An Enhanced Contextual Fire Detection Algorithm for MODIS

https://doi.org/10.1016/S0034-4257(03)00184-6Get rights and content

Abstract

Experience with the first 2 years of high quality data from the Moderate Resolution Imaging Spectroradiometer (MODIS) through quality control and validation has suggested several improvements to the original MODIS active fire detection algorithm described by Kaufman, Justice et al. [Journal of Geophysical Research 103 (1998) 32215]. In this paper, we present an improved replacement detection algorithm that offers increased sensitivity to smaller, cooler fires as well as a significantly lower false alarm rate. Performance of both the original and improved algorithm is established using a theoretical simulation and high-resolution Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) scenes. In general, the new algorithm can detect fires roughly half the minimum size that could be detected with the original algorithm while having an overall false alarm rate 10–100 times smaller.

Introduction

As part of NASA's Earth Observing System (EOS), the Moderate Resolution Imaging Spectroradiometer (MODIS) is carried on both the Terra and Aqua satellites. The MODIS instruments, which began collecting data in February 2000 (Terra) and June 2002 (Aqua), are being used to generate oceanic, atmospheric, and terrestrial data products Kaufman et al., 1998, Masuoka et al., 1998. Since launch, emphasis has been given to characterizing instrument performance, determining and monitoring the quality of the data products, and undertaking validation (Morisette, Privette, & Justice, 2002). Based on this understanding, improvements have been made to all of the algorithms. The MODIS active fire products fall within the suite of terrestrial products and provide information about actively burning fires, including their location and timing, instantaneous radiative power, and smoldering ratio, presented at a selection of spatial and temporal scales Justice et al., 2002, Kaufman et al., 1998. A detection algorithm that identifies the active fires within each MODIS swath forms the basis of these products.

Although the original MODIS fire detection algorithm of Kaufman, Justice et al. (1998) functioned reasonably well following several initial postlaunch revisions collectively known as “version 3” (Justice, Giglio et al., 2002), two significant problems limited the overall quality of the product. Firstly, persistent false detections occurred in some deserts and sparsely vegetated land surfaces, particularly in northern Ethiopia, the Middle East, and Central India. Not unexpectedly, most of these were caused by the algorithm's absolute threshold tests. Secondly, relatively small (yet generally obvious) fires were frequently not detected. In response to these problems, we have developed a replacement version 4 contextual algorithm that offers superior sensitivity to smaller, cooler fires and have yielded fewer blatant false alarms. In this paper, we describe this algorithm.

Section snippets

Algorithm description

The improved detection algorithm is based on the original MODIS detection algorithm (Kaufman, Justice et al., 1998), heritage algorithms developed for the Advanced Very High Resolution Radiometer (AVHRR) and the Visible and Infrared Scanner (VIRS) Giglio et al., 1999, Giglio et al., 2003, and experience with the first 2 years of high quality MODIS data.

The algorithm uses brightness temperatures derived from the MODIS 4-and 11-μm channels, denoted by T4 and T11, respectively. The MODIS

Algorithm performance

To date, four principal methods have been used to assess algorithm performance and evaluate the MODIS fire products. First, based on earlier work done by Dowty (1993) and Giglio et al. (1999), simulated MODIS imagery was used to quantify algorithm detection and false alarm rates under a wide range of environmental conditions within different biomes. Second, fire maps generated from high-resolution scenes acquired with the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER)

Conclusion

We have described an improved contextual active fire detection algorithm for the MODIS instrument. This algorithm, known as version 4, offers considerable improvement over previous versions. The version 4 algorithm is run as part of the MODIS land forward processing stream, as well as within the MODIS Rapid Response System (Justice, Townshend et al., 2002). It is also being run as part of the MODIS “Collection 4” reprocessing stream to reprocess all MODIS data starting from March 2000, the

Acknowledgements

We thank Jeffrey Morisette (NASA Goddard Space Flight Center) and four anonymous reviewers for their helpful comments on the manuscript.

References (22)

  • L. Giglio et al.

    A multi-year fire data set for the tropics derived from the TRMM VIRS

    International Journal of Remote Sensing

    (2003)
  • Cited by (1414)

    • Division of the tropical savanna fire season into early and late dry season burning using MODIS active fires

      2023, International Journal of Applied Earth Observation and Geoinformation
    View all citing articles on Scopus
    View full text