Accurately estimating tree volume and biomass is necessary for forest ecology and fundamental for fire management. However, traditional methods for their estimation are expensive, require a large amount of labor and material resources, and may require destructive sampling. In this study, we used a terrestrial laser scanner (TLS) and free software to estimate the volume and biomass of tree trunks in individual trees of two coniferous species, Abies religiosa (Kunth) Schltdl. & Cham and Pinus pseudostrobus Lindl, in two plots of the Monarch Butterfly Biosphere Reserve, Michoacán, Mexico. Comparisons between measurements with predictive models (destructive method) and TLS (non-destructive method) showed no significant differences. The TLS represents an alternative to the traditional methods of measurement, which allows for the estimation of diameters at different heights through the cloud of points, which opens new opportunities to characterize standing tree volume and biomass and their burning potential, as well as ladder fuels. The results show that this increasingly accessible technology can be used by local rural inhabitants to adequately estimate aerial biomass and fuel loads in a non-destructive manner and with a more efficient use of time.

The NASA Wildland Fire Management Program was initiated within NASA’s Earth Science Division in 2021. Its mission is threefold: 1) assemble communities of practice through collaborative efforts with government, academia, and the private sector; 2) co-develop knowledge and applications with relevant stakeholders in the wildfire community; and 3) improve wildland fire management through the transitioning of NASA technology and applied science to stakeholder organizations.

The program is focusing on supporting proactive fire management, including situational awareness, preparedness, and risk mitigation. This will be accomplished through selected projects that identify management challenges, relevant partnerships and users, and the NASA technology that will be utilized to deliver innovative solutions to managing wildfires. Examples include: 1) investigation of evaporative stress from OpenET to help predict the risk of wildfire occurrence in watersheds, 2) incorporation of space based LiDAR for the generation of 3-dimensional forest fuel metrics, used to improve wildfire risk and behavior models, 3) integration of global, multi-platform geostationary active fire data in near-real-time into NASA’s Fire Information for Resource Management System (FIRMS), and 4) identification of post-fire ecohydrological conditions using thermal, multispectral, synthetic aperture radar (SAR), and hyperspectral remotely-sensed data to improve flood hazard forecast models.

In addition, the program is hosting the Actionable Fire Science Hub, a NASA portal for stakeholders to share fire management information and to enable more effective collaboration. This will include open data and other content, and an environment for development and assessment of solutions for adapting to, managing, and living with wildfires.

Improvements in understanding fire behaviour and associated modelling drove a need to update Australia’s bushfire spread simulation capability to a more flexible, modular and computationally efficient. In 2020, the Spark bushfire spread framework, built by CSIRO, was selected by AFAC to form the foundations of the next generation national operational bushfire spread simulator. Spark Operational has been designed for, and with the input from, Australia’s rural fire and land management agencies.
The underlying design paradigm for Spark (and its subsequent operational variant) has been flexibility of construction, streamlining of common processes, and speed and efficiency of computation and data management. Key elements in Spark include a computationally economical fire perimeter expansion algorithm that allows thousands of fires to propagate simultaneously and efficiently deal with perimeters that merge or overlap; the ability to implement numerous fuel type-specific fire spread models concurrently, and the ability to utilise state-of-the-art computer hardware to accelerate processing.
The Spark framework implements fire science components (e.g., primary fire spread and ancillary models) as scripts that allow new models to be added and evaluated rapidly, unlike older simulators that hard-coded models and associated data. Similarly, the data requirements for Spark’s fire models are also defined such that new or different sources may be accessed or generated as required in the most streamlined manner.
This presentation will outline the intent, structure and operation of Spark Operational that makes it suitable as the next generation bushfire simulator with flexibility to be adopted internationally.

Objectives
In 2019, CAL FIRE started a process to implement technologies to support the wildfire crisis, which continued dramatically during the 2020 and 2021 fire seasons. This process led to the creation of an Intel community that implemented state of art fire risk modeling, monitoring, and tactical tools.
Methods
There was a request for innovative ideas in 2019 to support fire detection, monitoring, and operational wildfire risk modeling, both in advance and for individual incidents. After 131 proposals, Technosylva was selected to implement Wildfire Analyst and Tactical Analyst statewide as the state's authoritative solutions to evaluate operational fire behavior and support wildfire operations.
Results
During the last three years, a community of more than 4000 users has been trained and used these tools during the initial attack and in major incidents, from Dispatch centers, Emergency Command Centers, Incident Command Posts, and in the Field.
More than 30,000 incidents received an assessment, and more than 70000 simulations supported the intel mission.
Real-time tracking of resources, updates of fire progression every 15 min with DoD data, integration of > 1000 cameras, high-resolution weather and fuel models, and stations data, including WRF-SFire and other operational models, together with information from a mobile app is changing the paradigm of the Situation Units with these new capabilities.
Conclusions
Taking these proven tools to real operations has allowed capturing the needed data to improve the scientific models and make the advances needed to support the Intel Community.

The mapping of burned areas has been important for the implementation of public
policies for planning, control and suppression of fires. Deep learning based methods
have been considered state-of-the-art in several environmental applications, although
they have limitations in generalizability when training samples collected from a spatial
region different from the one used in the mapping. Therefore, our objective is to assess
the generalization and mapping capability of burned areas in the Brazilian Amazon
using deep learning and PlanetScope imagery. Respecting to generalization, we used
only images from the Brazilian Pantanal for training, which is the largest wetland in the
world. These images were divided into patches with 512 x 512 pixels and manually
labeled, comprising 4-band PlanetScope multispectral imagery with spatial resolution of
four meters. The dataset consisting of 5222 patches for training and 88 patches for
validation was used to train SegFormer, an efficient semantic segmentation method that
combines transformers and multilayer perceptron decoders. For the test, 448 patches
from the Brazilian Amazon were used to investigate the ability of the proposed
methodology to map burned areas, as well as its generalization. Our methodology
achieved 80.57% of IoU, 90.71% of pixel accuracy and 89.24% of F-Score. The
obtained results indicate that transformer-based networks are suitable to deal with the
mapping of burned areas based on high spatial resolution images, even when trained
and evaluated in different regions. Future studies should explore the potential of vision
transformer architectures to mapping burned areas.

When it comes to wildfire detection, high-altitude surveillance allows for more efficient monitoring of vast expanses, but the comparatively low-resolution is a source of ambiguity (false positives or negatives). Conversely, low-altitude flights will yield the accurate real-time data required for early detection but scaling it up would require an unfeasibly large fleet. The objective is to compare these two approaches in scenarios featuring different models of autonomous unmanned air vehicles (UAV).

We investigate how a network of patrol routes, the topology of which reflects the constraints of the different drone models, would allow for the detection and disambiguation of a fire event at variable stages. We use a simple simulation-based method informed by reasonable assumptions about the relationship between altitude, autonomy, maneuverability, observation range and accuracy, to produce quantitative performance estimates for several alternative types of UAV (e.g., fixed-wing vs. quadcopter).

Our results emphasise the benefits of a hybrid approach combining multiple types of autonomous UAVs operating on different scales. In such a scenario, a fleet of high-altitude, long range, fixed-wing aircraft is used to continuously update a coarse-grained picture of a large area. Upon detecting a suspected fire event, these units may call for a disambiguation mission to be executed by a smaller, low-altitude drone.

We conclude that machine intelligence enablers such as division of labour and flight plan coordination, as well as other important factors such as optimal positioning of automated UAV facilities, are critical to the performance of such a hybrid wildfire detection strategy.

Fire intensity is the most commonly used term describing fire behaviour in the wildfire community. It is, however, difficult to observe from space. Here, we assess fire spread and fire radiative power using infrared sensors with different spatial, spectral and temporal resolutions. The sensors used offer either high spatial resolution for fire detection, but a low temporal resolution, moderate spatial resolution and daily observations (VIIRS), and high temporal resolution with low spatial resolution and fire radiative power retrievals (Meteosat SEVIRI). We extracted fire fronts from Sentinel-2 and use the available fire products for S-NPP VIIRS and Meteosat SEVIRI. Rate of spread was analysed by measuring the displacement of fire fronts between the mid-morning Sentinel-2 overpasses and the early afternoon VIIRS overpasses. We furthermore tested assessing rate of spread from Planet and Sentinel -2 data, which are both available at high spatial resolution and overpasses are only a short time apart. We retrieved FRP from 15-min Meteosat SEVIRI observations and estimated total fire radiative energy release over the observed fire fronts. This was then converted to total fuel consumption per unit area. Using rate of spread and fuel consumption, Byram’s fire intensity could be derived. We tested this approach in a frequently burning West African savanna landscape. Comparison to field experiments and fire behaviour models showed similar results between field observations, model outputs and remote-sensing-derived estimates. Further development may lead to a remote sensing and model-based fire intensity product replacing widespread assumptions on fire intensity based on fire season.