GIS APPLIED TO THE STUDY OF TEMPORAL RECOVERY OF BURNED AREAS IN THE MUNICIPALITY OF PAI PEDRO LOCATED IN REGION NORTH OF THE STATE OF MINAS GERAIS

The northern region of Minas Gerais is considered a transition area between the biomes Caatinga and Savanna, with predominance of the Savanna and its variations. It is common to the occurrence of fires, which can be caused naturally or by anthropogenic actions, resulting in changes in the landscape. In this sense, the study aims to evaluate from time series of LANDSAT-5/TM and MODIS/TERRA images the process of natural recovery of vegetation after the identification of operational pixels corresponding to the burned areas. The results of the research are based on the response of the municipality test of Pai Pedro/MG. We can observe that was detected a hotspots pixel on the day 08/04/2008 through images MOD14A1, which validation was performed from the survey of burned areas conducted by INPE National Institute of Space Research. From the identification of operational pixel of burned was evaluated the dynamics of recovery of the area through the vegetation index NDVI (Normalized Difference Vegetation Index), generated by images MODIS/TERRA, product MOD13Q. The scars of burned analyzed, approximately 52% were detected. Despite this limitation, much of the area impacted was identified. The errors of omission were considered satisfactory, with a view to the technical limitations of spatial resolution of the sensors used.


Introduction
The state of Minas Gerais has valuable Environmental Conservation Units and a growing need to eliminate the risk of fires in these areas.The effective control of the sources of risk requires the knowledge of how these operate locally and when and where fires occur more commonly.1984) and changes in soil moisture, especially when associated with intensive grazing, as Pillar & Tables (1997).According to Coutinho (1990), the burning of vegetation such as

Material and Methods
The area under study involves the municipality of Pai Pedro located in the mesoregion north of Minas Gerais (Figure 1).
The municipality has an area of 839.804km² and according to the census of agriculture has 28.980ha of land for pasture and 5.262ha for practices of crops (IBGE, 2010).Selection of pixels of the places that were observed the fire spots in image MOD14A1, through the process of slicing, being the class interval corresponding to the grade level of the incidence of burned.For this paper was considered the value of level of gray more than The NDVI/LANDSAT-5 was compared to the NDVI/MODIS in order to establish a response parameter (consistency) between images, by choosing the resolution of space of 30m as a reference.The pixels were mapped and converted to a vector file, where it was possible to follow the behavior of the surface until the total disappearance of the scar of burned.

Results and Discussion
After validation of outbreaks of fire, about 75% of pixels of outbreaks diagnosed by  responsible for these variations (Tanaka et al., 1983;Frederiksen, 1990;Robinson, 1991).
According to Pereira et al. (1997), and it's important to know the time interval between the occurrence of fire and the date of acquisition of spectral data due to the changes that will occur in a of these areas, occasioned by the recovery of vegetation affected by fire.
Time series of NDVI/MODIS images (Figure 3) and submitted the response related to the regeneration of the area over the period studied.Note that in the month of February/2009 (Figure 3D), was observed, greater vegetative vigor of the analyzed area.
This response was similar to the month of May/2008 (Figure3A) period before burning.
However, the pixels of burned in some regions still had a value well below the expected, which shows that at some points there is presence of the scar of burned, and only in the month of March/2009 there is a complete recovery, i.e. there is no pixel with expected value of a burned area.This dynamics can be observed in Figure 3. low energy emitted at these events.Table 1 presents this monitoring by checking the actual area before and after burning.

Conclusions
The frequency of outbreaks of heat showed that the differences spectral, temporal and thresholds of the proposed algorithm interfere in the monitoring of the burning of vegetation, which assumes a greater amount of satellites, expanding coverage of information.
The scars of burned analyzed, approximately 52% were detected.Despite this limitation, much of the area impacted was identified.The errors of omission were considered satisfactory, with a view to the technical limitations of spatial resolution of the sensors used.

References
These information are linked to an individual record of occurrence and this record is the main source of all the statistics about the quantification of outbreaks of fire.The more frequent data for prevention programs are: the causes of fire; the time, place of occurrence and the extent of the burn area.The northern region of the state is considered a transition Corresponding author: crisrodnas@gmail.com.area between the biomes caatinga and savanna, with predominance of the Savanna and its variations, making it extremely sensitive as the fire action.In large part to the region, can be observed the use of agricultural practices, where it is common the use of fire as a management tool in opening up new areas for farming, as also in the control of pests of pastures, crops, and to eliminate leftovers of pasture aging, bringing the medium and longterm implications.The impacts caused by the action of fire may result, directly or indirectly, major changes in soil and vegetation, mainly due to the reduction of dead plant material, vegetation cover (De Castilhos & Jacques,

LANDSTA
practice management for the creation of cattle and one of the main activities associated with the fire in the region, making use of extensive areas of natural pasture in the forms of Cerrado more open, as field clean and dirty field.It is worth noting that there are also cases of burning that occurs by the need for renewal of the same, which is very common in the dynamics of the area.In accordance with Florenzano (2007), the detection and monitoring of the outbreaks of burned transcend to the problem of deforestation and its consequences in themselves, in this way the Remote Sensing can help in the acquisition of spatial information and temporal, which allow the characterization of occurrences of outbreaks of heat, in addition to the measurement of the area and biomass actually affected by the fire, providing important contributions to studies on this topic, relating these issues to the environment, and their ecological effects, climatic and atmospheric chemistry.The knowledge of the causes and the frequency with which events of fires occur is of paramount importance, especially taking into consideration that the starting point for the preparation of plans for prevention and know who (or what) started the fire (Santos, 2004).In this context, the National Institute for Space Research/INPE since the 1980s has been improving a system for detecting fires from images of sensors on board satellites polar and geostationary.They are the so-called "hotspots", which are geographic points captured by spaceborne sensors on the surface of the soil, when detected temperature above 47°C and minimum area of 900m² (Gontijo et al.,2011).The INPE disclose data derived from polar-orbiting satellites AQUA, TERRA, NOAA's-15, 16, 17, 18 and 19, and the geostationary satellites METEOSAT-02 and GOES-12.Each satellite polar produces two imaging per day, and the geostationary generate some images per hour, and in total the INPE handles more than 200 images per day, specifically to detect outbreaks of burning of vegetation.With regard to outbreaks MODIS (AQUA and TERRA), INPE has developed an algorithm for detection of outbreaks of burning in a conservative way, so as to minimize false alarms associated with noise and the solar reflection occasional in water bodies and exposed soil in daytime images (INPE, 2010).The accuracy of detection algorithms and the reliability of data generated should therefore be evaluated, to estimate its uncertainty and improve existing products.The information on the accuracy of the algorithms must be constantly updated because, over time, the performance of the sensors and their conditions radiometric characteristics are altered.This analysis can be performed from the mapping of scars burned in scenes of average resolution and the comparison of these with the spatial location of outbreaks of burned in images of spatial resolution low.In this context this paper had as its goal the creation of an algorithm that enables the identification of operational standards and/or scars of fires, from the association of a geographical information system and temporal data obtained from NDVI/MODIS images and MOD14A1, taking as reference archives of outbreaks of heat released by INPE.To characterize the recovery of the areas burned form used images
because in this class is the outbreaks that are detected by MODIS.2. Weighting of pixels selected images of MOD14A1 with the objective to create a final image that contains only the region of burned with his level of gray classified according to the above item related, i.e. it is attributed to pixels not belonging to this condition a value equal to zero. 3.In the detection of the area effectively burned maps to the extend of vegetation destroyed by fire from the change of spectral characteristics of the images obtained before and after the occurrence of fire, according to the methodology proposed by Rudorff et.al (2007).From the operational identification was performed to validate the focus by displaying the actual area burned, taking as reference images LANDSAT-5/TM.The procedure adopted was the Boolean operation that performs a test from the result of the operation of the difference of NDVI before and after the process of burned itself.The technique for the detection of changes used was the subtraction images according to the methodology proposed Mather(1999).It was established the following sentence of decision: If the image difference have a negative value and have pixel corresponding in weighted image MOD14A1 then burned area confirmed and/or detected.Once identified the burning was carried out the monitoring of surface, through the temporal analysis of the values of vegetation index NDVI.
MODIS and 42% of the points of hotspots identified a burned effective in the field.To analyze the points coincident MODIS x outbreaks of heat, the hit was 80% of pixels belonging to the class burned.The amount of outbreaks of heat coming from each detection methodology, carried out by satellites show that the spectral differences, temporal and thresholds of algorithms, interfere with the effective monitoring of the burning of vegetation.This result also highlights the fact that outbreaks of heat are diagnosed by different types of satellites, and then a higher sensitivity when there are sudden changes in temperature, and can be more precise the smaller the area burned, causing also in the incidence of observations from places that there was not a burned of fact.In remarks made by the junction of those files, the dense areas of points of hotspots tended to coincide with the pixels observed by MODIS, characterizing a burned significant that covers a larger area.The results obtained by the junction of the MODIS sensor and points of hotspots are valid and significantly increases the probability of identifying operational of outbreaks of burned by the algorithm, with a higher margin of safety thus reducing the need for the use of LANDSAT-5/TM images for validation.Time series of NDVI images/MODIS presented in Figure 2, it is possible to observe the response of temporal evolution of the vegetation in the area burned, taking the month of May/2008, as the beginning of the analysis.

Figure 2 .
Figure 2. Analysis of the sample mean of the values of NDVI/MODIS of temporal series.

Figure 3 .
Figure 3. Monitoring of the scar of burned and natural regeneration of the area in the municipality of Pai Pedro/MG.NDVI image from MODIS sensor product MOD13Q1, for the months may/2008 (A), august/2008 (B), january/2009 (C) and February/2009 (D).

Figure 4 .
Figure 4. LANDSAT-5/TM image in May/2008 showing the location of burned (A), and identification of operational scars burned in function of the response of NDVI/MOD13Q1 for the months August/2008 (B), January/2009 (C) and March/2009 showing the final result of the processing performed by the algorithm.

Table 1 .
Table quantification of areas in hectares, diagnosed by detection algorithm automatically burned areas.