Secondary Forest Age and Tropical Forest Biomass Estimation Using Thematic Mapper Imagery

The Brazilian Legal Amazon encompasses an area of approximately 5,030,000 km2, stretching from the states of Maranhão and Tocantins in the east to Amazonas and Acre in the west, and from Roraima and Amapá in the north to Mato Grosso in the south. Approximately 4,090,000 km2 of this area is forested, 850,000 km2 is cerrado (wooded grassland), and 90,000 km2 is water (Skole and Tucker 1993, Fearnside 1996). As of 1988, 230,324 km2, or 5.6%, of the approximately 4,090,000 km2 forested Brazilian Legal Amazon had been deforested (Skole and Tucker 1993, Skole et al. 1994) since theconstruction of the Belem-Brasilia Highway in 1958 (Moran et al. 1994). Each year, approximately 15,000–20,000 km2 of additional primary tropical forest in this region are cut and cleared (Skole et al. 1994), although estimates of clearing rates have varied greatly among studies and over the decades (approximately 8000–10,000 km2/yr in the 1970s, Mahar 1988; approximately 35,000 km2/yr in the 1980s, Fearnside 1989; approximately 15,200 km2/yr from 1978–1988, Skole and Tucker 1993). The carbon dynamics associated with this ongoing transformation of the Amazon are globally significant. Brazil now ranks fourth in atmospheric carbon emissions (behind the United States, the states of the former Soviet block, and China; Goldemburg 1989), in large part because of Amazonian deforestation (Moran et al. 1994). Although the general trend toward tropical forest loss in the Amazon is well documented (Setzer and Pereira 1991, Fearnside 1993, Skole and Tucker 1993, Houghton 1994), more subtle but significant issues also affect regional and subcontinental carbon budgets. Researchers now realize that forested areas that have been cleared do not necessarily remain deforested. Studies that assess biomass and carbon stocks or fluxes must consider afforestation—that is, the conversion of nonforested areas (e.g., pasture and cropland) to secondary forest—which typically results from abandonment or agricultural rotation. Secondary forests are those areas that have been abandoned and that have become revegetated after all or a significant portion of the original, primary forest has been removed. Researchers who use satellite data to monitor deforestation are recognizing that secondary forests are an important component of Amazonian land-cover change dynamics (Brown 1993, Moran et al. 1994, Skole et al. 1994, Alves and Skole 1996, Foody et al. 1996, Kimes et al. 1999). Rates of deforestation and afforestation in the Amazon are responsive primarily to political and economic policies—federal tax incentives or subsidies that promote ranching, mining, and logging—and only secondarily to population pressures (Moran et al. 1994). A significant proportion of areas that have been deforested cycle in and out of secondary forest regrowth from year to year (Alves and Skole 1996, Fearnside 1996), and this cycling has ramifications for the carbon budget of the Brazilian Legal Amazon, the world's largest remaining tropical forest. The cutting, burning, and clearing of primary forest and the subsequent cycles of afforestation and deforestation affect the amount of carbon that is sequestered and released from a particular parcel of land over time. The rate at which carbon is sequestered by secondary forests depends primarily on the age of the forest, on the number of times that the land has been cleared and used for agricultural or pastoral purposes, and on its use while cleared (Uhl et al. 1988). In this article, we explore the use of Landsat Thematic Mapper (TM) digital satellite imagery to estimate aboveground dry biomass and carbon budgets in a section of the Brazilian Amazon that is currently being settled. We developed a 7-year land-cover history of the colonized area around the city of Ariquemes in the state of Rondônia, Brazil, using multitemporal TM data (Figure 1). Seven scenes acquired during each dry season from 1989 through 1995 were each classified to identify primary forest, secondary forest, and nonforest/cleared areas. The seven forest classification maps were concatenated and used to develop land-cover trajectories that identified a particular pixel, for a given year, as primary forest, nonforest, or secondary forest. Each secondary forest pixel was further identified by age of regrowth and number of times that the land represented by the pixel had been cleared. This 7-year data set served as the digital ground reference for this study. 33,000 km2study area in Rondônia, Brazil; 25 July 1995 Thematic Mapper data. Forest conversion centers on the city of Ariquemes on route BR-364. The polygons in the expanded section identify secondary forest. Red, 1 year old; green, 2 years old; blue, 3 years old; magenta, 4 years old; yellow, 5 years old; aqua, 6 years old; white, 7 years old. Because forest age and land-use history (i.e., number of times cleared) can serve as surrogates for biomass and carbon accumulation rates in Amazonian secondary forest, the most recent of the seven images (i.e., 1995) was studied to see whether single-date satellite imagery could be used to accurately estimate the age and land-use history of the secondary regrowth. If such information can be gleaned from a single satellite acquisition, then one clear overpass might be all that is needed to develop accurate estimates of biomass and carbon sequestration rates. All too often, one clear overpass is all that is in the satellite because of from are by or If satellite data can be used to secondary forest important to biomass cycling (i.e., and land-use then more accurate carbon budgets could be developed for the Secondary forests are carbon and afforestation the carbon to the that results from forest the rate at which carbon is sequestered by tropical secondary forests This is by age of the secondary forest, land-use history of the cleared number of times that the area has been (Moran et al. 1994), and forest Houghton et al. carbon accumulation rates in tropical and forests of and 3 forests are from one from the tropical forest to the was from the forests were cleared and the land was carbon rates in secondary forests on cleared areas of tropical and forests were approximately and The carbon sequestration and biomass accumulation is although is important to be of which is being Fearnside used a conversion of for secondary forest, on by In the conversion of of aboveground dry forest biomass is to of Fearnside a conversion of for primary forest, the conversion by and et al. studied the of secondary forest age and land-use history on biomass accumulation rates in the studied abandoned in Brazil, and that used that had to forest (i.e., because were abandoned after being cleared and to dry and a rate of approximately biomass had to of primary forest abandoned approximately 1 used for years as biomass at a rate of 5 which had been to for years and then biomass at a rate of only on et al. that secondary forest age was a of biomass on and used but not on used and of tropical forest data from et in the a biomass accumulation of after for an rate of The rates over that were not Rates varied from approximately for the years to over years that tropical forests on (Brown and this the of of forest biomass in have been used to tropical deforestation is for the carbon budget because results in a carbon from the This that primary which carbon or not at be with secondary et al. although secondary forests are carbon the conversion of primary forest to secondary forest results in a to the carbon sequestration in is the of secondary forests from primary forests not atmospheric In given that significant areas of primary forest have been conversion of to be as as from secondary forests because secondary forest carbon are those of primary forests (Brown 1993). Secondary forests that from cleared areas are an important land-cover component in the Amazon, the and of the Brazilian Legal Amazon, most clearing Alves and Skole documented in over a from to that the area from approximately to over 6 a forest conversion rate of approximately clearing rate is and is of an area on the year, from to of the area by was secondary forest. Alves and Skole that abandonment of cleared areas and the clearing of secondary forests are in this particular data on and and associated in (Skole et al. 1994) and (Moran et al. 1994), Fearnside used to a of Amazonian land In now of the area and the in the year and secondary forest. Approximately of the of the and of the to secondary forest each year in this years be for years to land Secondary forest times be on the of after which to or pasture 1996). (Skole et al. a of 5 years for secondary forest in the Rondônia, is from that secondary forests a in areas being from primary forest to an forest The of secondary forests in carbon and biomass only become as more of the Amazon is cleared. Because the tropical forests biomass and carbon (Brown and et al. because tropical deforestation for approximately of the in the Houghton 1994), because to be the of carbon and because satellite data is the only to over the of tropical has been to to of forested land in the using satellite digital and data. data can be used to secondary forests from primary forests and nonforested areas in the using TM digital that primary and secondary tropical forest were to approximately the secondary forest with primary forest. In on the Moran et al. identified secondary regrowth that was years years and years old. with primary forest and were identified at using TM data. Moran et al. at in areas of at years the secondary regrowth to be to Amazonian forest. This age at which secondary and primary forest be or in tropical et al. in satellite as the of land-cover from the satellite digital data are to age from satellite digital data. Foody et al. used Landsat TM data to clearing 2 years and more Brazil, with of and As age were the of most classification Kimes et al. that secondary forest age could not be using single-date imagery from results that age that secondary with an age of 5 years or can be with using TM but that year are too to As and in the satellite data are to one clear year over a particular of tropical because the satellite cycles (e.g., for the number of and because in the are typically associated with the dry and and cleared areas. The of seven and TM scenes approximately 1 year over a area around Rondônia, a This multitemporal data set served as a or data set that to to see whether we could accurately secondary forest age using single-date TM The TM scenes were acquired from over a 7-year from 1989 to The seven TM scenes the for July 2 1991, 7 1993, 4 1994, and 25 July Because were in this such as those by Kimes et al. were not of of the seven TM were used from each the was not because of its of the from each for use in that is, the have been identified as for the of in tropical et al. 1996). All images were to the because in the satellite the of from are typically as that is, as the of the in the and from to from to Each of the seven scenes was classified forest nonforest/cleared secondary forest or and water As the secondary forests are from the primary forest, at to a secondary forest age of approximately on TM classification have of more for (Moran et al. 1994, and 1996, Kimes et al. 1999). the seven digital (i.e., the land-cover were The area to all seven scenes an area of approximately 1). the seven scenes been and a 7-year land-cover history was for each of the more A was then used to an data that the land-cover information in the land-cover This the of of land-cover trajectories to an the in The identified areas that in primary forest the 7-year areas that were nonforested the 7-year areas and areas because of or on one or more were to be of each of the in 1995 on 7-year land-cover 1989 to Mapper of each of the in 1995 on 7-year land-cover 1989 to Mapper particular were the of secondary forest with to age and clearing history on land-cover The age of a particular secondary forest pixel was as the number of years since with abandonment as Fearnside years since the The number of years since abandonment could of be for area that was in secondary forest in the 1989 and in secondary forest through the secondary forest age was a of age or to 7 an accurate of secondary forest age depends on the areas one year and not the This on the of 1 year is a of with to an accurate of secondary forest the biomass on abandoned tropical forest and et al. 1988, Houghton et al. the of this portion of forests that we classified as secondary in be agricultural (e.g., The of this secondary forest is but we to be on during to on the 7-year the secondary forest in the was identified to the number of times that the particular area had been cleared and to its age since We to the number of times cleared as areas were those that had been cleared only the TM secondary forest had been cleared 1989 and As an a for a particular pixel might be is, this pixel was classified as primary forest in 1989 and as in 1991, as secondary forest in as cleared in 1993, and as secondary forest in and This pixel be identified in the as secondary forest in The used to the ground reference data set are from In the secondary forest could change primary forest as secondary forest its to primary forest a such as be identified as secondary forest. The is that the in and 1995 (i.e., the pixel as primary forest of were given the that secondary forest to primary forest. the change from primary forest to secondary forest an of clearing was not to with a conversion was the because the to be that were with secondary forest and not digital or in a pixel being classified as because its age or clearing history was a pixel identified as water or on that pixel the for the 7-year This was developed primarily to that the secondary forest age were identified as accurately as to in for a more accurate and set of secondary forest and the area of water and and a significant number of an The land areas represented by were to the land-cover on a of a of The ground reference by the the of primary forest, nonforest, and secondary forest in the 1995 TM Each of was to identify areas to and the results from used to identify secondary forest age areas of that were of the were needed that a particular could to the of were used to an estimate of accurately a particular the that were scenes to in the polygons and to the (i.e., from the areas identified as a single and for and were in the ground reference data were then Approximately of the TM were is, represented a of trajectories that not given the were studied to of the was The of was associated with secondary forest. A of was and that the could be among the primary forest, nonforest, and secondary forest to biomass Approximately of this on clearing because of or secondary forest and primary forest of the that secondary forest from primary forest were in areas that to have been or and through a or the were primary forest in were nonforest, and one of the secondary forest The area associated with the more was the land-cover for biomass budget on from this this for the area Rondônia, are in that are have in to more accurate biomass and carbon the and associated with the land-cover and areas were identified in the 1995 TM polygons (i.e., polygons from the of were as primary forest or or as one of the of the secondary forest secondary forest was not to its and each of a of all in polygons was Each was to develop with for use in subsequent This that the were and that only a from each were in (i.e., for each of the land-cover The pixel the TM from the 1995 and digital The which are of were using the TM and were in the on the results of classification and and on by Kimes et al. The are from and were used to estimate classification associated with of primary forest, nonforest, and all of secondary to the with which the age of tropical secondary forest can be using TM or to whether the clearing history of a particular of land (i.e., number of times cleared) could be and to the of secondary forest age using et al. The results of were then used to estimate biomass and carbon budget with the use of single-date TM imagery to assess secondary forest. rates were for the in the Rondônia, area as by the history as of for biomass and carbon were also for a as by Fearnside in which the land are and secondary forest. of the area around Ariquemes is primary forest 1). which the biomass and carbon to the as using the that over of the biomass and carbon in this in 1995 was in this primary forest. aboveground dry biomass and carbon budgets for the 1995 Landsat Thematic Mapper aboveground dry biomass and carbon budgets for the 1995 Landsat Thematic Mapper that all of the secondary forest and and the of the areas in the were primary forest, a estimate of biomass and carbon for this area can be the 1970s, the Rondônia, and Mato was and et al. 1996), have been cleared or an biomass of of biomass were of carbon over this The cleared areas in this have those in of aboveground dry or of carbon by the nonforest/cleared and all secondary forest The 33,000 km2 area and of Rondônia, that is in the TM released approximately of carbon year from approximately of forest biomass the carbon to the from this area to the biomass loss rate was do not carbon Houghton et al. that biomass be as approximately of the aboveground although the biomass is (i.e., of the aboveground data such as the TM scenes that were acquired over 7 years from 1989 to 1995 are for the and are to we whether general (i.e., age or history (e.g., or could be using single-date TM were used to primary forest, nonforest, and secondary forest. a and a were that the results of a to tropical forest could be to results using a more of the of results that the use of a in with primary forest classification to Approximately of the was to the use of the was to the use of the The results of were to those in a et al. which the use of data for tropical secondary forest the we were to primary forest all of the secondary forest (Figure The was then to each of the secondary forest to whether as the secondary forest to more and more primary The was also to secondary forest age with clearing (i.e., to the of history on 2 that classification although do not with forest at the of we results by Moran et al. and who that secondary forest to the age of approximately years can be from primary forest using satellite data. As 2 also that forest on that have been cleared times are as not more from primary forest. of developed to primary forest from all of secondary forest and then to secondary forest age for each on the are on for the primary forest and for each of the secondary forest age or satellite of the Amazon, on single-date imagery to identify land-cover We to see well secondary forest could be given only a single 3 the age a was using TM and The the secondary forest year with a of years and an of In estimates of secondary forest age were in on years age 1 years and 7 years were age only of the in the age a is not to accurately secondary forest age using and single-date TM and data. secondary forest age secondary forest age using a with for each age The for age is the for age is Each of age is a a areas a of Secondary forest age and clearing history are As by et al. secondary forests that have been cleared and used for or pasture more forest. and forests to be to forests (Figure in 4 the of is the the around of the the of the remaining information clearing history be using single-date TM Thematic Mapper (TM) digital for secondary forest age and clearing in the 1995 TM The digital number the of in a particular from one or a of TM 1 TM 2 TM 3 TM 4 TM 5 TM 7 In the of the TM are to the amount of in the of TM 4 is to in and is this for significant are and The TM to water secondary forest, secondary forest, secondary forest, primary forest. Although the associated with is not one around of the of the remaining The the of the of secondary forests with clearing that single-date TM imagery be used to accurately estimate age or to the clearing history of tropical secondary forest, large are the biomass that this we to biomass budget that might be only single-date satellite imagery were in the use of single-date TM imagery is that secondary forest must be classified as a represented the (i.e., in which approximately of the Landsat TM has been from primary forest cleared or secondary forest as is the In this the biomass budget in 2 was used as ground reference (i.e., the estimate of was on the ground Ariquemes in This the biomass budget using the forest and clearing history as by the with the budget from a in which secondary forest was identified only as a single because of the to accurately the age of the regrowth. The a in one that has an as by Fearnside of agricultural and secondary forest biomass budgets were with and secondary forest age biomass budgets were to that might be secondary forest age was not in a by forest In this the of the TM is primary forest, biomass estimates were with and secondary forest age and clearing history In biomass estimates for secondary forest (i.e., age rates were 5 and (Uhl et al. 1988). was also that secondary forest on 5 years (Skole et al. 1994, Fearnside 1996). We used the estimates in 2 in with and age to biomass for forest and for the 33,000 km2 TM The results age secondary forest biomass were approximately 5 years to 5 years were to the secondary forest biomass as a of age in 2 (i.e., to the TM the biomass with and age information were In to estimate the biomass of secondary forest in and of age information is The of that information can to which on the age and accumulation rates of the secondary forest. one is in the biomass over an TM in which approximately of that has been to secondary forest, information the age of that secondary forest is of This the that approximately of the area is in primary forest, which the of the aboveground in areas such as Rondônia, primary forest conversion the of primary forest with and the area of secondary forest budget become more as the of secondary forest to primary forest The the of secondary forest age information in a all of the forest has been 2 at a in which of the primary forest a TM has been to agricultural or secondary forest. Fearnside that colonized areas of in be of primary forest, and secondary forest. primary forest as being secondary forest more years the amount of biomass as the primary forest. that the areas be of pasture and pasture the secondary forest, of the area from abandoned and on abandoned were with the of the TM in 1995 to the in area and biomass for the Thematic Mapper in estimates were that the area of water and not change over the years and that the of the secondary forest was to the 1995 area and biomass for the Thematic Mapper in estimates were that the area of water and not change over the years and that the of the secondary forest was to the 1995 As 3 years secondary forest for over of the land area in this TM and of the biomass in this secondary forest. additional of the biomass in secondary forest more years old. as area secondary forest a in biomass and carbon state and In the of this article, the is, are in biomass estimates secondary forest age information is not 3 a and ground reference biomass biomass estimates were the age and clearing history As in was that the secondary forest on 5 years and that biomass at 5 and Because the age and clearing history were used in the 1995 and the rates associated with secondary forest biomass were a a because secondary forest such a large of the TM and the of the associated with biomass estimates were in the 1995 accumulation rate of 5 a biomass rate of an accumulation rate of in a biomass are the rates of for the 1995 as forest conversion and as secondary forest the the age and clearing history of the secondary forest have a on biomass and carbon state The we have are the biomass accumulation rates were used to the ground reference data age and the age secondary forest biomass were as not to and we that to was to the of secondary forest age information on the biomass and carbon budgets of a tropical forest that secondary forest age information is to accurate estimates of biomass and carbon in secondary forests to and also in those in which the of the biomass in a given area in secondary forest. also that single-date Landsat TM data can be used to accurately and secondary forest from primary forest and single-date TM and data be used to secondary forest age and clearing The estimates of that might be secondary forest age information is not biomass on the of to were in the in which secondary forest age information was not and the region was by and secondary forest studies that have the of Amazonian deforestation in of biomass or carbon have a transformation from forest to pasture or from forest to The is to remain in that state the of secondary forest in of biomass accumulation and to biomass and carbon state and 1996). of biomass and carbon in the Amazon have to secondary forests in the budget and consider at the age of the secondary forest in as more land is biomass secondary forest land is on the of be the age of the secondary forest is not in the of that single-date TM imagery secondary forest age on TM and data. studies have age (e.g., and years Moran et al. and years et al. and studies have that secondary forest and primary forest satellite after approximately years (Moran et al. 1994, Foody et al. 1996, in Amazonian is that secondary tropical forests 1 and years biomass at a rates then in secondary forest and and a of secondary forest age (e.g., those of Moran et al. 1994) and the of general biomass accumulation rates a in is in the as to biomass accumulation rates might be to general age In and rates and and et al. 1988, and Moran et al. 1994, et al. for a of We on that the rates might of aboveground dry biomass in secondary forest years estimate on et al. 1988, and Houghton et al. 1991, and Moran et al. 1994) and 4 in secondary forest of years (Brown and as in Foody et al. 1996). Although that single-date TM imagery be used to accurately estimate age in secondary tropical forest, studies by that age and years can be of biomass accumulation rates to age associated with biomass and carbon and of in carbon estimates become as issues and the economic of those issues to the in the This was by the and of to We the of who the data for