The Regional Mainland Southeast Asian FIRE Product, 1992-1993, from JRC
Entry ID:
SAI_JRC_Asian_FIRE_Product
|
[
Update this Record
]
|
Updating this record requires registration.
|
Summary
Abstract:
Comparatively little is known empirically about the vegetation fire regime of Southeast Asia when viewed at larger scales. This is despite the importance of fire as an agent of regional land cover change and in modifying atmospheric chemistry. Fire is widely used in rice cultivation in Asia where 94% of the world's crop is grown [Nguyen et al., 1994]. It also has a ... high incidence within forests in tropical Asia [Hao and Liu, 1994] where it is mainly associated with shifting cultivation [McNeely et al., 1991]. As with the neo-tropics and the African tropics, Southeast Asian tropical forests are of considerable ecological and economic importance and makeup about 20% of the world's tropical forest resource [after FAO, 1993]. Information on biomass burning within the Indo-Malayan region is needed to assist in the modelling of large-scale atmospheric pollution and climate change phenomena and for regional use by landuse managers, habitat conservationists, and national and regional policy makers. Mainland Southeast Asia is the focus of the Southeast Asian fire product since it is more strongly seasonal and less humid than many parts of insular Southeast Asia [Nix, 1983] and thus both favours the use of fire as a land management tool and supports more fire-prone ecosystems [54% of forest formations are tropical seasonal forest compared to 4% within insular regions, FAO, 1993]. The mainland Southeast Asian product offers an analysis of the spatial and temporal distribution of vegetation fire in mainland Southeast Asia using AVHRR 1 km resolution data for the period of a single dry season [that chosen is from November 1992 to April 1993]. - Pre-processing: Images selected for pre-processing had [i] low cloud cover [cloud cover over land generally < 30%], [ii] a central swath position to avoid excessive distortion of pixel geometry which is pronounced towards swath extremities, and [iii] low digital corruption [dropped lines and partial line reading]. Thirty images from the available 55 fulfilled these criteria. The pre-processing rational and methodology used are given in Achard and D'Souza [1994], with the exception that that [i] channel-3 radiance was additionally converted to brightness temperature and [ii] a sixth data band is provided which is the difference between channel-3 and channel-4. Accordingly, pre-processing calibration gave channel output as summarised in Table 5.1 below. First-order geometric correction was made using calculated 'tie-points'; data was transformed to the Plate Carree [Lat/Lon] map projection and an area of interest window of 5deg 30' - 25deg 0' N, 97deg 30' - 110deg 0' E was extracted. Table: Band characteristics of AVHRR data used to derive point-fire maps Image band Band description: Radiometric units and scaling: number: Bands 1,2 AVHRR channel-1, 2 Top of atmosphere reflectances x 10 Bands 3,4,5 AVHRR channel-3, 4, 5 [Brightness temperatures - 223] x 10 Band 6 AVHRR [channel-3, 4] Brightness temperature x 10 - Processing: There are three stages in data processing, these are outlined below. [i] Geometric correction: Second-order geometric correction to within c. 1 pixel accuracy was made using user-defined ground control points. [ii] Cloud masking: Cloud cover was determined after visual inspection of the images showed that land-cloud and cloud-smoke differentiation was clearest using the algorithm {[channel-1 + channel-2] > 25% top of atmosphere reflectance and channel-5 < 288 K}. [iii] Fire algorithm derivation: Active fire detection was achieved by classi- fying images using a multi-spectral thresholding technique that used channel-3 and channel-4 brightness temperature. The thresholds of the algorithm were defined using a training-set of pixels that showed [i] saturation, or near saturation, in channel-3, but which [ii] were not saturated on consecutive days. The second criterion was to reduce false-fire detection by minimising the possibility of including semi-permanent hot pixels that saturate due to sunglint or solar heating of the ground. In addition [and especially when next-day consecutive images were not available or when cloud cover was obstructive], other features were used to identify fire and build the training-set, i.e. the spatial distribution of candidate fire pixels, their configuration, background-pixel spectral characteristics, and the presence of smoke. A new training-set was created for each image to account for temporal changes in land surface characteristic and changes in atmospheric conditions and viewing angle. The use of multi-spectral algorithms is discussed by Stuttard et al. [1995] and Justice and Dowty [1994], here the algorithm derived for each image was of the form: Step 1: Bt3 > k1 Step 2: Bt3 - Bt4 > k2 Step 3: Bt4 > k3 Where Bti is brightness temperature in channel i, k1 is fixed to 320 K [slightly below the saturation temperature of channel-3], and k2 and k3 are determined by the lower 95% confidence interval [ - 2s] of the distributions of the fire training-set radiance values for each image. If the criteria of all three steps are met then the pixel is classified as a fire. NB this image-specific approach could not be used to determine an image-dependent algorithm threshold for channel-3 since this channel's response is influenced by the saturation phenomenon of the sensor. There are two main categories of products and two subsidiary categories of products. The main products are point-fire maps [daily and daily-composite] and fire-density maps [composite]. These products are described in detail below. The subsidiary products are cloud cover [daily and daily-composite] and the source images [AVHRR pre-processed data] from which the fire and cloud maps are derived. Point-fire maps: Files are produced from sea and cloud masked AVHRR 1 km resolution images using image-specific multi-threshold fire detection algorithms. Temporal coverage: - Daily: 30 images, November 1992 to April 1993 [image dates are 10-11/11, 10-14/12, 19-21/12, 27-28/12, 30/12/1992 and 30-31/1, 1-2/2, 6-9/2, 3-6/3, 4-6/4, 16/4, 23/4/1993]. The point-fire map for 4 April 1993 is illustrated in Figure 9a. - Two-monthly: Nov/Dec 1992 [early dry season], Jan/Feb 1993 [mid dry season], Mar/Apr [late dry season]. - Six-monthly: November 1992 to April 1993 [30-image composite over the 1992/93 dry season]. Fire-density maps: Files are produced from the 30-image composite of point-fires described in Section 5.3.1 above; fire density within 25 x 25 km grid cells is plotted on a relative scale: low, medium, high and very high fire activity [categories are defined]. Temporal coverage: - A single map covers the entire 1992/93 dry season [November to April]. Users will find the active fire data product most useful if it is integrated with GIS support. A summary of the scales, land cover/land use, infrastructure, and applications that the data may be used at, integrated with, and applied to is given below. An exhaustive list of potential uses is not given here since the user-community is well aware of its own needs. * Scale: Local [e.g. national park studies], country [national policies], regional [Southeast Asian perspectives]. * Land use: Permanent agriculture, shifting cultivation, timber, plantation, grazing. * Land cover/terrain: Forest, savanna, savanna woodland, elevation, * national parks. * Infrastructure/communications: Roads, rivers, habitation. * Applications: Atmospheric chemistry, land cover change, deforestation, fire ecology, population and migration pressure.
Related URL
|
Description:
USGS Coastal Marine Time Series Browser
|
Multimedia Sample
 View full image
Caption:
Map showing mooring locations for the Middle Atlantic Bight
|
|
Geographic Coverage
(Click for Interactive Map)
Spatial coordinates
| |
N: 41.47
|
|
S: 38.8
|
|
E: -73.78
|
|
W: -75.69
|
|
Data Resolution
Temporal Resolution:
60.00 minutes
Temporal Resolution Range:
Hourly - < Daily
Personnel
Role:
TECHNICAL CONTACT
Phone:
508-457-2356
Email:
emontgomery at usgs.gov
Contact Address:
U.S. Geological Survey Woods Hole Science Center
384 Woods Hole Rd
City:
Woods Hole
Province or State:
MA
Postal Code:
02543-1598
Country:
USA
Role:
INVESTIGATOR
Phone:
(508) 457-2229
Fax:
508-457-2310
Email:
rsignell at usgs.gov
Contact Address:
Gulf of Maine Information System
Quissett Campus
384 Woods Hole Road
City:
Woods Hole
Province or State:
MA
Postal Code:
02543-1598
Country:
USA
Creation and Review Dates
Last DIF Revision Date:
2011-09-29
|
[
Update this Record
]
|
Updating this record requires registration.
|
|