. Task 3.1 Build software that can create radio coverage probability fields. This task was completed and deployed as an interactive web page that generates coverage maps based on user-specified locations. Customer discovery, particularly within the USDA (Missoula Fire Lab and Athens Prescribed Fire Technology Hub) as well as at the WFSI prescribed burn demonstrated a clear need for both static and user-specific data products, which we will create during Phase II. Task 3.2 Expose building radio coverage probability field as a service via a RESTful endpoint. This task was completed and deployed via Google Cloud through Continuous Integration / Continuous Deployment (CI/CD) pipelines. Users can request repeater coverage outputs via URL request. Task 3.3 Integrate radio coverage probability field from the RESTful endpoint into ATAK platform (or other high-priority end-user platform). During Phase I, we integrated into our publicly-available MVP. Customer discovery during Phase I presented two next steps for Phase II. First, we will add shapefile exports to pass along to GIS teams and GeoPDF support for Avenza integration for firefighters. Second, we will integrate static data into the Fire Lab's Fire Weather Alert System (FWAS), discussed further in our Phase II proposal. Task 4.1 Expose WindNinja as a containerized service. Containerization was accomplished early during Phase I. Although we deployed a first-pass RESTful WindNinja service to Google Cloud (Task 4.3), we learned from conversations with the creators and maintainers of WindNinja, that they had already contemplated and decided against building / running such a service due to insufficient demand, particularly when weighed against the effort and maintenance costs required to stand up a scalable cloud architecture to achieve the customer-desired speed-ups. Thus, after deploying this first-pass service, we discontinued pursuing any further speed-ups. Task 4.2 Run the WindNinja service with real-time environmental data. This task was originally intended to focus on TAK-accumulated data, which was not pursued further, for reasons discussed in detail in the "Describe major changes" section below. Conversations between GDA and the WindNinja maintainers revealed that currently, WindNinja does not allow users to factor in both NWPs and environmental data into wind modeling. GDA worked with the Fire Lab to propose this as a task in the proposed Phase II work plan (see Letter of Support attached to Phase II proposal). A CRADA agreement with the Fire Lab is also in the works. Task 4.3 Provide Incident Command access to real-time WindNinja outputs as a service via a RESTful endpoint. This task was completed and deployed via Google Cloud through Continuous Integration / Continuous Deployment (CI/CD) pipelines. Users can request WindNinja outputs via URL request. Improvement for this task was discontinued after conversations with the Missoula Fire Lab, discussed in Task 4.1 above. Task 4.4 Integrate WindNinja outputs from the RESTful endpoint into the ATAK platform (or other high-priority end-user platform). Task was not pursued further as a result of conversations with the Missoula Fire Lab, discussed in Task 4.1. Task 5.1 Create software framework and proof-of-concept implementation for interpretable intervention recommendations. Task was not pursued further, for reasons discussed in detail in the "Describe major changes" section below. Task 5.2 Create user stories and visualizations for modeled intervention impact. Task was not pursued further, for reasons discussed in detail in the "Describe major changes" section below. Task 6.1 Demonstrate the data products over past fires, such as the Dixie Fire. Examples of our existing repeater coverage and wind uncertainty for the Dixie Fire are included in our Phase I Final Progress Report. Details can be found in those figure descriptions. Task 6.2 Simulate the data ingest and data product generation using Kestrel Weather Meters. Task was not pursued further, for reasons discussed in detail in the "Describe major changes" section below. Task 6.3 Secure a partner to perform experiments during a controlled burn. We will collaborate with Joe O'Brien (Athens Prescribed Fire Technology Hub) during Phase II (see Letter of Support inPhase II proposal). We will also work with Michigan Tech Research Institute (MTRI, proposed Phase II subcontractor) to design and run our experiments. MTRI has extensive experience with prescribed burn experimentation, including through the WFSI and FASMEE burns

Publications