DEFORESTATION DYNAMICS IN MATO GROSSO, CENTRAL-WEST BRAZIL USING GIS AND NOAA/AVHRR DATA

Transcription

1 DEFORESTATION DYNAMICS IN MATO GROSSO, CENTRAL-WEST BRAZIL USING GIS AND NOAA/AVHRR DATA Sanga-Ngoie Kazadi, Professor Ritsumeikan Asia Pacific University, Beppu, Japan Sayaka Yoshikawa, PhD Candidate Graduate School of Bioresources, Mie University, Tsu, Japan ABSTRACT Since the early 1980s, large scale and deep environmental alterations in the Amazon River Basin owing to land development, population inflow, and the consequent deforestation, have become a serious ecological problem in this region which is known to be, both climatologically and biogenetically, one of the most important ecosystem in the world. Mato Grosso has emerged as the Brazilian state with the highest deforestation rate, and with the most dynamic changes in vegetation and land covers. In this paper, we focused on the following two main objectives: (i) to quantitatively assess the extent of vegetation change over the past two decades for more accurate eco-climatic impact analysis, and (ii) to clarify the causes to these changes, with special focus to human factors, with Mato Grosso as our case study. 5-year Digital Vegetation Model (DVM) Maps, aggregated into Phases I and IV, were created for every five years during the period, using the first components of the principal components analysis (PCA) of NOAA/ AVHRR multi-spectral data (Channels 1, 2 and 4). Vegetation and land cover changes are characterized by broad destruction of primary forests in the north and large-scale savanization expanding from the south. Change rates are shown to be larger over non-inhabited areas (56%) and far away from the main highways (52%), than over the populated zones in the south (42%), within 50km of the roads (44%). This emphasizes not only the role of population density and road building in accelerating deforestation, but also the importance of navigable rivers, especially over the roadless areas in the north. PREFACE About 48% of tropical rainforests on earth are found in the Brazilian Amazon Basin (Park, 1992), playing an important role in the regional weather and CO2 storage (Dale, 1994). On the other hand, since 1970 s large scale deforestation due to population inflow and the improvement of infrastructure such as construction of roads and

2 highways along with building large-scale mechanized plantations and pastures is increasingly becoming a very serious issue. In addition, a remarkable decline of annual precipitation over the past 50 years is observed (Bruijnzeel, 1996). Increased discharge of surface water, especially in the Amazon River flow rate, has been observed, more likely because of deforestation (Gentry and Lopez-Parodo, 1980). Soil erosion is another related big issue. As a result, the possibility of frequent disasters such as landslide and flood is to be considered. Similar phenomena are noted to be also taking place in tropical regions in Africa: increase in albedo and long term decline of evaporation and precipitation caused by tropical deforestation (Sanga-Ngoie and Fukuyama, 1996). Salati and Vose (1984) have noted that 74.1% of precipitation in the tropical region originates from evapotranspiration from forested lands in the Amazonia, pointing to the important role of forests in the process of moisture adjustment in the atmosphere and heat balance. In general, it is known that deforestation in tropical regions is very likely to have a great impact on the eco-climatic system on the global scale, including such phenomenon as the global warming. The Instituto Nacional de Pesquisas Espaciais (INPE), has been conducting assessment research on deforestation rate in the Amazonia using LANDSAT since 1978 (INPE, 1999). In recent years, Real-time Deforestation Monitoring System (DETER) was developed jointly with NASA using MODIS data (Morton et al., 2005). INPE (2006) made it clear that the highest deforestation rates are found in Mato Gross State (abbreviated as MT hereinafter). However, INPE did assess deforestation rates only, with no reference to changes in other vegetation covers. Errors due to climatic effects were not considered. Figure 1. The 9 states of the Legal Amazon: Acre (AC), Amazonas (AM), Amapá (AP), Maranhão (MA), Pará (PA), Tocantins (TO), Rondônia (RO), Roraima (RR), and Mato Grosso (MT), within the Amazon River Basin. MT, our study area, is located in the Southern Brazilian Amazon. MT is one of those states designated as part of the Legal Amazon and is situated in the mid-western part of Brazil (Figure 1). MT s total land area is 903,386km 2, which is about 2.4 times that of Japan. Up to 67.8% of MT s total land belongs to the southern basin of the Amazon River (ANA, 2002). MT primary vegetation is characterized by latitudinal variation: forests in the north, mixed ecotomes in the middle, and cerrado in the south (Fearnside,

3 2003). Recently, large-scale mechanized plantations and pasture lands have increased in the mixed zone and the cerrado. In particular, these are the places where the largest quantity of soybean is produced in Brazil. Further lands are therefore needed for large-scale supporting infrastructures such as roads and dams, by cutting forests. The objectives of our research are: (i) to produce more accurate vegetation maps of MT with the least dependency on climate variability, (ii) to assess vegetation changes in quantity and quality, and (iii) to elucidate the causes to these changes, with special attention to population pressure and roads building. Hereafter, the used data and analysis method is presented in section 2, respectively. Our analysis results are given in section 3. Summary are highlighted in section 4. USED DATA AND ANALYSIS METHOD We used visible band (Channel 1: 0.58~0.68μm), near-infrared band (Channel 2: 0.73~1.10μm) and thermal infrared band (Channel 4: 10.3~11.3μm) from NOAA/NASA Pathfinder AVHRR Land Data Set in four periods; (Phase I: July 1981~June 1986, Phase II: July 1991~June 1991, Phase III: July 1991~June 1996, Phase IV: July 1996~June 2001) for creating the vegetation maps for each phase. As ground truth information, we used the National Atlas of Brazil 4 th edition issued by the Instituto Brasilieiro de Geografia e Estatìstica (IBGE, 2005). Furthermore, we used the population data made available by the United Nations Environment Program (UNEP/GRID, 2000). Vector data, including those of roads downloaded from IBGE (2005) were re-edited and used for analysis. Our analytical methodology is given in Figure 2, based on Nonomura et al. (2003). This method made it possible to deferentiate open forests from dense ones, and non-vegetative regions such as deserts, taking into consideration the difference in sensible heat distribution. First, Principal Components Analysis (PCA) is performed so as to extract the first components of each of the selected channels of NOAA/AVHRR data. Using the so-obtained first principal components of each channel, false color composite were obtained, and cluster-analysis performed for unsupervised classification. We reclassified the obtained clusters using the National Atlas of Brazil 4 th edition (IBGE, 2005), the Brazilian climatic zone maps and DEM, to yield the 5-year Digital Vegetation Model Maps I, II, III and IV. The GIS Software IDRISI32 (Eastman, 2001) along with the database producing software, CARTALINX. Were used as analytical tools.

4 Figure 2. Flowchart for our analysis method. ANALYSIS RESULT 5-year DVM Maps and Vegetative Changes Between 1981 and 2001 Figure 3 shows the 5-year DVM Maps obtained through the aforementioned analysis method, and classified into 7 vegetation types: Evergreen Broadleaf Forests (EBF), Semi-deciduous Seasonal Forests (SdF), Broadleaf and Seasonal Forests (BSF), Savanna Woodlands (SW), Savanna Grasslands (SG), Savanna (S), and Deeply Altered Areas (DA). Accuracy assessment of the classified map for each phase showed consistently high values as follows:: 75.2% for Phase I, 71.9% for Phase II, 70.2% for Phase III, and 61.9% for Phase IV. Spatial vegetation cover of the 5-year DVM Maps I (Figure 3 (a)) is characterized by four main distributions: (i) EBF are largely in the north-west and the mid-north (38.1% of the entire MT), (ii) BSF in the north-east and the upper basin of the Xingu River, (iii) SW (24.05), S (2.9%) and SG (1.0%) mostly in the south, and (iv) DA (15.3%) in state capital, Cuiabá City and along the main roads and highways. The following features characterize the chronological changes. EBF are dramatically reduced from ha (38.1% out of MT s total area) in Phase I to ha (13.8%) in Phase IV. On the other hand, BSF (DA) increased a little from 13.7 (13.6) 10 6 ha in Phase I to 21.2 (18.7) 10 6 ha in Phase IV. Especially, S showed a radical increase of 7.5 times from ha (2.9%) in Phase I to ha (21.6%) in Phase IV. The emphasizes the wide-spreading savanization in the area.

5 Figure 3. (a) 5-year DVM Maps I for the July 1981 to June 1986 period in MT. (b) 5-year DVM Maps II for the July 1986 to June (c) 5-year DVM Maps III for the July 1991 to June (d) 5-year DVM Maps IV for the July 1996 to June Blue lines denote rivers, black bold lines are paved roads, red meshes are Parks inhabited by Indian and orange meshes are Areas inhabited by Indian. Causes of Vegetation Changes and its Mechanism Vegetation changes vs. population density. 5-year mean population density distribution maps of each period were classified into 12 categories and compared with the corresponding 5-year DVM Maps. Population density (x) is low in the mid-north and high toward the state capital in the south. In terms of vegetation, it became clear that EBF are mostly found in the x=0 (inhabitants/km 2 ) category, while SW and DA are dominant in 0<x=<20 (inhabitants/km 2 ) category. DA are found in the highly populated 20<x<=300 (inhabitants/km 2 ) category, with sporadically distributed S. When compared with vegetation cover changes, we could find that 56% of regions with devastated land covers (such as deforestation and savannization) were found in the x>=20 (inhabitants/km 2 ) category, while out of the remaining devastated areas, 42% were in the non-inhabited areas. Vegetation changes vs. highways. The distance from the paved main highways (y) was classified into 6 categories and compared with population density. Through all the phases EBF decreased from Phase I ( ha) to Phase IV ( ha) in the 0<y<=50 (km) band, while S increased from Phase I ( ha) to Phase IV ( ha) or 8.7 times in the same zone. Similar results were obtained for the 50<y<=100 (km) category. In the 200<y<=300 (km) category, especially in the periphery of the Juruena River, EBF decreased from Phase I ( ha) to Phase IV ( ha), while SW increased from Phase I ( ha) to Phase IV ( ha). Since 1990 s DA s entry was also observed. This shows that navigable waterways contribute to the deforestation process by facilitating transportation, in the same ways as highways. In fact, it is shown that 44% of regions with devastated land covers were in the 0<y<=50 (km) category, 43% in the 50<y<=200 (km) category, and 9% in the x>200 (km) category. Nepstad et al. (2001) indicated that 2/3 of deforestation to be within 50 km from roads. Our findings clearly show that 52% of deforestation is taking place more than 50 km away from main roads.

6 DISCUSSION AND CONCLUDING REMARKS In this paper, we obtained seven-category and high quality 5-year DVM maps Phases I IV (Figures 3(a) (d)) for four 5-year periods over the past two decades ( ). Temporal vegetation change patterns over the analysis period (summarized in Table 1) are characterized by a sharp decrease in EBF covers from the initial 38.1% of the lands to 18.5% (Phase II), to 16.5% (Phase III) and finally to 13.8% in Phase IV on one side, and the remarkable increase of thinner forest cover types (SdF and BSF), of herbaceous vegetation types (savanna complex: from 27.9% in Phase I to 35.2%, then to 38.4%, and finally to 39.7% in Phase IV), and of the DA (from 15.3% in Phase I to 20.7% in Phase IV) on the other side. A sharp decrease by 19.6% in EBF between Phase I and II is to be noted. These findings confirm, and update, those by Fearnside (1993) who notes a decrease of about 16.3% in forests covers by 1991 in Mato Grosso. The main trend in the changes (Table 1) shows the dominance of deforestation of natural forests between Phase I and Phase II, and of deforestation and savanization between Phase II and Phase III. Savanization is quite remarkable from Phase II, and is dominant later on. It results either from deforested lands changing into pasture or farm lands, or from those DA that are abandoned to savanization. S covers spread from 2.9% in Phase I to 4.1%, and then to 10.8% and finally to 21.6% in Phase IV. In general, changes tend to spread from the populated regions in the Southeast to other regions in the west and the north, along the main roads and highways and along navigable rivers in remote and roadless areas. Table 1. Trend of Vegetation changes from Phases I to IV. Solid arrows denote recovery (green), degradation (red) and transitional process (yellow), respectively. The solid arrows thickness denotes the change rates, in decreasing order: 7 10%, 5 7%, 3 5% and 1.5 3%.

7 REFERENCES Agência Nacional de Àguas, ANA, The Evolution of Water Resources Management in Brazil, ANA, Setor Policial Sul, Brazil. Bruijnzeel, L.A., Predicting the hydrological impacts of land covers transformation in the humid tropics: The need for integrated research, Amazonian Deforestation and Climate, Gash, J. H. C., Nobre, C. A., Roberts, J. M., and Victoria, R. L., John Wiley & Sons, Chichester, pp Dale, V.H., Terrestrial CO 2 Flux: The Challenge of Interdiscipilinary Research. Effects of Land-Use Change on Atmospheric CO 2 Concentrations, Dale, V. H., Springer Verlag New York, pp Eastman, J.R., Guide to GIS and Image Processing, 1, Clark University, Massachusetts. Fearnside, P.M., Deforestation in Brazilian Amazonia: The effect of population and land tenure, Ambio, 22: Fearnside, P.M., Deforestation control in Mato Grasso: A new model for slowing the loss of Brazil s Amazon forest, Ambio, 32: Gentry, A.H. and J. Lopez-parodo, 1980, Deforestation and increased flooding of the upper Amazon, Science, 210: 19. Instituto Brasilieiro de Geografia e Estatìstica, IBGE, Mapas Interativos IBGE. Instituto Nacional de Pesquisas Espaciais, INPE, Monitoramento da Floresta amazônica brasileira por satélite Instituto Nacional de Pesquisas Espaciais, INPE, PROJETO PRODES Monitoramento da Floresta amazônica brasileira por satélite. Morton, D.C., R.S. Defries, Y.E. Shimabukuro, L.O. Anderson, B.E.S. Del, F. Santo, M.C. Hansen, and M.Carroll, Rapid Assessment of Annual Deforestation in the Brazilian Amazon Using MODIS Data, Earth Interactions, 9: Nepstad, D., G. Carvalho, A.C. Barros, A. Alencar, J.P. Capobianco, J. Bishop, P. Moutinho, P. Lefebvre, U.L. Silva Jr., and E. Prins, Road paving, fire regime feedbacks, and the future of Amazon forests, Forest Ecology and Management, 154: Nonomura, A., K. Sanga-Ngoie, and K. Fukuyama, Devising a new digital vegetation model for eco-climatic analysis in Africa using GIS and NOAA AVHRR data, International Journal of Remote Sensing, 24: Park, C.C., Tropical Rainforests, Rutledge, New York. Salati, E. and P.B. Vose, Amazon Basin: A system in equibrium, Science, 225: Sanga-Ngoie, K. and K. Fukuyama, Interannual and long-term climate variability over the Zaire River Basin during the last 30 years, Journal of Geophysical Research, 101: United Nations Environment Program, UNEP/GRID, Latin America and Caribbean Population Distribution Database, United Nations Environment Programme Environment for Development, Global Resource Information Database - Sioux Falls.