About hotspots and burnt area mapping

Hotspot Data

Hotspots are produced from thermal (heat) sensors on a number of different satellites. They are usually accurate to within 1km of their actual location.

The great majority of hotspots represent actual fires or heat sources such as smokestacks. Occasionally hotspots may be triggered by very hot rock surfaces, smoke plumes or errors in the automated process that detects the hotspots from a satellite image. In the latter case, the hotspots may appear in a line running diagonally across the map. These are usually removed as soon as they are detected.

Active fires large enough to be detected by the satellites can have their position updated as often as four or five times a day. The NAFI site sources data from five satellites which have orbits that pass over the poles. These satellites only get to see a given
location twice a day: once in the day and once at night. Broadly speaking there are five hotspot updates a day corresponding to
the different satellites passing over twice a day:

  1. Very early morning pass (around 1-3am CST) Aqua and NPP satellites
  2. Early morning pass (around 4-6am CST) NOAA 18 and NOAA 19
  3. Late morning pass (around 10am-12pm CST) Terra satellite
  4. Early afternoon pass (around 1 – 3pm CST) Aqua and NPP satellites
  5. Late evening pass (around 10pm-12am CST) Terra satellite

Note that these times are Central Standard Time (NT time).

Most hotspots are displayed on the NAFI site between 30 minutes and 3 hours after they are detected by the satellites, as they need to be processed at a downlink station and sent across the network to the NAFI server. A smaller number of hotpots are supplied to NAFI many hours after they were detected; these are from earlier satellites passes that have been processed in a different way, (e.g. sourced from NASA in the US), and are displayed to provide a more complete record. Because of this inherent delay the NAFI site is not suitable for tracking fast moving fires.

Due to the way the heat image is converted into hot spot points, one large fire could show up as many hotspots or a number of smaller fires could appear as one hotspot.

Image: MODIS Fire Ground Observation. Source: (NASA N.D.)

Sometimes smaller fires do not produce a hotspot and other times hotspots may appear where there are no fires.

Cloud cover also affects the satellites ability to capture hotspots. So, hotspots are often a useful guide, but do not always tell the whole truth.

Burnt area maps

Burnt area maps use a satellite imagery to identify (classify) only the areas that have been burnt. The example below shows a burnt area (fire scar) in red on a Landsat image and the same burnt area mapped, with burnt shown as black and not burnt shown as white.

The burnt areas on NAFI are mostly mapped from MODIS satellite images of fire with a pixel size of 250m. This contrasts with the approximately 1 km resolution of the hotspot mapping. This 250m resolution is suitable for broadly mapping the areas burnt by large open-country fires, but it won’t pick up the smaller scale patchiness that often characterises early dry season fires.

How frequently are the fire scars mapped, and how accurately are they dated?

In the northern fire season, the fire scar mapping in a given area of the fire-prone far north (north of 20 degrees South) is usually updated once a week – occasionally twice a week.

NAFI mapping is undertaken using a combination of both automated and manual processes. Using MODIS imagery two (cloud-free if possible) image dates are selected, generally about a week apart. A computer program creates polygons/areas grouping the image into similar regions based on changes between the image dates.

The rest of the process is largely manual, with the NAFI mapper visually identifying burnt areas based on shape, pattern and context, with the help of additional data such as hotspots and fire lines. Polygons that have been visually identified as containing burnt areas are selected and copied into a final burnt area map for upload to the NAFI website.

If the interval between the two images dates covers the end of one month and the beginning of the other, then the fire scar is given the colour corresponding to the month with the most days in the interval. So, the cut-off date between one month and the next in the fire scar mapping is only approximate for most months.

July-August cut-off The exception is the cut-off date between July and August, which is important for Carbon Farming Initiative Savanna Burning projects. Images are chosen to ensure that mapping identifies all July fire scars
correctly and does not assign August scars to July.

The advantage of burnt area maps is that you can calculate area burnt and other information. By adding multiple burnt area maps together over a year, you can calculate the total area burnt. By adding multiple yearly burnt area maps together, you can calculate the fire frequency for a particular landscape.

Many of these derived fire history maps, such as the one above show how many times an area has been burnt since a certain year (in this instance year 2000) and are available via the NAFI website.

The trouble with cloud

Cloud cover prevents satellite sensors from capturing information on the ground and is a constant issue for providing comprehensive burnt area and hotspot coverage.

Fortunately, in Northern Australia during the dry season when most fires occur, we often have cloud free days. Issues related to cloud cover are generally more common early and later in the dry season during build-up and build-down times.

Sometimes directly above a fire might appear with clear skies however the direct line of sight between the satellite and the fire at the time of overpass may result in parts of the image or hotspot detection being blocked. If the fire goes out before the next satellite overpass then there will be no hotspots for that fire.

To accurately map the burnt area, mappers will need to wait for the next overpass with cloud free visualisation over the burnt area. If an area of interest on a Landsat image is covered by cloud, the mapper will need to wait another 16 days before they can get another, hopefully cloud-free, image. For Sentinel the revisit period is 5 days.

Sentinel 2 Short-wave Infrared Imagery 5 days apart.

The high frequency (every 1-2 days) of MODIS images and other hotspot sensors however, means that they are good for capturing data during gaps in the cloud. The high spatial extent and temporal resolution of MODIS allows NAFI to provide more frequently updated burnt area mapping – important for monitoring active fire.

A MODIS image comparison.