Since the early days of satellite-based remote sensing, instruments have been used to detect wildfires around the globe. However, what if satellite could monitor, in real-time, where wildfires emerged and provide critical data that can aid firefighters? This is what new work is now focusing on, making satellite-based remote sensing able to determine the intensity and direction of wildfires in real-time and not just monitor their presence.
Recently, NASA scientists have been working on this problem using Moderate Resolution Imaging Spectroradiometer (MODIS) data collected which is then used to monitor the presence of wildfires. Recent work has used neural network models on many years of wildfire data to then automatically determine where a wildfire has emerged to the extent that the method is over 99% accurate. Using older data now improves prediction of future fires. However, the problem is MODIS is not an ideal resolution if one were interested in gaining more data about the emergence of a wildfire, such as its direction of travel and heat intensity. One solution is to have CubeSats, or miniature satellites, flying and monitoring in real-time areas for the emergence of wildfires that can also use thermal sensors which can map intensity of fires or even sensors that obtain data on carbon and soot produced. Such data can then be linked with fire officials who can obtain data quickly as a fire develops, enabling a response to be better shaped towards the type of fire present.
For over a decade, scientists have been trying to apply different instruments for real-time fire monitoring, including measuring intensity of fires using thermal measures and using such instruments as unmanned aerial vehicles (UAVs) that have needed equipment such as infrared cameras, where data can provide live feeds and be used to predict information such as directionality of a fire. More recent approaches in real-time or near real-time fire monitoring has generally focused on balancing spatial coverage and temporal resolution, such as having a 10 minute window between images, and good spatial resolution (~500 m) over a wide area (e.g., all of Japan) in a single image. Such approaches try to balance monitoring large areas, in short time intervals, and enable a good enough resolution such that vital data can be obtained to determine the directionality and other important data on wildfires. This was the case with the Himawari-8 data satellite, launched by Japan, that showed some promise in balancing these characteristics and still show potential in near real-time monitoring of fires.
Other approaches attempt to integrate simulations with satellite monitoring. Scientists see that wildfires can be complex in that they incorporate a variety of factors that include temperature, wind, fuel, and other future weather conditions. Sensor data could be used to directly feed simulations that then attempt to forecast change in a fire. This is what is called a reinitialization approach, which takes fire models and reinitializes them using the latest data to run a forecast simulation. From this, it has been demonstrated the forecasting accuracy of knowing what a wildfire would do significantly improved as near real-time data are incorporated.
What is changing is also how satellites, such as the new generation of GOES satellites, are now being viewed. In recent years, commercial-based satellites were often launched that enabled capabilities such as real-time weather monitoring. Increasingly, governments are realizing that real-time weather monitoring is also critical in integration with such activities as wildfire monitoring and forecasting, including other weather-related or environmental phenomena that can develop quickly such as lighting. In such cases, newer weather satellites incorporate thermal imaging that can be processed using algorithms such as WF-ABBA, which processes thermal images to detect active fires and can be used to determine likely changes within the fire that may lead to a given directional spread.
New satellites and improve image processing methods are better enabling our understanding of wildfires. Still, wildfires, as demonstrated by recent fires in California and elsewhere, are difficult to understand, often affected by multiple factors that are very dynamic and not always easy to monitor given their spatial and resolution scales. Nevertheless, detection and forecasting techniques are improving in acquiring the needed data to better understand wildfires in real-time, where better algorithms and improved imagery are beginning to make a difference.
 For more on monitoring and investigating fires in real-time data from satellites, including machine methods used, see: https://eos.org/articles/new-eyes-on-wildfires.
 For more on UAV instruments used to monitor wildfires, see: Casbeer, D. W., Sai-Ming Li, Beard, R. W., Mehra, R. K., & McLain, T. W. 2005. Forest fire monitoring with multiple small UAVs. Proceedings of the 2005, American Control Conference, 2005.: 3530–3535. Presented at the Proceedings of the 2005, American Control Conference, 2005., https://ieeexplore.ieee.org/document/1470520/, May 8, 2019, Portland, OR, USA: IEEE.
 For more on Himawari-8 and its use in monitoring fires, see: Liu, X., He, B., Quan, X., Yebra, M., Qiu, S., Yin, C., et al. 2018. Near Real-Time Extracting Wildfire Spread Rate from Himawari-8 Satellite Data. Remote Sensing, 10(10): 1654.
 For more on a recent approach in integrating simulation and remote sensing data, see: Cardil, A., Monedero, S., Ramírez, J., & Silva, C. A. 2019. Assessing and reinitializing wildland fire simulations through satellite active fire data. Journal of Environmental Management, 231: 996–1003.
 For more on next generation of weather satellites and potential for real-time fire monitoring, see: Jones, S., Hally, B., Reinke, K., Wickramasinghe, C., Wallace, L., & Engel, C. 2018. Next Generation Fire Detection from Geostationary Satellites. IGARSS 2018 – 2018 IEEE International Geoscience and Remote Sensing Symposium: 5465–5468. Presented at the IGARSS 2018 – 2018 IEEE International Geoscience and Remote Sensing Symposium, https://ieeexplore.ieee.org/document/8518812/, May 8, 2019, Valencia: IEEE.
Radiant Earth has launched Radiant MLHub, a cloud-based open library for training geospatial data used by machine learning algorithms.
Researchers have compiled an enormous global dataset with over four trillion satellite-based measurements of sea surface temperature.