Operational Satellite Monitoring ofFires in Brazil

(IFFN No. 9 – July 1993, p. 8-11)

---

Fire detection and control are complex tasks in Brazil. The country has old and popular established traditions of burning the vegetation whenever possible, weak environmental concerns, and little or nor capability of fire detection in most of its territory (8 million km2). Among its main uses, fire is normally employed to renew pastures all over the country, to clear felled trees and shrubs in areas of new deforestation, and in sugar cane plantations prior to manual harvesting. A pronounced dry season of about four months in the southern and central regions during the austral winter create very favorable conditions for the wide spread use and propagation of fires. Natural fires (e.g. lightning fires) represent a negligible fraction of fire events.

Since 1987 an operational system of fire detection based on orbital remote sensing has been used in Brazil. This report summarizes the technique used and the products available, pointing to its main advantages, limitations and future needs. Readers interested in more details should be directed to the references listed at the end of the text.

The satellites used for fire detection and monitoring are the NOAA-series meteorological satellites of polar orbit (~98° , 840 km altitude). There are always two of these satellites in operation, resulting in at least four overpasses per day for any tropical regions of the globe. At present, four NOAA satellites can be used (NOAA 9-12) and an additional one (NOAA-13) is expected to be in use by July 1993. The early afternoon images (at 2 p.m. local solar time) are particularly used for fire monitoring because most fires lit around noon are still active during the satellite pass. Each image covers a ground strip of about 2,500 km oriented in the SSE-NNW direction for day-time ascending passes, or NNE-SSW in night-time descending passes. The length of the strip is about 4,000 km, centered in the latitude of the receiving station. Such wide-area coverage in a short time interval (14 minutes) available a few times per day is the main advantage of these satellites in fire detection. No cost or restrictions exist to receive these AVHRR images/HRPT mode. Accredited commercial complete receiving stations have been sold for less than US$100,000; ones of 1/10 of this price are currently offered!

The detection of fires relies on the picture elements (pixels) of the image above a certain threshold in the thermal channel 3 (3.9 m m) of the 4-channel AVHRR sensor aboard the NOAA satellites. This channel, with nominal saturation at 43° C, was designed for ocean and temperature measurements. Nevertheless, it is the most sensitive one to targets with temperatures of vegetation fires. Its signal, therefore, can be used only to detect fires but not to estimate their temperature or size. Hot targets like cities of dense construction or exposed soils under a hot sun are seen by channel 3 with much lower temperatures than active fires and are not erroneously mistaken with them. Any fire front with ca. 50 m or more will be detected by channel 3, and will be indicated by at least one pixel. Since the size of a pixel is at least 1.1 x 1.1 km, area estimates of fire fronts for field operations planning is impossible. The precision of fire detection is one pixel, which at nadir is ± 500 m; at the edges of the image, because of pixel geometric distortion, it may reach ± 3,000 m. A major constraint is caused by sun glint in water bodies and in very reflective hot exposed soils in particular cases of sun-target-satellite geometry. In Brazil this problem was greatly reduced by using always the most west pass of the satellites for the region of interest. When this pass occurs the sun is at a lower angle for that region and the soil temperature has decreased in relation to the previous pass which was ca.25 degrees eastward, highly reducing the chances of sun glint.

Two types of products are made and operationally distributed in Brazil by the National Space Institute (Instituto Nacional de Pesquisas Espaciais – INPE), which receives the NOAA images and processes them on real time. The first one is directed to those with operational needs of fire detection and fighting, like organizations in charge of protected areas or commercial forests, and environmental agencies. A special image processing software coupled with a geographical information system identifies all “fire pixels” in the images received, determines their geographical coordinates, selects those in each individual area of interest, and prepares and sends a telex message to specific users. Within about 30 minutes of the satellite pass individual users have at their telex machines a list of geographical coordinates of the fires detected in their own regions. Telex transmissions are preferred to avoid noise problems found in regular fax/phone lines. Until 1992 only the afternoon satellite passes were used on a regular basis, with night and early morning passes processed only on special request; in 1993 an early morning image should also be included in the monitoring.

This fire monitoring system is operational on a daily basis in Brazil for six months, from 1 June until 30 November. The size of the areas monitored varies according to the users needs, ranging from small ones with just a few km2 to those of large states of many thousands of km2. Users usually relay the locations of the fires received by telex to fire brigades in the field by radio. Fire fighting brigades which checked hundreds of the fires detected by this satellite system reported that ca.98% of them were correctly identified; the remaining 2% were never reached because of logistical problems. In relation to other sources of information on fires, like calls from the public, park guards, aircraft pilots, the satellite detection accounted for ca.96% of all fires detected. For statistical and control purposes, the system also generates a monthly statement for each area monitored, indicating the total number of fires detected each day.

A second operational product of the system gives an estimate of fire distribution and density for the country with cumulative weekly and monthly data obtained from the daily image processing. Fire pixels are counted in Grid cells of 0.5° of latitude by 0.5° of longitude and sent every week to users by E-mail. Three matrixes of data are produced and sent: the first contains the number of fire pixels detected in each grid cell during the week; the second contains the number of times each cell was imaged by the satellite; and the third has the average number of fire pixels in each grid cell. The E-mail file containing the three matrix has ca.70 Kbytes. Users of these weekly products include newspapers that publish maps of fire density in the country or in specific states, and different scientific groups interested in studies on vegetation, atmospheric chemistry and climate modelling.

 

Fig.1. Fire activities in Brazil during July 1992 as recorded by NOAA satellite data.
Source: National Institute for Space Research INPE, Sao Paulo, Brazil.
(will be added later)

 

Fig.2. Fire activities in Brazil during August 1992 as recorded by NOAA satellite data.
Source: National Institute for Space Research INPE, Sao Paulo, Brazil.
(will be added later)

 

The two figures show an example of maps made from such matrixes for the months of July and August 1992. The month of August, which includes the peak of the fire activity, had over 100,000 fires, while July, still at the beginning of the fire season, had only 15,000 fires detected by satellite. The highest density of fires is found along the current south limit of the Amazon forest. This is a region subject to intense deforestation through fire. The vegetation is cut at the end of the wet season or start of the dry season, and then left to dry for one or two months. Fire is used to burn the dead organic matter and will be used in the same place for many years until all organic matter from the original forest is consumed.

Although restricted by the Brazilian government, deforestation and the associated burnings are extremely difficult to control in such extensive and inhospitable regions. The control has been made in recent years largely based in the satellite detection technique described above: helicopters stationed in the region have been directed to the geographical coordinates of the fire pixels in AVHRR’s channel 3 images. This enabled the Environmental Institute of Brazil to enforce existing legislation on fires and deforestation with millions of dollars in fines imposed on law offenders. In fact, AVHRR images in 1987 were responsible for showing the scientific community and the public in general that biomass burning in Amazonia was taking place at unprecedented rates and out of control.

Without discussing extensive existing field and validation work or theoretical considerations, the following pros and cons of the AVHRR detection of fires are listed:

Main advantages

• methodology of detection regular and uniform
• use of satellite data not restricted, costs are less than other images
• four images daily anywhere in the globe, at least
• coverage ranging from few to millions of km2
• precise location of fires for fire control purposes
• fast acquisition and distribution of information on fire locations
• simple detection principle, field proven
• products of simple output
• answers individual needs of different users
• fast, simple and diversified delivery of products
• low product cost for users 

Main limitations

• no detection of fires not active during the satellite time overpass
• no detection of fire fronts smaller than ca.50 m
• obscuring clouds (but not smoke) in the fire-satellite line-of-sight
• no detection of fires not reaching the canopy
• solar reflection in a few cases
• very coarse estimate of area in fire
• misses advance of fires between consecutive images

 

 References

Pereira, M.C. and A.W. Setzer 1991. Spectral characteristics of deforestation fires in NOAA/ AVHRR images. Int. J. Remote Sensing 14 (3): 583-597.

Setzer, A.W. and M.C. Pereira 1991a. Amazonia biomass burning in 1987 and an estimate of their tropospheric emissions Ambio 20: 19-22.

Setzer, A.W. and M.C. Pereira 1991b. Operational detection of fires in Brazil with NOAA-AVHRR. 24th ERIM Int.Symp.Remote Sensing of Environm. Rio de Janeiro, Brazil.

Setzer, A.W. et al. 1992. The use of NOAA satellites in the detection of fires in Brazil. Climanalise (INPE/Brazil) 7(8), 41-53. (in Portuguese)

 

 

From:   Alberto Setzer
Address: 
INPE/DSM
Caixa Postal 515
BR – 12201 San José dos Campos

Phone: (++55) 123-418977 ext 347
Fax: (++55) 123-218743
at present
c/o CEC-JRC-ISRA – T.P.440
I-21020 Ispra

Phone:  (++39) 332-785018
Fax:     (++39) 332-789073

---

 Top
Country Notes

 

Fire Activity in the Guyana Shield,
the Orinoco and Amazon Basins During March 1998

(IFFN No. 19 – September 1998,p. 35-39)

---

Introduction

The analysis reported here was done in the framework of two projects of the Global Vegetation Monitoring Unit (GVM, ex Monitoring of Tropical Vegetation [MTV] Unit) of the Space Applications Institute (SAI) of the Joint Research Centre (JRC):

• the FIRE project (Fire in Global Resources and Environmental Monitoring) is dedicated to a permanent monitoring of vegetation fires at regional level, but all over the globe, and
• the TREES project (Tropical Ecosystem Environment observation by Satellite) is dedicated to the development of an operational tropical forest monitoring system.

Both projects have to provide the services of the Commission with up-to-date, independent and policy-relevant information on three domains of strategic importance for the European Union:

• the Framework Convention on Climate Change (Kyoto Protocol on source reduction and sink enhancement of green-house gases),
• the implementation of projects which take into account sustainable development, and
• the protection of forests of global importance.

In this context, a team of the GVM Unit was sent to Paramaribo (Suriname) in South America, from 8 to 21 March 1998, in order:

• to map on real time the vegetation fires over Venezuela, Guyana, Suriname, French Guyana and Northern Brazil,
• to assess the type of vegetation affected by the fires,
• to format the information collected in order to support the evaluation of the role of these fires in the deforestation dynamics and of their impacts on the emission of green-house gases, and
• to provide the scientific team of the LBA Project CLAIRE in Mainz (Germany) with in-situ real time information on the fire distribution.

The focus of the CLAIRE (Cooperative LBA Airborne Regional experiment; LBA = Large Scale Biosphere-Atmosphere Experiment) project is to provide the basis of knowledge required to determine the net exchanges of important gases and aerosols between the atmosphere and the Amazon region, and to understand the processes regulating these exchanges.

Data collection and processing in South America during the campaign of March 1998

From 9 to 20 March 1998, a total of 36 NOAA-AVHRR-HRPT images (1 km resolution) have been acquired and processed on an area of 2000 km x 2000 km covering Colombia, Venezuela, Guyana, Suriname, French Guyana, Northern Brazil and Northeastern Peru.

The acquisition of the NOAA-AVHRR-HRPT images has been done in-situ, three times a day, with a PC-based portable satellite receiver and a horn antenna. The pre-processing, radiometric calibration and geometric correction were carried out using the PANAIS software package developed by GVM (Janodet et al. 1996). The processing for vegetation fire location and GIS analysis were done using ERDAS IMAGINE 8.3 and ArcView packages, on Unix workstation.

Results were made available by e-mail within 24 hours of the satellite overpass, to the CLAIRE/LBA scientific team in Mainz (Germany) and to the services of the Commission, i.e. the DG JRC in Brussels (Belgium) and the direction of the SAI in Ispra (Italy).

General considerations on fire distribution in the region during March 1998

The intense cloud coverage over the region of interest during most of the observation period did not always allow the detection of all individual fires. Therefore, regional sets or groups of fires have been defined and located. Further investigation on individual fires has been conducted locally on specific areas of interest such as conservation areas and “indigenous areas”.

The regional groups of fires have been divided into the following categories:

• a fire set is a group of individual fire events observed in a given area of a limited extent and clearly separated from the surrounding, located by the lat/long coordinates of its centre.
• a fire front is a group of individual fire events distributed within an elongated cluster, positioned by the lat-long coordinates of its extremities.
• a fire zone is a group of individual fire events distributed over large area, still clearly separated from the surrounding, and defined by the lat/long coordinates of its corners.

The various groups of fires have been mapped and analyzed in combination with relevant information such as administrative boundaries, river networks, vegetation cover types (D’Souza et al. 1995) and protected areas.

The satellite observations done from 9 to 20 March 1998 have lead to the definition of four main regions affected by fire activity in South America North of the Equator, called Fire Regions (Fig.1).

These four Fire Regions have been characterized in terms of:

• the spatio-temporal distribution of fire and the type of vegetation affected,
• the probable environmental impacts, relevant for the three European Union’s key issues.

These environmental impacts include:

• the emission of greenhouse gases and aerosols leading to the degradation of air quality with possible health problems, and to the increase of CO2 emissions with consequences on climate change,
• specific disturbances of the ecosystems, the vegetation cover and the soil, of their function and dynamics, leading to the degradation or even destruction of the natural resources like in the case of soil erosion.

Fire Region I: The Northeastern slopes of the Andes (Venezuela and Colombia)

The activity of the fires, which was previously limited in the Llanos, extended on the 12th and the 13th of March over the slopes of the Cordillera de Merida (Venezuela) and the northern part of the Cordillera Oriental (Colombia). These fires were distributed on a SW-NE transect some 200 km long and created smokes covering about 60,000 km2.

This fire event affected a mixture of “sub-montane forest” and “lowland dense forest”. Even if limited in space, it must be considered as having a very high environmental impact with important socio-economic consequences. The sub-montane forest ecosystem is indeed poorly adapted to fire and its disturbance exposes the soil to increased erosion on the slopes.

Fig.1. The four main fire regions in South America (North of the Equator) as derived from satellite observations between 9 and 20 March 1998.

Fig.2. Detaild map of fire patterns in the four fire regions during march 1998 (222 KB)

 

Fire Region II: The Llanos of the Orinoco (Venezuela and Colombia)

Spread over most of the Llanos, the fires in this second region affected more especially two sub-regions: along the Meta river (Colombia) and over the Apure and the Arauca river basins (Venezuela).

At the beginning of the observed period, fires were limited to the central part of the Llanos (Venezuela and Colombia), over most of the Apure and Arauca river basins. A progressive movement of the fire activity was then observed toward East, following the Orinoco river until the end of the reporting period, on 18 March, when the fires were mainly concentrated between Ciudad Bolivar and Ciudad Guyana. The fire activity was intense during the whole period of observation, but with a peak on 13 March.

The main vegetation type affected by the fires in this region was a mosaic of “savannah woodland and xeromorphic forest”, with patches of “lowland dense moist forest”. Here again the forest ecosystem affected by the fires is far from being adapted to the burning practices and the soil degradation and erosion can be enhanced in region such as south of the Orinoco river.

Fire Region III: The coastal zones (Venezuela, Guyana, and Suriname)

The third region corresponds to the coastal zone of the Atlantic Ocean, where the fires were distributed according to three sub-regional patterns:

• on the relief along the coast of the Caribbean Sea (Venezuela): the fires were observed mainly within the “dense moist forest” and some of them in the “agricultural mosaic”;

• within the Orinoco delta: the fires showed a peak of activity on the 13th and 14th of March, affecting essentially the “coastal swamp forest”, and “low land dense moist forest” at the edges of the delta.

• within the 50 km wide belt, parallel to the coast of Suriname and Guyana, liable to flooding from important rivers such as the Essequibo, Berbice or Nickerie: fire activities were particularly dense around Marlborough, Georgetown, New Amsterdam (Guyana), along the Courantyne river and Wageningen (Suriname). A peak of activity was observed from the 15 to 18 March. The vegetation types mostly affected in Guyana were the “lowland dense moist forest” and the “savannah woodland”. But in Suriname, the mixed class “coastal swamp forest and mangroves” was affected.

Due to the very high fragility of most of the coastal ecosystems affected here, one can expect very high environmental impacts of this third fire event. This includes the disturbance of the coastal geomorphologic equilibrium and increased erosion, with consequences on the terrestrial resources but also on the fishery capacities of the coastal zone.

Fire Region IV: The Roraima State (Brazil) and the Kanuku Mountains region (Guyana)

This fourth fire region covers most of the Rio Branco basin, upstream of the confluence with the Catrimani river (Brazil), and the upper basin of the Repununi and the Essequibo rivers (Guyana).

Fires were initially detected within the mosaic of “savannah woodland, xeromorphic forest” and “shrublands”. But after 11 March they clearly extended within the “lowland dense moist forest” in Guyana, with a peak of activity on 14 and 15 March. They also penetrated the “fragmented forest” and “dense moist forest” in Brazil after 12 March, with a peak of activity between 15 and 18 March.

Even when they were not in the non-forested domain, the fires were not positioned randomly. They were detected right at the edge of the forested types of vegetation, indicating clearly the purpose of this burning activity.

This fourth fire event has very strong environmental as well as social impact. Further investigations have to be conducted to assess and quantify in which way it affected, directly by burning fronts, or indirectly by dense smokes coverage, many indigenous areas as well as forest reserves and biological reserves.

Conclusions and Perspectives

From the present analysis of the fire patterns in this South American region during March 1998, and the comparison conducted with historical data from the fire season of 1992-93 (Dwyer et al. 1998; Stroppiana et al. 1998), one can make the following remarks.

The fire event was not totally exceptional and unpredictable. Reasons for burning are known and not new. But the intensity and extension of the fires in the dense moist forest domain (Brazil, Guyana) was effectively unusual. All the fires occurring in the region, and not only the ones affecting the forest, have strong environmental impacts and socio-economical consequences, especially because they affect vegetation on slopes exposed to soil erosion or very fragile coastal ecosystems.

Both of these first aspects were moreover minimized by the media, which focussed on the fires in the forested area of the Roraima State in Brazil.

The evaluation of the ecological and socio-economical consequences of the burning activity in this region requires an almost permanent flow of information on the location and timing of fires, at a scale that fits with the one of the phenomenon. This information must be first collected and then made available in a suitable format to the services of the Commission.

Earth Observation Systems provide here the only practical solution for such information collection. In this context the GVM unit of the SAI is currently developing the World-Fire-Web Network (Pinnock et al. 1998), a system for globally mapping fires in vegetation: a worldwide network of receiving stations will collect and process NOAA-AVHRR images for regional detection of fires. Daily global fire maps will be built up by automatically sharing regional maps over the Internet. This global fire information will then be available on-line and in near real-time. After a 12 months test period of a pilot network with partial global coverage (June 1998), the network will be progressively enlarged to give full global coverage.

As far as the information flow to the services of the Commission is concerned, the Tropical Forest Information System (Wood and Janvier 1995) is a good starting point. It is designed to be a tool for organizing and archiving information related to fire, vegetation and all topographic, geographic, and non-spatial information. It brings together the relevant datasets and allows performing the assessment and analysis of the impacts of fires on the environment. Finally it provide a tool for query, browsing and visualization of the fire products and results.

A prototype version has been installed by the TREES project at DG XI/D4. It is planned to install two others entries to the system in DG VIII and DGI/B.

References

D’Souza G., J.-P.Malingreau, and H.Eva. 1995. Tropical forest cover of South and Central America as derived from analyses of NOAA-AVHRR data, 50. Publ. European Commission, EUR 16274 EN, Luxembourg.

Dwyer N., J.-M.Grégoire, and J.-P.Malingreau. 1998. A global analysis of vegetation fires: spatial and temporal dynamics. Ambio 27 (3), 175-181.

Janodet E., A.Tournier, S.Brownlee, B.Glénat, and J.-M.Grégoire. 1996. Software package developed and implemented by MTV for the pre-processing and processing of NOAA-AVHRR images acquired with a portable receiving station. JRC Technical Note No. I.96.245, 32. Publ. European Commission, Joint Research Centre, Ispra, Italy, December 1996.

Pinnock S., J.-M.Grégoire, and J.-P.Malingreau. 1998. The World-Fire-Web project. Web Site : http://www.mtv.sai.jrc.it/projects/fire/wfw/fact_sheet.html

Stoppiana D., S.Pinnock, and J.-M.Grégoire. 1998. The Global Fire Product. Int. J. Remote Sensing (submitted 1998).

Wood R., and P.Janvier. 1995. Tropical Forest Information System: past and future priorities. Proceedings of Eurocarto XIII, Workshop on Scale and Extent, 35-49. Publ. European Commission, EUR 16406 EN, Luxembourg.

 

 

From: J.-M. Grégoire, B. Glénat,
P. Janvier, E. Janodet, A. Tournier and J.M.N. Silva
Address:
Global Vegetation Monitoring Unit
Space Applications Institute
Joint Research Centre
I – 21020 Ispra, Varese
ITALY
Department of Forestry
Technical University
Lisbon
PORTUGAL

For more information on the FIRE and TREES project’s activities: http://www.mtv.sai.jrc.it/

---

Top
IFFN No. 19
Country Notes