Performance Modeling for Space-based Observations of Forest Fires using Microbolometers

Transcription

1 Performance Modeling for Space-based Observations of Forest Fires using Microbolometers Peyman Rahnama 1, Linda Marchese 2, François Chateauneuf 3, John Hackett 4, Tim Lynham 5 and Martin Wooster 6 1 Peyman Rahnama, COM DEV Ltd., 155 Sheldon Drive, Cambridge, Ontario, Canada N1R 7H6. Tel: x peyman.rahnama@comdev.ca 2 Linda Marchese, INO (National Optics Institute), 2740, rue Einstein, Quebec City, Quebec, Canada G1P 4S4. Tel: Linda.Marchese@ino.ca 3 François Chateauneuf, INO (National Optics Institute), 2740, rue Einstein, Quebec City, Quebec, Canada G1P 4S4, Tel: Francois.Chateauneuf@ino.ca 4 John Hackett, COM DEV Ltd., 155 Sheldon Drive, Cambridge, Ontario, Canada N1R 7H6. Tel: x john.hackett@comdev.ca 5 Tim Lynham, Natural Resources Canada, 1219 Queen St. East, Sault Ste. Marie, Ontario, Canada P6A 2E5. Tel: tlynham@nrcan.gc.ca 6 Martin Wooster, Department of Geography, King's College London, Strand, London, UK WC2R 2LS. Tel: martin.wooster@kcl.ac.uk Introduction Space-based observations of hot-spot events have numerous direct benefits to life on Earth. Such observations would enhance the health and safety of human beings and would protect quality of the natural environment. Hot-spot data can be used for fire detection and fire monitoring, volcanic monitoring, land cover change monitoring as well as studies related to biomass burning, carbon emissions and climate change. COM DEV Ltd., the largest Canadian designer and manufacturer of space hardware subsystems and scientific instruments, has recently completed a technology development study under a contract by the Canadian Space Agency (CSA). The study was related to the observation of wildfires from satellite platforms. The main objective of the study was to develop software models to allow the simulation and retrieval of satellite imagery for a thermal imaging system to be used to detect and monitor hot events on the Earth with an emphasis on forest fire observation applications. As part of this study, COM DEV has developed a conceptual instrument design, an end-to-end simulation model, fire data retrieval models and analysis functions for sensitivity analyses, error analyses, performance modeling, trade-off analyses and design optimization. Satellite imagery of fire observations has been simulated to help potential users of forest fire satellite data to better understand the utility of the data and the appropriate level of specification to provide the essential data product at the most effective cost. This paper presents a brief summary of the study. Some future plans and the usefulness of the instrument concept and the simulation models for future fire missions are discussed. IUFRO Division 4 Meeting (2009): Extending Forest Inventory and Monitoring over Space and Time Page 1

2 Rationale for Space-based Observations of Wildfires Degradation of the environment, air quality issues and loss of human lives are pressing issues in the 21 st century. Wildfires, whether caused naturally or by humans, are a dominant disturbance in all vegetation zones throughout the world. Furthermore, wildfires threaten human safety and property. Wildfire smoke can travel thousands of kilometers from its source. Smoke exposure is a health hazard for both fire fighters and people in exposed communities. This gaseous material threatens the lives of humans, animals, and vegetation. Wildfires are a major concern for air quality and climate change issues. Wildfires release large amounts of greenhouse gases and atmospheric pollutants such as carbon monoxide, smoke aerosols and atmospheric mercury. It is believed that greenhouse gases and atmospheric pollutants are the major causes of global warming. Because wildfire emissions are a significant source of carbon gases, they are of interest to many governments for air quality related issues, such as the international carbon gas tax budgets and the Kyoto Protocol. Wildfires not only harm valuable ecosystems and are a danger to human lives, but they also have a huge economic impact. According to the European Commission, each hectare of forest lost to fire costs Europe's economy over $3,000. According to the Canadian Wildland Fire Strategy, on average, ten thousand fires burn more than two million hectares of Canadian forests annually and the cost of fire suppression is more than $600 million per year. For effective control and management of wildfires, early fire detection and monitoring is essential. Space based platforms provide an ideal means for global, frequent, accurate, and yet cost-effective fire detection and monitoring. In addition to fire detection and fire monitoring, data from hot-spot instruments can be used for volcanic monitoring, land cover change monitoring, cloud imaging, aerosol measurements, as well as studies related to biomass burning, carbon emissions, air quality and climate change. Hot-spot data can also be used for forestry applications such as forest inventory, mapping forest carbon and biomass, forest change detection and greenhouse gas reporting. Instrument Concept Channel selection analyses using radiative transfer modeling are critical for properly designing multi-channel imagers. Based on the channel selection analyses performed as part of this study, in order to observe wildfires from satellite platforms, at minimum, three channels are required. The instrument is composed of three compact cameras: Mid-Wave IR (MIR) Camera (~ 3.5 µm) for the detection of wildfires, Thermal IR (TIR) Camera (~ 9 µm) for estimation of background surface temperature, Visible or Near IR Camera for cloud detection and rejection of sun-glint off the water, and sun reflection from clouds. Both photon-counting detectors and uncooled microbolometers were considered in this study. End-to-end Simulation Model The End-to-end Simulation Model consists of a forward model, inversion models, a noise visualization model, a Graphical User Interface (GUI) and analysis functions. The forward model simulates expected observations of space-based thermal imaging fire missions and the noise IUFRO Division 4 Meeting (2009): Extending Forest Inventory and Monitoring over Space and Time Page 2

3 levels for different instrument characteristics, atmospheric conditions and measurement scenarios. The inversion models use the mock raw data simulated by the forward model to retrieve fire characteristics such as fire temperature and Fire Radiative Power (FRP). The forward model is composed of a Scene Model, an Atmospheric Model, an Instrument Model, a Viewing Geometry Model and an Observational Simulation Model. The model also includes radiative transfer models for various applications such as channel selection studies. The expected wildfire measurements for combinations of different scene characteristics (e.g., different cloud type, different atmospheric conditions, different surface temperature), different instrument and orbit characteristics and different measurement scenarios (e.g., different swath width, different ground resolution) have been simulated. Level 0 to level 2 fire data have been simulated. The simulation of the data chain will be invaluable for performance modeling and scientific assessment of the instrument performance. Using MATLAB, a Graphical User Interface (GUI) for the end-to-end simulation model has been developed for interactive user interfaces. The GUI allows the user to change some of the model parameters such as ground sampling distance and see the impact on signal and noise images to obtain optimum fire observation conditions. A sample GUI input window and sample GUI output images are presented in figures 1 and 2 respectively. The GUI outputs temperature map(s), signal at the detector with no noise, signal with noise added after is readout by the electronics, and the corresponding Signal-to-Noise Ratio (SNR) or Noise Equivalent Power (NEP) on a pixel basis. The GUI uses adjustable parameters. Figure 1: A sample GUI input window. IUFRO Division 4 Meeting (2009): Extending Forest Inventory and Monitoring over Space and Time Page 3

4 Figure 2. Sample GUI output images. The signal reaching the detector has been calculated for each pixel. The noise level is calculated for each pixel signal after it is read out by the electronics. As a mean of visualizing the noise influence on the image, signal images are generated many times with a noise statistic model added to the signal for each pixel and the resulting images are overlaid on specific geographic locations on Earth. Animated sequences have been used for this illustration of noise. The visualization model can help users evaluate satellite imagery products for various instrument and measurement scenarios. An inversion model based on an improved version of Dozier method (Dozier 1981), which is similar to the approach suggested by Giglio and Kendall (2001) has been developed. The main improvement is accounting for atmospheric transmittance of the radiant emission from the fire. A Fire Radiative Power (FRP) inversion model based on MIR radiance method (Wooster 2003) has also been developed. FRP is important for studies related to rate of biomass combustion and rate of production of carbon emission from biomass burning. Using the inverse models, fire characteristics such as fire temperature, fire fractional area and Fire Radiative Power (FRP) have been retrieved from the mock raw data simulated by the forward model. Analyses Having adjustable parameters for the forward and inverse models allows instrument and orbit parameter optimization. The end-to-end simulation model along with the analysis functions have been employed to perform error analyses, sensitivity analyses, performance modeling, trade-off studies and design optimization. The simulation and retrieval models have been used to assess the concept feasibility of hot spot measurements using Canadian microbolometer technology. A IUFRO Division 4 Meeting (2009): Extending Forest Inventory and Monitoring over Space and Time Page 4

5 preliminary performance assessment of space-based observations of wildfires using the instrument employing microbolometer technology has been undertaken by simulating the expected performance of the instrument and performing error analyses accordingly. The suitability of microbolometer detector technology for wildfire observations and the performance limitations have been determined. Summary There is a need for fire detection and monitoring in Canada. A fire-observing mission would advance our knowledge and understanding of our atmosphere, air quality issues, carbon emissions, forestry, climate change and global warming, and would provide a scientific foundation for the sound policy needed to protect the future health of our planet. Such a mission would provide the data necessary for Canada to respond effectively to local and global fire issues as well as some of the atmospheric and environmental issues. The simulation and retrieval models can be used as a powerful tool for defining the requirements and evaluating the performance of fire imaging systems. The simulation and retrieval models developed by COM DEV as part of the study are critical to optimize the design of fire instruments and to refine the fire retrieval algorithms. The conceptual design of the instrument as well as the simulation, retrieval, and radiative transfer models will be very useful for future space missions that are intended to generate imagery of the earth for forestry applications. It is planned to consolidate the user requirements. The multi-channel imager will be customdesigned to meet needs of the users in a cost-effective manner. For instance, NIR and Visible channels can be added to the instrument for Directional Spectral Reflectance measurements for fire scar and post-burn studies. The end-to-end simulation models will be employed to derive the instrument and mission design requirements based on the measurement requirements. The performance analysis functions will be used to verify the feasibility of achieving the mission requirements. The instrument design will be optimized using the simulation models and analysis functions. Acknowledgements The authors acknowledge the support of the Canadian Space Agency (CSA). Literature Cited Dozier, J A Method for Satellite Identification of Surface Temperature Fields of Subpixel Resolution, Remote Sensing Env., 11, Giglio, L. and J. D. Kendall Application of the Dozier retrieval to wildfire characterization- A sensitivity analysis, Remote Sens. Environ., 77, Wooster, M., B. Zhukov and D. Oertel Fire radiative energy release for quantitative study of biomass burning: derivation from the BIRD experimental satellite and comparison to MODIS fire products, Remote Sens. Environ., 86, IUFRO Division 4 Meeting (2009): Extending Forest Inventory and Monitoring over Space and Time Page 5