Amazon Biome ATBD_R. Algorithm Theoretical Basis Document & Results. Instituto do Homem and Meio Ambiente da Amazônia - Imazon
|
|
- Karen Warren
- 6 years ago
- Views:
Transcription
1 Amazon Biome ATBD_R Algorithm Theoretical Basis Document & Results Instituto do Homem and Meio Ambiente da Amazônia - Imazon Centro de Geotecnologia do Imazon - CGI February,
2 Executive Summary This ATBD_R is a dynamic and evolving document. Each time we launch a new collection, this document will be revised accordingly. By reading the document one should be able to understand how the collection of maps was produced, what datasets and methodological approaches were utilized and the rationale and theory behind these approaches. 2
3 Content 1. Introduction Study Area Key Science and Applications 8 2. Overview and Background Information Context and Key Information Existent Maps and Mapping Initiatives Algorithm Descriptions, Assumptions, and Approaches ImgTools Algorithms Collection 1: MapBiomas Amazonia Algorithm Data availability and screening Classification scheme and parameters Data volumes and processing issues Collection 2 - MapBiomas Amazonia Algorithm Data availability and screening Classification scheme and parameters Data volumes and processing issues Validation Strategies Map Collections and Analysis Practical Considerations Concluding Remarks and Perspectves References Appendices Supplementary Material Algorithm Code 35 List of Figures 3
4 Figure 1. Location of the Amazon biome in Brazil. 6 Figure 2. Forcing factors are indicated with red ovals; processes addressed in this Review are indicated by green boxes and arrows; and consequences for human society are indicated by blue boxes with rounded corners. 7 Figure 3. Methodology applied in processing Landsat images and implemented in the ImgTools software. 12 Figure 4. Stages for processing of Landsat images in order to generate Collection 1 of the MapBiomas Amazon project. 13 Figure 5. Distribution of mapping charts for MapBiomas Amazonia. 14 Figure 6. Coverage of Landsat scenes to make up the temporal mosaic of the charts processed for the Amazon biome for the period of 2008 to 2015, with up to 30% of cloud cover. 15 Figure 7. Spectral curves of the pure components utilized in the MME applied to Brazilian biomes; (a) for Landsat 5 and 7 in surface reflectance, (b) for Landsat 8 for reflectance at the top of the atmosphere. 16 Figure 8. Empirical decision tree applied to the NDFI, GV and GVs to map the classes of Forest and Non-Forest. Masks for Water and Cloud are applied to make up the final map. The variables (V1, V2, V3, V4) are adjusted for each chart. 18 Figure 9. Frequency distribution of the values assumed by variables v1, v2, v3 and v4. 18 Figure 10. Empirical decision tree applied to Collection 2 for the amazon biome. 22 Figure 11.Coverage of Landsat scenes to make up the temporal mosaic of the charts processed for the Amazon biome for the period of 2000 to 2016, with up to 30% of cloud cover. 26 Figure 12. Spatial distribution of the sampling points for analyzing the accuracy of Collections 1 and 2 of MapBiomas Amazonia. 28 Figure 13. Sampling scheme and interpretation grade of Landsat pixels utilized in visual interpretation for analysis of accuracy. 29 Figure 14. Preliminary result of the accuracy analysis of the Amazon biome for Collection 1. Accuracy by year (Level 3) 30 Figure 15. Data classification panel for Collection 2 MapBiomas Amazonia (A), and a more detailed view of the classification map for 2008 (B). 31 4
5 List of Tables Table 1. Reference maps utilized in the MapBiomas Amazon project. 10 Table 2. Number of images utilized per year, sensor and maximum percentage of cloud cover accepted for Collection Table 3. Rules for reclassification applied by the temporal filter. 19 Table 4. Temporal filter rules applied to Collection 2 land cover classes. 23 Table 5. Number of images utilized per year, sensor and maximum percentage of cloud cover accepted for Collection Table 6. Classification scheme of Collection 2 for the Amazon Biome. 26 5
6 1. Introduction 1.1 Study Area The Amazon Biome has an extension of approximately 4,196,943 km², representing 49.29% of Brazilian territory. It fully occupies the States of Acre, Amapá, Amazonas, Pará, and Roraima, and sizable portions of Rondônia (99%), Mato Grosso (54%), Maranhão (34%) and Tocantins (9%), according to the IBGE (2004). The largest biodiversity reserve in the world, it is mainly covered by Dense Tropical Rainforest, characterized by evergreen vegetation and a high density of shrubby and arboreal species. The region is marked by an occupation process many times associated with social conflicts, deforestation and degradation of natural resources. Data from the National Institute of Space Research (INPE) reveal that, by 2015, almost 20% of Amazon forests had already been deforested (PRODES, 2016). Imazon indicates that more than 300 thousand kilometers of clandestine roads have already been opened in the Amazon biome, providing access to illegal logging and mining and the degradation of natural resources. Despite those problems, the Amazon biome still retains 80% of its forest cover and vast expanses of protected areas (PRODES, 2016). Figure 1. Location of the Amazon biome in Brazil. 6
7 1.2. Key Science and Applications The Amazon biome is one of the most intensely studied biomes in Brazil. Several multidisciplinary projects have been conducted in this region, notably The Large-Scale Biosphere-Atmosphere Experiment in Amazonia ( LBA ) conducted from and led by Brazil. This region has also been monitored by INPE since 1988 through the PRODES project. Independent forest monitoring has also being conducted by national and international organizations (e.g., Imazon, University of Maryland GLAD systems, Terra-i and Global Forest Watch ). None of these projects focus on annual historical (1985 through the present) mapping of land cover and land use classes with Landsat imagery. MapBiomas Amazonia has the ambitious proposal of doing that. There are several applications that can be developed for the Amazon region with the MapBiomas maps, including: Mapping and quantifying land cover and land use transitions. Quantification of gross forest losses and gains. Monitoring of secondary growth forests. Monitoring of water resources and their interaction with land cover classes. Expansion of infrastructure and urbanization. Regional planning. Management of Protected Areas. Estimation of greenhouse gas emissions. Scientific applications can also benefit from the MapBiomas results for the Amazon biome. Interactions between global climate, land use, fire, hydrology, ecology and human dimensions, already reported as critical for the future of the Amazon region (Figure 2) (Davidson et al., 2012), can be better understood with long time-series of land cover and land use maps, as provided by MapBiomas Amazonia. 1 7
8 Figure 2. Forcing factors are indicated with red ovals; processes addressed in this Review are indicated by green boxes and arrows; and consequences for human society are indicated by blue boxes with rounded corners. 2. Overview and Background Information In March 2015, at a technical seminar on use of remote sensing and vegetation mapping data, the Project for Annual Mapping of Land Cover and Use in Brazil (MapBiomas) began its research activities. In that seminar, held at the Getúlio Vargas Foundation (FGV) in São Paulo, specialists in the areas of remote sensing of the various Brazilian biomes and on mapping of land use gathered to discuss the possibility of producing annual maps of land cover and use for the Brazilian territory in a rapid and low cost form. The purpose of the Project is to create updated maps and recover information on land use and cover over the last thirty years with Landsat images. Additionally, MapBiomas utilizes computer technology on clouds to process the Landsat images and web platforms for accessing and disseminating data. Imazon has been present since the beginning of the MapBiomas project and contributed to the technical and scientific development of the project. The algorithms for processing Landsat images for application of a linear pixel mixture model (MLMP) and for calculation of the NDFI (Normalized Difference Fraction Index) (Souza et al., 2003) are examples of contributions by Imazon. Additionally, the Institute has contributed through research, development of the ImgTools software (Souza Jr. & Siqueira, 2013) and mapping of deforestation and forest degradation in the Amazon biome, publication of several papers on this issue and capacitation of several professionals working in this area. That knowledge learned and accumulated by the Institute has been shared with researchers interested in mapping pressures on tropical forests around the world. MapBiomas received that collection of knowledge, the algorithms implemented in ImgTools, and adapted a large part of the methodologies and algorithms for processing the satellite data utilized (and in some cases developed) by Imazon in its research carried out over the last decade. Imazon s experience in processing of large-scale satellite images contributed to MapBiomas obtaining the first version of its annual forest maps for all Brazilian biomes in only six months. The methodology utilized by Imazon for processing satellite images and generating data on deforestation and forest degradation gave rise to the ImgTools software (Souza Jr. & Siqueira, 2013). This software is a package of tools and algorithms for processing and classifying satellite images of the Landsat series (5, 7 and 8) and served as the basis for developing processing codes for the MLMP, NDFI, Empirical Decision Tree Builder, and spatial and temporal filters, currently implemented on the MapBiomas Workspace platform. That platform is a web application that uses the Earth Engine (EE) computer cloud to carry out the processing of remote sensing data collections in a parallel and distributed manner. EE was developed by Google in order to in a few seconds process images that a single computer would take hours or even days to do. The combination of the algorithms base from ImgTools and the EE processing power was one of the points that made possible production of land cover and use data for the Amazon biome in this project Context and Key Information 8
9 Land use in the Amazon biome is quite heterogenous and complex throughout its 4.1 million square kilometers (MMA, 2013), a reflex of the historical dynamics of occupation of that territory, and understanding that process is essential in studies on biodiversity, climate change and environmental modeling, as well as for conceiving and following up policies directed towards land use (Almeida et al, 2016). Up to 2014 land cover of altered areas in the Amazon was made up predominantly of pastures, with approximately 479 thousand km² or 63% of the biome, and in smaller proportions by secondary vegetation (22.8%) and areas with annual agriculture (6%) (INPE, 2016). This reflects the current and historical dynamics of occupation of the Amazonian space, which, without planning, generate negative socioeconomic and environmental impacts to this day, such as illegal deforestation and land-grabbing on public lands, all of which ends up impacting ecological services, biodiversity and water and carbon cycles (Fearnside, 2006) Existent Maps and Mapping Initiatives The growing concern with the consequences of climate changes has generated interest in questions related to monitoring of tropical forests and to actions for fighting deforestation in the Amazon. The first initiatives were developed for the purpose of mapping and quantifying the occurrence of deforestation in the region in order to advise the Brazilian government in devising public policies. The National Institute for Space Research (INPE) was a pioneer in monitoring deforestation of primary forests through clear-cutting, with a minimum area mapped of 6.25 ha, beginning in 1988 through the PRODES project, which utilizes Landsat satellite images to generate the official annual deforestation rates for the biome. The dynamics of deforestation in the region have presented changes over time and the significant increase in rates at the beginning of the years 2000 pointed to the need for obtaining information on deforestation in a more expedited manner in order to enable identification and punishment of the offenders, not allowing the expansion of deforested areas to increase until the closing of the deforestation calendar. In that context, the deforestation alerts systems come into play, with the objective of making available to enforcement agencies the location of occurrences of clearcutting and forest degradation, as well as informing society with respect to the situation of forests in the Amazon. In 2004 the System for Detecting Deforestation in Real Time (DETER) began operating and generating monthly deforestation alerts. DETER employs daily images from the MODIS sensor of the Terra satellite and is able to detect deforestation events in areas as low as 25 ha. In 2006 Imazon s Deforestation Alert System (SAD) began its activities with its independent monitoring also using images from the MODIS sensor, but with a distinct method for processing data, which makes it possible to detect alerts of both clearcutting and forest degradation, due to timber harvesting or to forest fires, on areas greater than 10 ha. Making available information on deforestation in a rapid and precise fashion is linked to speed in processing the data. SAD has carried out a partnership with Google so that the stages of processing and validating the alerts are performed on the Google Earth Engine platform, which makes it possible to access, organize and visualize geospatial data in the cloud. Thus, beginning in 2012, processing of MODIS images, which was previously done on Imazon computers and Imazon, began to be carried out in the Google computer clouds, making it possible to have communication of deforestation alerts to enforcement agencies and society in a shorter period. 9
10 The growing access to timely and free of charge geospatial data via computers (clouds) has favored the development of applications on larger scales and with more precise sensors. That is the case, for example, with the Global Land Analysis and Discovery (GLAD), a system that generates deforestation alerts in tropical forests of Brazil, Peru, Republic of the Congo and Indonesia through Landsat satellite images. The main initiatives for mapping and monitoring of the Amazon biome are presented below in Table 1. These data serve as a reference for mapping of the Amazon biome through the MapBiomas project. Table 1. Reference maps utilized in the MapBiomas Amazon project. Systems for Monitoring and Mapping Acronym Start of Activities Objective Website Projeto de Monitoramento do desmatamento na Amazônia Legal por Satélite PRODES 1988 Generate annual rates of deforestation e.br/prodes/index. php Detecção de Desmatamento em Tempo Real DETER 2004 Monthly deforestation alerts e.br/deter/ Sistema de Alerta de Desmatamento do Imazon SAD 2006 Monthly deforestation alerts and forest degradation ngeo.org.br/imazon geo.php Global Land Analysis and Discovery GLAD 2016 Monthly deforestation alerts u/ Levantamento de informações do uso e cobertura da terra na Amazônia TerraClass 2004 Map land use and cover /cra/projetos_pesq uisas/dados_terracl ass.php Mapeamento da Degradação Florestal na Amazônia Brasileira DEGRAD 2007 Annual mapping of forest degradation e.br/degrad/ Red Amazónica de Información Socioambiental Georreferenciada RAISG 2010 Map deforestation and pressure factors on the forest mbiental.org/ Global Forest Watching GFW 1997 Map loss and gain of forests orestwatch.org/ 3. Algorithm Descriptions, Assumptions, and Approaches The algorithms, mathematical models for processing of Landsat images, assumptions and a methodology applied in the Amazon MapBiomas project are presented in the sections below. First we present the algorithms for ImgTools, developed by Imazon and which serve as the basis for processing the Landsat images in order to generate Collections 1 and 2 of MapBiomas 10
11 Amazonia. The adjustments and improvements applied in those collections are presented in the correspondent sections of those collections. 3.1 ImgTools Algorithms Collection 1 of the MapBiomas Amazon project, which covered the period of 2008 to 2015, was generated based on methods developed by Imazon (Souza et al., 2005; 2013, Souza and Siqueira 2013; Souza 2012). The processing follows the approach originally implemented in the ImgTools software, which was re-codified in the MapBiomas Workspace platform for processing in Earth Engine. We initially presented the ImgTools approach and the adaptations and improvements developed in MapBiomas Amazonia. ImgTools was developed to integrate fundamental image processing steps in order to accomplish classification of very large Landsat data sets across space and time. These steps include: (1) Pre-processing; (2) Construction of a spectral library; (3) Spectral Mixture Analysis; (4) Building Empirical Decision Tree; (5) Post-classification. All these steps are explained in detail below (Figure 2). Step 1 Pre-Processing. The first step in the methodology consisted of producing annual data sets of Landsat images for mapping in a per scene basis. The Landsat data used had already been orthorectified and radiometrically corrected (atmospheric correction and removal of smoke noise using the Carlotto (1999) method (Souza Jr. and Siqueira, 2013). Step 2 Construction of a spectral library. This consisted of identifying pure spectral components (endmembers) in order to estimate their abundance in each pixel. The endmembers of interest for forest monitoring are: green vegetation, non-photosynthetic vegetation (NPV), soil, cloud and shadow. The process for identifying and creating the spectral library involves identifying the spectral curves that have the potential to be endmembers. Next, a visual inspection of those curves is made for the final selection of the set of those pure components. Detailed information about these procedures are described in Souza et al., (2005). Step 3 Spectral Mixture Analysis (SMA). The reflectance images processed in Step 1 are combined with the endmembers obtained in Step 2 to estimate the abundance of those pure components in each pixel of the image. To do that, we applied the SMA technique, using the spectral library of those pure components. The fraction images resulting from the SMA were used to calculate the NDFI (Normalized Differencing Fraction Image) (Souza Jr. et al., 2005), and combined in a classifier through a generic empirical decision tree in the next stage. Step 4 Empirical Decision Tree Classification. We built the decision tree empirically in order to classify the deforested areas, degraded forests, water bodies, shadow and cloud (Figure 3). As an entry for classification we used the fraction images obtained with the MME and the NDFI spectral index NDFI. 11
12 Figure 3. Methodology applied in processing Landsat images and implemented in the ImgTools software. Step 5 Post-Classification. Spatial and temporal filters were constructed in order to remove spurious classifications and non-allowed temporal transitions between mapping classes, respectively. The spatial power makes it possible for given classes to be defined together with their least mappable area. That way, isolated pixels classified in a given image can be eliminated and substituted by the dominant class in their neighborhood. The temporal filter is a set of rules for non-allowed transitions that are applied to each image classified in a given year. That way, it is possible to remove clouds and correct non-allowed transitions. 3.2 Collection 1: MapBiomas Amazonia Algorithm Collection 1 of the MapBiomas Amazon project was generated based on collections of 1 Landsat images (5, 7 and 8) with standard for correction L1T, available at the Earth Engine platform. Processing of Landsat images was conducted on the Google Earth platform following the stages described in Figure 1. Those stages were implemented in the MapBiomas Workspace Platform (Figure 4), following the processing steps and the algorithms of ImgTools for calculating the SMA, NDFI, empirical decision tree and spatial and temporal filters. The principal difference in applying those algorithms in Collection 1 do MapBiomas is in the mapping area that went from scene to mosaic (see Stage 1 below for more details). [English translation of table below on page 52] 1 Landsat image collection IDs for: i) Landsat 5: ; ii) Landsat 7:., and Landsat 8. 12
13 Figure 4. Stages for processing of Landsat images in order to generate Collection 1 of the MapBiomas Amazon project Data availability and screening The mapping unit for MapBiomas is the International Map of the World to the Millionth on the scale of 1: , covering an area of 1 30 of longitude by 1 of lattitude (Figure 5). A total of 294 charts cover the Amazon biome, and 2 to 6 Landsat scenes are necessary to cover the area of each chart, with the great majority having 4 scenes. Table 2 shows the number of images utilized in each year analyzed for the Amazon biome. For each chart, the following parameters of the Landsat images are defined: i) the year of mapping (from 2008 to 2015 for that phase of the project). ii) the period during the year for selecting images (t0 and t1 on day/month/year) que that enables the maximum spectral contrast between the forest and non-forest classes. In the case of the Amazon biome, we concentrated mainly (>90%) on the period of June 1 to October 31 for selecting the Landsat images. iii) the maximum cloud cover in the Landsat images that will be utilized in processing, which was 30% for the majority of charts, although in areas with less frequent cloud cover we utilized up to 70% of cloud cover. iv) the type of sensor (Landsat 5 [L5], Landsat [L7], and Landsat 8 [L8]). 13
14 Figure 5. Distribution of mapping charts for MapBiomas Amazonia. Those parameters are obtained from the metadata file of the collections of Landsat images available on the Earth Engine platform. We utilized collections of ortho-rectified surface reflectance images (L1T) for L5 and L7, with reflectance from the top of the atmosphere for L8. Table 2. Number of images utilized per year, sensor and maximum percentage of cloud cover accepted for Collection 1. Year Number of Landsat images Amount size (GByte) Biome data cover (%) , , , , , Total , The application of the parameters described above (period and cloud cover) made it possible to obtain a set of scenes from Landsat images for a given year, in a defined period, 14
15 covering the area of the chart selected for processing (Figure 6). That set of Landsat scenes, shown in Figure 6, is processed with the objective of generating a mosaic [temporal] with the lowest cloud cover for each chart, optimizing the period of greatest spectral contrast of the land cover classes that will be mapped. One may observe that the central-southern portion of the arc of deforestation has the largest number of Landsat scenes with more than 6 scenes for the period of June 1 to October 31. In the northern portion of the Amazon biome, we can observe gaps with scenes having up to 30% of cloud cover. For those areas, we had to expand the cloud percentage (up to 70%) in order to obtain the temporal mosaic. Figure 6. Coverage of Landsat scenes to make up the temporal mosaic of the charts processed for the Amazon biome for the period of 2008 to 2015, with up to 30% of cloud cover. The Landsat scenes selected in order to generate the mosaics for the charts were then processed to generate the temporal mosaic. Initially, a buffer zone is applied along the scene of the Landsat image to remove possible edge noises. Landsat scenes that do not have data on all bands are considered to be noise. The next step is to apply an algorithm for detecting and removing clouds, based on spectral mixture modeling (MEM; see Stage 3 for details regarding that method), and images of the thermal band of the electromagnetic spectrum (Band 6 for L5 and L7, and Band 10 for L8; all converted to the Celsius temperature scale). Pixels with a cloud fraction greater than or equal to 10% and temperature lower than or equal to 22 degrees Celsius are detected initially as clouds. A surrounding region of 10 pixels (300 meters, considering a Landsat pixel Landsat as 15
16 30m) is generated along the areas initially detected as having having clouds, utilizing a circular matrix filter ( kernel ). That process allows detection of areas with shadow generated by the clouds originally detected. In those areas the cloud fraction percentage in the pixel diminishes, and because of that, all the pixels surrounding the clouds, with values greater than or equal to 7% of the cloud fraction are added to the areas initially classified as clouds. That process generates the areas with clouds for all of the Landsat scenes that make up the area of the chart, for the year and period selected. The areas classified as Clouds are ignored in the subsequent processing through generation of a mask applied to those regions. Finally, a median filter is applied to the pixels not covered by clouds in the images, to compose the final image of the chart, known as the temporal mosaic Classification scheme and parameters The inputs for classification were generated using the MEM technique, which makes it possible to estimate composition of the pixel in terms of pure components (spectral signatures). The MEM applied in MapBiomas utilizes a spectral library defined by five pure components: Vegetation, Non-Photosynthetically Active Vegetation, Soils and Clouds (Figure 7). That spectral library is the same one utilized in the ImgTools software, defined by Souza et al. (2005). (a) (b) Figure 7. Spectral curves of the pure components utilized in the MME applied to Brazilian biomes; (a) for Landsat 5 and 7 in surface reflectance, (b) for Landsat 8 for reflectance at the top of the atmosphere. The MME assumes that the reflectance of each pixel, in each band of the Landsat image, results from the linear combination of the product of the percentage of pure component (i.e., proportion in the pixel) by its respective value in the band, based on the equation: for R b = Fi R i,b + ε b (1) ΣF i = 1 (2) 16
17 where R b is a reflectance in band b, R i,b is a reflectance of the pure component i, in band b, F i the fraction (or proportion) of the pure component I, whose sum is 1 (or 100%), and ε b is the residual error of each band. The MME is resolved to estimate the F i values (i.e., the fraction or proportion of the pure component), thus obtaining Fraction images for Vegetation, Non-Photosynthetically Active Vegetation, Soil and Cloud, which were utilized for calculating the NDFI spectral difference, obtained by the normalized different of the Vegetation fraction image and the sum of the Non-Photosynthetically Active Vegetation and Soil fraction images, according to the equation below: Where, and, The Fraction and NDFI images are utilized as entry parameters for classifying ground cover (see Stage 4 below, and legend) and to generate masks for clouds in order to generate the temporal mosaic (Stage 2). The Fraction images obtained with the MME are utilized to generate the ground cover map for classes of Level 1 of the Classification Legend, including: Forest, Non-Forest, Water and Cloud. Classification is conducted according to three steps (Figure 1). First, we apply the cloud mask described above, to map areas covered by clouds that could not be filtered during the process of generating the temporal mosaic for the chart. Next, we apply the mask for water based on the following criteria: pixels with Shadow Fraction greater than 75%, and Vegetation less than 10% and Soil less than 5%, and that are not in areas of shaded relief (we detected shaded relief using the Hill Shadow model and illumination data from each scene available in their metadata). Finally, we apply a decision tree (empirically defined) to map the forest and non-forest areas. The cloud and water masks are combined with the map resulting from the decision tree to generate the final map (Figure 8). The statistical distribution of the parameters is presented below in Figure 9. 17
18 Figure 8. Empirical decision tree applied to the NDFI, GV and GVs to map the classes of Forest and Non-Forest. Masks for Water and Cloud are applied to make up the final map. The variables (V1, V2, V3, V4) are adjusted for each chart. Figure 9. Frequency distribution of the values assumed by variables v1, v2, v3 and v4. Application of the decision tree generates the land cover map with the Forest, Non-Forest, Water and Cloud classes. The maps generated in the Classification and Land Cover Stage are subjected to processes of spatial and temporal filtering. The spatial filter segments and indexes the Forests, Non-Forest and Water classes in contiguous regions that are later on identified and reclassified based on the following criterion: areas smaller than or equal to half a hectare (i.e., approximately 5 pixels) are reclassified based on the majority of the neighboring classes. Thus, for example, a segment of the Non-Forest class with up to 5 pixels is identified, the neighboring pixels are identified, and the segment is reclassified with the predominant value of the neighboring class of pixels (Figure 7). That process is applied for all of the segments of the classes selected for filtering (i.e. Forests, Non-Forest and Water). The next step is to utilize temporal information to identify transitions of classes that are not allowed between consecutive years. For example, if a pixel is classified as Forest in a given 18
19 year t i (with i = 2008, 2009,..., 2015), and Non-Forest in year t i+1 and in t i+2, that pixel is reclassified as Forest in year t i. There are several rules for transitions listed in Table 3 that are applied by the temporal filter. Table 3. Rules for reclassification applied by the temporal filter. Rules Input (Year) Output General ti ti+1 ti+2 ti+5 ti ti+1 ti+2 ti+3 RG-1 NF F F F F F F F RG-2 NF F NF NF NF NF NF NF RG-3 NF SI * * NF NF * * RG-4 NF A NF NF NF NF NF NF RG-5 F SI * * F F * * RG-6 F A F F F F F F RG-7 F NF F F F F F F RG-8 A F F F F F F F RG-9 A NF NF NF NF NF NF NF RG-10 A * A A A A A A RG-11 A SI * * A A * * RG-12 * NF SI * * NF NF * 19
20 RG-13 * F SI * * F F * RG-14 * A SI * * A A * RG-15 * * NF SI * * NF NF RG-16 * * F SI * * F F RG-17 * * A SI * * A A Specific Rules RE-1 NF NF F NF NF NF NF NF RE-2 NF NF A NF NF NF NF NF RE-3 F F NF F F F F F RE-4 F F A F F F F F NF = Non-forest; F = Forest; A = Water; SI = No information or Cloud; * = Any class; i = {2008, 2009, 2010,..., 2015}. After application of the spatial and temporal filters, the resulting land cover maps were evaluated based on analyses of spatial concordance with maps [reference] of the biomes published by other initiatives (Table 1). The Prodes deforestation maps Prodes were selected for analysis of the spatial agreement. A 5km x 5km grid was created to compare the proportion of forest area classified by the MapBiomas maps and the reference maps. Those cells of the analysis grid are classified according to three levels of agreement: High (>80%), Medium (80-60%) and Low (>60). That procedure was applied to a set of 10 charts and the results point to a high correspondence between those results for the Forest class. Finally, the maps for areas of agricultural crops, pastures and planted forests, independently generated with automatic classifiers based on statistical decision trees, were integrated with the map of forests produced for Collection 1 of MapBiomas Amazonia. Those maps were combined with the land cover maps of the Amazon biome to generate the final consolidated map of Collection 1. 20
21 Data volumes and processing issues The volume of Landsat imagery data processed for Collection 1 of the Amazon Biome reached GigaBytes. These data generated several outputs that comprise XXX TeraBytes, and are available at MapBiomas platform. 3.3 Collection 2 - MapBiomas Amazonia Collection 2 covered the period from 2000 to 2016 and followed the image process protocol described for Collection 1. The differences in the image processing approach are highlighted in this section Algorithm The algorithm applied to Collection 2 followed the image processing steps applied to Collection 1. There are changes regarding the decision trees and new rules to temporal filter. A new empirical decision tree was built to deal with the new classification scheme of Collection 2. Figure 10. Empirical decision tree applied to Collection 2 for the amazon biome. 21
22 Two empirical decision trees were defined - one that does not include Non-Forest Natural Formation class (Figure 10A) and another that does (Figure 10B). The only structural difference in these two trees is in node 4-1 (NDFI>=5), which is split in the second decision tree into Agriculture or Pasture and Non-Forest Natural Formations classes (Figure 10B). A new set of temporal filters rules was developed to deal with the new land cover classes (Table 4). Table 4. Temporal filter rules applied to Collection 2 land cover classes. Rule Type Kernel Notes t-2 t-1 T t+1 t+2 Result RG01 RG 3 F-N-F null null 3 RG02 RG 3 F-AP-F null null 3 RG03 RG 3 F-FNNF-F null null 3 RG04 RG 3 AP-F-AP null null 15 RG05 RG 3 AP-N-AP null null 15 RG06 RG 3 AP-FNNF-AP null null 15 RG07 RG 3 AP-A-AP null null 15 RG08 RG 3 FS-N-FS null null 8 RG09 RG 3 FNNF-N-FNNF null null 10 RG10 RG 3 FNNF-A-FNNF null null 10 RG11 RG 3 ANV-F-ANV null null 22 RG12 RG 3 ANV-N-ANV null null 22 RG13 RG 3 A-N-A null null 26 22
23 RG14 RG 3 A-FNNF-A null null 26 RG15 RG 3 A-AP-A null null 26 RG16 RG 3 FS-FNNF-FS null null 8 RG17 RG 3 A-ANV-A null null 26 RG18 RG 3 ANV-FS-ANV null null 22 RG19 RG 3 FNNF-FS-FNNF null null 10 RG20 RG 3 FNNF-FD-FNNF null null 10 RP01 RP 3 N-A-A null null 26 RU01 RU 3 A-A-N null null Data availability and screening The same mapping unit (i.e., 1 degree by 1.5 degree map sheet area) was used for Collection 2. The main difference was in selection of the temporal window to generate temporal image mosaics. For Collection 2, we used data collected for each map sheet during 2000 through 2016 to define the period for selecting Landsat images to build the temporal mosaic. The optimal period for the Amazon biome was from June 1 to October 31. Therefore, the initial selection of a time period for building image mosaics was automatic. At that point, image analysts inspected the results and made changes to improve the mosaics (i.e., reduce cloud and/or increase surface observation areas within the mosaic) if necessary. The total amount of Landsat scenes processed per year to produce the image mosaics is presented in Table 5. Table 5. Number of images utilized per year, sensor and maximum percentage of cloud cover accepted for Collection 1. Year Number of Landsat images Amount size (GByte) Biome data cover (%)
24 , , , , , , Total Figure 11.Coverage of Landsat scenes to make up the temporal mosaic of the charts processed for the Amazon biome for the period of 2000 to 2016, with up to 30% of cloud cover Classification scheme and parameters The classification scheme for Collection 2 of MapBiomas Amazonia is presented in Table 6. These classes are a sub-set of the whole MapBiomas classification scheme and will be the main input for classification integration with other classes (which is discussed in the MapBiomas Integration ATBD). 24
25 Table 6. Classification scheme of Collection 2 for the Amazon Biome. Value Color Color code Class 0 #FFFFFF No Information 3 # Dense Forest 6 #76A5AF Inundated Forest 7 #29EEE4 Degraded Forest 8 #77A605 Secondary Forest 10 #FF9966 Nature Non-Forest Formations 14 #FFFFB2 Agriculture and Pasture 22 #EA9999 Non-Vegetated Areas 26 #0000FF Water Surface 27 #D5D5E5 Unobserved Data volumes and processing issues The volume of Landsat imagery data processed for Collection 2 of the Amazon Biome reached XXXX TeraBytes. These data generated several outputs that include: Collection 2 Asset for the Amazon Biome, composed of temporal mosaics of Landsat imagery, SMA, NDFI and derived spectral indices. These datasets together comprise XXX TeraBytes, and are available at the MapBiomas platform. 4. Validation Strategies The methodology for analyzing the accuracy of the annual land cover and use maps of the Amazon was applied with the following objectives: a) define the sample plan; and b) conduct an analysis of the accuracy of Collections 1 and 2 of the annual mapping (in other words, of all of the years of mapping those collections) of land cover in Brazil performed by the project: The sampling plan is the methodology for calculating the number of samples and for allocating the sampling points necessary for estimating the accuracy of mapping of Brazilian biomes. The analysis of accuracy is the estimate of the Fraction of the correctly classified number of pixels from mapping, their sampling errors and confidence intervals, besides the estimate of the real area occupied by each class of land cover. 25
26 For the Amazon biome a number of 938 samples per year was defined for the period of 2008 to 2015, which were interpreted by 3 evaluators (Figure 11). Figure 12. Spatial distribution of the sampling points for analyzing the accuracy of Collections 1 and 2 of MapBiomas Amazonia. Landsat images were utilized for visual interpretation of the sampling points (Figure 11). AccTools software was utilized to collect and visually interpret those points. A grid of 3x3 pixel points was employed, using the location of the sampling as the central pixel of that grid as a reference. Thus, the resulting sample is formed by the central pixel and its neighboring pixels, totaling 9 pixels, in other ways, an area of approximately one hectare (Figure 12). 26
27 Figure 13. Sampling scheme and interpretation grade of Landsat pixels utilized in visual interpretation for analysis of accuracy. The samples from each Collection were interpreted by three independent evaluators. In each pixel, interpretation is made of the fractions occupied by each class of land cover in the pixel. The fraction occupied by classes not identified by the evaluator is included in the other class. For the accuracy analysis, the dominant class visually interpreted with the Landsat images will be taken as the reference class of the pixel. The dominant class is the class with the largest proportion in the pixel, automatically calculated by the system. After data collection using AccTools, the data collected are exported to the Excel format, for analysis of map accuracy. Besides the proportions of the pixel occupied by each class of cover, the following information is extracted: biome, dominant class, page, pixel number in the block, sample number, geographic coordinate of the sample and observations. The Amazon presented an overall accuracy of 95% for Collection 1, considering all of the pixels analyzed. In evaluating the historical series of 2008 to 2015, the accuracy results presented values greater than 90%, reaching 96% in 2008 (Figure 13). The accuracy analysis in Collection 2 has not yet been concluded. 27
28 Figure 14. Preliminary result of the accuracy analysis of the Amazon biome for Collection Map Collections and Analysis The annual time-series classification dataset produced for MapBiomas Amazonia can be used in several types of spatial Analyses. For that purpose, only Collection 2 will be used, since it represents a more robust classification dataset as compared to the previous one. An example of the classification panel for Collection 2 is presented in Figure 14. Applications of these datasets have not yet been conducted. A potential list of such applications includes: 1. Quantification of gross annual deforestation. 2. Quantification of annual forest degradation. 3. Annual quantification of secondary growth forests. 4. Quantification of annual forest fragmentation. 5. Quantification of annual net forest balance. 6. Land cover and land use transitions. 7. Spatial analyses of agreement and disagreement among existing forest and land cover maps. 8. Quantification of the natural dynamics of water surfaces. Figure 15. Data classification panel for Collection 2 MapBiomas Amazonia (A), and a more detailed view of the classification map for 2008 (B). 28
29 6. Practical Considerations This is a live document and we expect to update it as soon as Collection 2 becomes available. New information about the results will be included. 7. Concluding Remarks and Perspectves The proposal algorithms for pre-processing and classifying Landsat imagery hold promise for revolutionizing the production of land cover and land use maps at a large scale. Thanks to Google Earth Engine and open source technology it is possible to access and process large scale datasets of satellite imagery such as the one presented in this document. The replication of this type of project is viable for other areas of the planet. The next step of this project is to expand mapping and monitoring of the Pan-Amazon and other tropical forest regions. This will involve producing Collection 3, which will include a large time-series (1985 to the present). Future developments include using the entire spectral-temporal information of Landsat data and integration with other sensors such as Sentinel-2 and AWiFs-Resourcesat. 29
30 8. References Almeida et al High spatial resolution land use and land cover mapping of the Brazilian Legal Amazon in 2008 using Landsat-5/TM and Modis data. Acta Amazônica, 46(3), Carlotto, M. J. Reducing the effects of space-varying, wavelength dependent scattering in multispectral imagery. International Journal of Remote Sensing, v. 20, n. 17, p , Fearnside, P.M Desmatamento na Amazônia: dinâmica, impactos e controle. Acta Amazônica, 36, INPE Instituto Nacional de Pesquisas Espaciais, Projeto TerraClass Dinâmica do uso and cobertura da terra para o período de 10 anos nas áreas desflorestadas da Amazônia Legal Brasileira. ( Acesso in 03/13/2017. MMA Ministério do Meio Ambiente, Biomas. ( Access on 03/13/2017. Souza Jr., C., Roberts, D. A. and Cochrane, M. A. Combining spectral and spatial information to map canopy damage from selective logging and forest fires. Remote Sensing of Environment, v. 98, n 2-3, p a. Souza Jr., C. and Siqueira, J. V. ImgTools: a software for optical remotely sensed data analysis. In: XVI Simpósio Brasileiro de Sensoriamento Remoto (SBSR). Foz do Iguaçu-PR. 8p,
31 9. Appendices 9.1. Supplementary Material 9.2. Algorithm Code 31
Atlantic Forest - Appendix
Atlantic Forest - Appendix Collection 3 Version 1 General coordinator Marcos Reis Rosa Team Fernando Frizeira Paternost Jacqueline Freitas Viviane Cristina Mazin Eduardo Reis Rosa 1 Landsat image mosaics
More informationCaatinga - Appendix. Collection 3. Version 1. General coordinator Washington J. S. Franca Rocha (UEFS)
Caatinga - Appendix Collection 3 Version 1 General coordinator Washington J. S. Franca Rocha (UEFS) Team Diego Pereira Costa (UEFS/GEODATIN) Frans Pareyn (APNE) José Luiz Vieira (APNE) Rodrigo N. Vasconcelos
More informationImage interpretation and analysis
Image interpretation and analysis Grundlagen Fernerkundung, Geo 123.1, FS 2014 Lecture 7a Rogier de Jong Michael Schaepman Why are snow, foam, and clouds white? Why are snow, foam, and clouds white? Today
More informationNORMALIZING ASTER DATA USING MODIS PRODUCTS FOR LAND COVER CLASSIFICATION
NORMALIZING ASTER DATA USING MODIS PRODUCTS FOR LAND COVER CLASSIFICATION F. Gao a, b, *, J. G. Masek a a Biospheric Sciences Branch, NASA Goddard Space Flight Center, Greenbelt, MD 20771, USA b Earth
More informationEVALUATION OF THE EXTENSION AND DEGRADATION OF MANGROVE AREAS IN SERGIPE STATE WITH REMOTE SENSING DATA
EVALUATION OF THE EXTENSION AND DEGRADATION OF MANGROVE ABSTRACT AREAS IN SERGIPE STATE WITH REMOTE SENSING DATA Myrian M. Abdon Ernesto G.M.Vieira Carmem R.S. Espindola Alberto W. Setzer Instituto de
More informationAPCAS/10/21 April 2010 ASIA AND PACIFIC COMMISSION ON AGRICULTURAL STATISTICS TWENTY-THIRD SESSION. Siem Reap, Cambodia, April 2010
APCAS/10/21 April 2010 Agenda Item 8 ASIA AND PACIFIC COMMISSION ON AGRICULTURAL STATISTICS TWENTY-THIRD SESSION Siem Reap, Cambodia, 26-30 April 2010 The Use of Remote Sensing for Area Estimation by Robert
More informationUniversity of Texas at San Antonio EES 5053 Term Project CORRELATION BETWEEN NDVI AND SURFACE TEMPERATURES USING LANDSAT ETM + IMAGERY NEWFEL MAZARI
University of Texas at San Antonio EES 5053 Term Project CORRELATION BETWEEN NDVI AND SURFACE TEMPERATURES USING LANDSAT ETM + IMAGERY NEWFEL MAZARI Introduction and Objectives The present study is a correlation
More informationAlberto Setzer 1 Demerval Aparecido Gonçalves 2 Fabiano Morelli 1.
Validation of fire pixels detected by satellites with small format aerial photos (Validação de focos de queima detectados por satélites com fotos aéreas de pequeno formato) Alberto Setzer 1 Demerval Aparecido
More informationRemote Sensing. The following figure is grey scale display of SPOT Panchromatic without stretching.
Remote Sensing Objectives This unit will briefly explain display of remote sensing image, geometric correction, spatial enhancement, spectral enhancement and classification of remote sensing image. At
More information2017 REMOTE SENSING EVENT TRAINING STRATEGIES 2016 SCIENCE OLYMPIAD COACHING ACADEMY CENTERVILLE, OH
2017 REMOTE SENSING EVENT TRAINING STRATEGIES 2016 SCIENCE OLYMPIAD COACHING ACADEMY CENTERVILLE, OH This presentation was prepared using draft rules. There may be some changes in the final copy of the
More informationLANDSAT-TM DATA TO MAP FLOODED AREAS
LANDSAT-TM DATA TO MAP FLOODED AREAS Sergio dos Anjos Ferreira Pinto Teresa Gallotti Florenzano Instituto de Pesquisas Espaciais-INPE Caixa Postal 515-12201 Sao Jose dos Campos-SP - Brazil Comission Number
More informationREMOTE SENSING. Topic 10 Fundamentals of Digital Multispectral Remote Sensing MULTISPECTRAL SCANNERS MULTISPECTRAL SCANNERS
REMOTE SENSING Topic 10 Fundamentals of Digital Multispectral Remote Sensing Chapter 5: Lillesand and Keifer Chapter 6: Avery and Berlin MULTISPECTRAL SCANNERS Record EMR in a number of discrete portions
More informationSatellite Remote Sensing: Earth System Observations
Satellite Remote Sensing: Earth System Observations Land surface Water Atmosphere Climate Ecosystems 1 EOS (Earth Observing System) Develop an understanding of the total Earth system, and the effects of
More informationA SYNERGETIC USE OF REMOTE-SENSED DATA TO ASSESS THE EVOLUTION OF BURNT AREA BY WILDFIRES IN PORTUGAL
A SYNERGETIC USE OF REMOTE-SENSED DATA TO ASSESS THE EVOLUTION OF BURNT AREA BY WILDFIRES IN PORTUGAL Teresa J. Calado and Carlos C. DaCamara CGUL, Faculty of Sciences, University of Lisbon, Campo Grande,
More informationIKONOS High Resolution Multispectral Scanner Sensor Characteristics
High Spatial Resolution and Hyperspectral Scanners IKONOS High Resolution Multispectral Scanner Sensor Characteristics Launch Date View Angle Orbit 24 September 1999 Vandenberg Air Force Base, California,
More informationPreparing Remote Sensing Data for Natural Resources Mapping (image enhancement, rectifications )
Preparing Remote Sensing Data for Natural Resources Mapping (image enhancement, rectifications ) Why is this important What are the major approaches Examples of digital image enhancement Follow up exercises
More informationLand Cover Analysis to Determine Areas of Clear-cut and Forest Cover in Olney, Montana. Geob 373 Remote Sensing. Dr Andreas Varhola, Kathry De Rego
1 Land Cover Analysis to Determine Areas of Clear-cut and Forest Cover in Olney, Montana Geob 373 Remote Sensing Dr Andreas Varhola, Kathry De Rego Zhu an Lim (14292149) L2B 17 Apr 2016 2 Abstract Montana
More informationCanImage. (Landsat 7 Orthoimages at the 1: Scale) Standards and Specifications Edition 1.0
CanImage (Landsat 7 Orthoimages at the 1:50 000 Scale) Standards and Specifications Edition 1.0 Centre for Topographic Information Customer Support Group 2144 King Street West, Suite 010 Sherbrooke, QC
More informationSatellite data processing and analysis: Examples and practical considerations
Satellite data processing and analysis: Examples and practical considerations Dániel Kristóf Ottó Petrik, Róbert Pataki, András Kolesár International LCLUC Regional Science Meeting in Central Europe Sopron,
More informationImprovements in Landsat Pathfinder Methods for Monitoring Tropical Deforestation and Their Extension to Extra-tropical Areas
Improvements in Landsat Pathfinder Methods for Monitoring Tropical Deforestation and Their Extension to Extra-tropical Areas PI: John R. G. Townshend Department of Geography (and Institute for Advanced
More informationF2 - Fire 2 module: Remote Sensing Data Classification
F2 - Fire 2 module: Remote Sensing Data Classification F2.1 Task_1: Supervised and Unsupervised classification examples of a Landsat 5 TM image from the Center of Portugal, year 2005 F2.1 Task_2: Burnt
More informationSommersemester Prof. Dr. Christoph Kleinn Institut für Waldinventur und Waldwachstum Arbeitsbereich Fernerkundung und Waldinventur.
Basics of Remote Sensing Some literature references Franklin, SE 2001 Remote Sensing for Sustainable Forest Management Lewis Publishers 407p Lillesand, Kiefer 2000 Remote Sensing and Image Interpretation
More informationIntroduction to Remote Sensing
Introduction to Remote Sensing Spatial, spectral, temporal resolutions Image display alternatives Vegetation Indices Image classifications Image change detections Accuracy assessment Satellites & Air-Photos
More informationM. Ellen Dean and Roger M. Hoffer Department of Forestry and Natural Resources. Purdue University, West Lafayette, Indiana
Evaluation of Thematic Mapper Data and Computer-aided Analysis Techniques for Mapping Forest Cover M. Ellen Dean and Roger M. Hoffer Department of Forestry and Natural Resources Laboratory for Applications
More informationWGISS-42 USGS Agency Report
WGISS-42 USGS Agency Report U.S. Department of the Interior U.S. Geological Survey Kristi Kline USGS EROS Center Major Activities Landsat Archive/Distribution Changes Land Change Monitoring, Assessment,
More informationCLASSIFICATION OF VEGETATION AREA FROM SATELLITE IMAGES USING IMAGE PROCESSING TECHNIQUES ABSTRACT
CLASSIFICATION OF VEGETATION AREA FROM SATELLITE IMAGES USING IMAGE PROCESSING TECHNIQUES Arpita Pandya Research Scholar, Computer Science, Rai University, Ahmedabad Dr. Priya R. Swaminarayan Professor
More informationRemote Sensing for Rangeland Applications
Remote Sensing for Rangeland Applications Jay Angerer Ecological Training June 16, 2012 Remote Sensing The term "remote sensing," first used in the United States in the 1950s by Ms. Evelyn Pruitt of the
More informationLecture 13: Remotely Sensed Geospatial Data
Lecture 13: Remotely Sensed Geospatial Data A. The Electromagnetic Spectrum: The electromagnetic spectrum (Figure 1) indicates the different forms of radiation (or simply stated light) emitted by nature.
More informationTEMPORAL ANALYSIS OF MULTI EPOCH LANDSAT GEOCOVER IMAGES IN ZONGULDAK TESTFIELD
TEMPORAL ANALYSIS OF MULTI EPOCH LANDSAT GEOCOVER IMAGES IN ZONGULDAK TESTFIELD Şahin, H. a*, Oruç, M. a, Büyüksalih, G. a a Zonguldak Karaelmas University, Zonguldak, Turkey - (sahin@karaelmas.edu.tr,
More informationAdvanced satellite image fusion techniques for estimating high resolution Land Surface Temperature time series
COMECAP 2014 e-book of proceedings vol. 2 Page 267 Advanced satellite image fusion techniques for estimating high resolution Land Surface Temperature time series Mitraka Z., Chrysoulakis N. Land Surface
More informationSpatial Analyst is an extension in ArcGIS specially designed for working with raster data.
Spatial Analyst is an extension in ArcGIS specially designed for working with raster data. 1 Do you remember the difference between vector and raster data in GIS? 2 In Lesson 2 you learned about the difference
More informationOn the use of synthetic images for change detection accuracy assessment
On the use of synthetic images for change detection accuracy assessment Hélio Radke Bittencourt 1, Daniel Capella Zanotta 2 and Thiago Bazzan 3 1 Departamento de Estatística, Pontifícia Universidade Católica
More informationUrban Classification of Metro Manila for Seismic Risk Assessment using Satellite Images
Urban Classification of Metro Manila for Seismic Risk Assessment using Satellite Images Fumio YAMAZAKI/ yamazaki@edm.bosai.go.jp Hajime MITOMI/ mitomi@edm.bosai.go.jp Yalkun YUSUF/ yalkun@edm.bosai.go.jp
More informationMonitoring land cover in Acre State, western Brazilian Amazonia, using multitemporal remote sensing data
Monitoring land cover in Acre State, western Brazilian Amazonia, using multitemporal remote sensing data Yosio E. Shimabukuro Valdete Duarte Egidio Arai Ramon M. Freitas Paulo R. Martini André Lima Instituto
More informationRemote sensing in archaeology from optical to lidar. Krištof Oštir ModeLTER Scientific Research Centre of the Slovenian Academy of Sciences and Arts
Remote sensing in archaeology from optical to lidar Krištof Oštir ModeLTER Scientific Research Centre of the Slovenian Academy of Sciences and Arts Introduction Optical remote sensing Systems Search for
More informationApplication of Linear Spectral unmixing to Enrique reef for classification
Application of Linear Spectral unmixing to Enrique reef for classification Carmen C. Zayas-Santiago University of Puerto Rico Mayaguez Marine Sciences Department Stefani 224 Mayaguez, PR 00681 c_castula@hotmail.com
More informationUSGS Welcome. 38 th CEOS Working Group on Calibration and Validation Plenary (WGCV-38)
Landsat 5 USGS Welcome Prepared for 38 th CEOS Working Group on Calibration and Validation Plenary (WGCV-38) Presenter Tom Cecere International Liaison USGS Land Remote Sensing Program Elephant Butte Reservoir
More informationField size estimation, past and future opportunities
Field size estimation, past and future opportunities Lin Yan & David Roy Geospatial Sciences Center of Excellence South Dakota State University February 13-15 th 2018 Advances in Emerging Technologies
More informationKeywords: Agriculture, Olive Trees, Supervised Classification, Landsat TM, QuickBird, Remote Sensing.
Classification of agricultural fields by using Landsat TM and QuickBird sensors. The case study of olive trees in Lesvos island. Christos Vasilakos, University of the Aegean, Department of Environmental
More informationSan Diego State University Department of Geography, San Diego, CA. USA b. University of California, Department of Geography, Santa Barbara, CA.
1 Plurimondi, VII, No 14: 1-9 Land Cover/Land Use Change analysis using multispatial resolution data and object-based image analysis Sory Toure a Douglas Stow a Lloyd Coulter a Avery Sandborn c David Lopez-Carr
More informationBrazilian Amazon Fire Frequency Data in Raster Format. Summary:
Brazilian Amazon Fire Frequency Data in Raster Format Summary: This dataset contains fire frequency data for the subregion of the Brazilian Amazon. These data were converted to flat raster binary image
More informationBlacksburg, VA July 24 th 30 th, 2010 Remote Sensing Page 1. A condensed overview. For our purposes
A condensed overview George McLeod Prepared by: With support from: NSF DUE-0903270 in partnership with: Geospatial Technician Education Through Virginia s Community Colleges (GTEVCC) The art and science
More informationThe techniques with ERDAS IMAGINE include:
The techniques with ERDAS IMAGINE include: 1. Data correction - radiometric and geometric correction 2. Radiometric enhancement - enhancing images based on the values of individual pixels 3. Spatial enhancement
More informationHow to Access Imagery and Carry Out Remote Sensing Analysis Using Landsat Data in a Browser
How to Access Imagery and Carry Out Remote Sensing Analysis Using Landsat Data in a Browser Including Introduction to Remote Sensing Concepts Based on: igett Remote Sensing Concept Modules and GeoTech
More informationData Requirements Definition and Data Services Options for RAPP
Data Requirements Definition and Data Services Options for RAPP Brian Killough CEOS Systems Engineering Office (SEO) 5 th GEOGLAM RAPP Workshop Frascati, Italy May 16-17, 2017 Requirements Update The observation
More informationAssessment of Spatiotemporal Changes in Vegetation Cover using NDVI in The Dangs District, Gujarat
Assessment of Spatiotemporal Changes in Vegetation Cover using NDVI in The Dangs District, Gujarat Using SAGA GIS and Quantum GIS Tutorial ID: IGET_CT_003 This tutorial has been developed by BVIEER as
More informationIn late April of 1986 a nuclear accident damaged a reactor at the Chernobyl nuclear
CHERNOBYL NUCLEAR POWER PLANT ACCIDENT Long Term Effects on Land Use Patterns Project Introduction: In late April of 1986 a nuclear accident damaged a reactor at the Chernobyl nuclear power plant in Ukraine.
More informationEVALUATION OF MEDIUM-RESOLUTION SATELLITE IMAGES FOR LAND USE MONITORING USING SPECTRAL MIXTURE ANALYSIS
EVALUATION OF MEDIUM-RESOLUTION SATELLITE IMAGES FOR LAND USE MONITORING USING SPECTRAL MIXTURE ANALYSIS Florian P. Kressler Austrian Research Centers, Seibersdorf, Austria florian.kressler@arcs.ac.at
More informationAUTOMATIC DETECTION OF HEDGES AND ORCHARDS USING VERY HIGH SPATIAL RESOLUTION IMAGERY
AUTOMATIC DETECTION OF HEDGES AND ORCHARDS USING VERY HIGH SPATIAL RESOLUTION IMAGERY Selim Aksoy Department of Computer Engineering, Bilkent University, Bilkent, 06800, Ankara, Turkey saksoy@cs.bilkent.edu.tr
More informationRemote Sensing Phenology. Bradley Reed Principal Scientist USGS National Center for Earth Resources Observation and Science Sioux Falls, SD
Remote Sensing Phenology Bradley Reed Principal Scientist USGS National Center for Earth Resources Observation and Science Sioux Falls, SD Remote Sensing Phenology Potential to provide wall-to-wall phenology
More informationPresent and future of marine production in Boka Kotorska
Present and future of marine production in Boka Kotorska First results from satellite remote sensing for the breeding areas of filter feeders in the Bay of Kotor INTRODUCTION Environmental monitoring is
More informationActivity Data (AD) Monitoring in the frame of REDD+ MRV
Activity Data (AD) Monitoring in the frame of REDD+ MRV Preliminary comments REDD+ is sustainable low emissions, high carbon rural development Monitoring efforts should support this effort Challenges Diversity
More informationLandsat 8 and Sentinel 2 higher order products: input to S2DUP. Chris Justice (UMD) Curtis Woodcock (BU), Martin Claverie (UMD/GSFC)
Landsat 8 and Sentinel 2 higher order products: input to S2DUP Chris Justice (UMD) Curtis Woodcock (BU), Martin Claverie (UMD/GSFC) MODIS Land Products Energy Balance Product Suite Surface Reflectance
More informationVisualizing a Pixel. Simulate a Sensor s View from Space. In this activity, you will:
Simulate a Sensor s View from Space In this activity, you will: Measure and mark pixel boundaries Learn about spatial resolution, pixels, and satellite imagery Classify land cover types Gain exposure to
More informationLand cover change methods. Ned Horning
Land cover change methods Ned Horning Version: 1.0 Creation Date: 2004-01-01 Revision Date: 2004-01-01 License: This document is licensed under a Creative Commons Attribution-Share Alike 3.0 Unported License.
More informationNON-PHOTOGRAPHIC SYSTEMS: Multispectral Scanners Medium and coarse resolution sensor comparisons: Landsat, SPOT, AVHRR and MODIS
NON-PHOTOGRAPHIC SYSTEMS: Multispectral Scanners Medium and coarse resolution sensor comparisons: Landsat, SPOT, AVHRR and MODIS CLASSIFICATION OF NONPHOTOGRAPHIC REMOTE SENSORS PASSIVE ACTIVE DIGITAL
More informationInter-Calibration of the RapidEye Sensors with Landsat 8, Sentinel and SPOT
Inter-Calibration of the RapidEye Sensors with Landsat 8, Sentinel and SPOT Dr. Andreas Brunn, Dr. Horst Weichelt, Dr. Rene Griesbach, Dr. Pablo Rosso Content About Planet Project Context (Purpose and
More informationHyperspectral image processing and analysis
Hyperspectral image processing and analysis Lecture 12 www.utsa.edu/lrsg/teaching/ees5083/l12-hyper.ppt Multi- vs. Hyper- Hyper-: Narrow bands ( 20 nm in resolution or FWHM) and continuous measurements.
More informationEnhancement of Multispectral Images and Vegetation Indices
Enhancement of Multispectral Images and Vegetation Indices ERDAS Imagine 2016 Description: We will use ERDAS Imagine with multispectral images to learn how an image can be enhanced for better interpretation.
More informationTexture characterization in DIRSIG
Rochester Institute of Technology RIT Scholar Works Theses Thesis/Dissertation Collections 2001 Texture characterization in DIRSIG Christy Burtner Follow this and additional works at: http://scholarworks.rit.edu/theses
More informationRemote Sensing Part 3 Examples & Applications
Remote Sensing Part 3 Examples & Applications Review: Spectral Signatures Review: Spectral Resolution Review: Computer Display of Remote Sensing Images Individual bands of satellite data are mapped to
More informationInt n r t o r d o u d c u ti t on o n to t o Remote Sensing
Introduction to Remote Sensing Definition of Remote Sensing Remote sensing refers to the activities of recording/observing/perceiving(sensing)objects or events at far away (remote) places. In remote sensing,
More informationAn Introduction to Remote Sensing & GIS. Introduction
An Introduction to Remote Sensing & GIS Introduction Remote sensing is the measurement of object properties on Earth s surface using data acquired from aircraft and satellites. It attempts to measure something
More informationThe Utility and Limitations of Remote Sensing in Land Use Change Detection and Conservation Planning
The Utility and Limitations of Remote Sensing in Land Use Change Detection and Conservation Planning Steffen Mueller, PhD, Principal Economist Ken Copenhaver, CropGrower LLC Presentation to: US Environmental
More informationAn NDVI image provides critical crop information that is not visible in an RGB or NIR image of the same scene. For example, plants may appear green
Normalized Difference Vegetation Index (NDVI) Spectral Band calculation that uses the visible (RGB) and near-infrared (NIR) bands of the electromagnetic spectrum NDVI= + An NDVI image provides critical
More informationDISTINGUISHING URBAN BUILT-UP AND BARE SOIL FEATURES FROM LANDSAT 8 OLI IMAGERY USING DIFFERENT DEVELOPED BAND INDICES
DISTINGUISHING URBAN BUILT-UP AND BARE SOIL FEATURES FROM LANDSAT 8 OLI IMAGERY USING DIFFERENT DEVELOPED BAND INDICES Mark Daryl C. Janiola (1), Jigg L. Pelayo (1), John Louis J. Gacad (1) (1) Central
More informationIntroduction of Satellite Remote Sensing
Introduction of Satellite Remote Sensing Spatial Resolution (Pixel size) Spectral Resolution (Bands) Resolutions of Remote Sensing 1. Spatial (what area and how detailed) 2. Spectral (what colors bands)
More informationIntroduction to Remote Sensing
Introduction to Remote Sensing Outline Remote Sensing Defined Resolution Electromagnetic Energy (EMR) Types Interpretation Applications Remote Sensing Defined Remote Sensing is: The art and science of
More informationCostal region of northern Peru, the pacific equatorial dry forest there is recognised for its unique endemic biodiversity
S.Baena@kew.org http://www.kew.org/gis/ Costal region of northern Peru, the pacific equatorial dry forest there is recognised for its unique endemic biodiversity Highly threatened ecosystem affected by
More informationFundamentals of Remote Sensing
Climate Variability, Hydrology, and Flooding Fundamentals of Remote Sensing May 19-22, 2015 GEO-Latin American & Caribbean Water Cycle Capacity Building Workshop Cartagena, Colombia 1 Objective To provide
More informationAtmospheric Correction (including ATCOR)
Technical Specifications Atmospheric Correction (including ATCOR) The data obtained by optical satellite sensors with high spatial resolution has become an invaluable tool for many groups interested in
More informationLesson 9: Multitemporal Analysis
Lesson 9: Multitemporal Analysis Lesson Description Multitemporal change analyses require the identification of features and measurement of their change through time. In this lesson, we will examine vegetation
More informationSUGAR_GIS. From a user perspective. Provides spatial distribution of a wide range of sugarcane production data in an easy to use and sensitive way.
SUGAR_GIS From a user perspective What is Sugar_GIS? A web-based, decision support tool. Provides spatial distribution of a wide range of sugarcane production data in an easy to use and sensitive way.
More informationLand Remote Sensing Lab 4: Classication and Change Detection Assigned: October 15, 2017 Due: October 27, Classication
Name: Land Remote Sensing Lab 4: Classication and Change Detection Assigned: October 15, 2017 Due: October 27, 2017 In this lab, you will generate several gures. Please sensibly name these images, save
More informationCoral Reef Remote Sensing
Coral Reef Remote Sensing Spectral, Spatial, Temporal Scaling Phillip Dustan Sensor Spatial Resolutio n Number of Bands Useful Bands coverage cycle Operation Landsat 80m 2 2 18 1972-97 Thematic 30m 7
More informationUSING LANDSAT MULTISPECTRAL IMAGES IN ANALYSING FOREST VEGETATION
Technical Sciences 243 USING LANDSAT MULTISPECTRAL IMAGES IN ANALYSING FOREST VEGETATION Teodor TODERA teotoderas@yahoo.com Traian CR CEA traiancracea@yahoo.com Alina NEGOESCU alina.negoescu@yahoo.com
More informationThe availability of cloud free Landsat TM and ETM+ land observations and implications for global Landsat data production
14475 The availability of cloud free Landsat TM and ETM+ land observations and implications for global Landsat data production *V. Kovalskyy, D. Roy (South Dakota State University) SUMMARY The NASA funded
More informationModule 3 Introduction to GIS. Lecture 8 GIS data acquisition
Module 3 Introduction to GIS Lecture 8 GIS data acquisition GIS workflow Data acquisition (geospatial data input) GPS Remote sensing (satellites, UAV s) LiDAR Digitized maps Attribute Data Management Data
More informationThe studies began when the Tiros satellites (1960) provided man s first synoptic view of the Earth s weather systems.
Remote sensing of the Earth from orbital altitudes was recognized in the mid-1960 s as a potential technique for obtaining information important for the effective use and conservation of natural resources.
More informationTHE THIRD GENERATION RELATIVE DETECTION EFFICIENCY MODEL FOR THE BRAZILIAN LIGHTNING DETECTION NETWORK (BRASILDAT)
THE THIRD GENERATION RELATIVE DETECTION EFFICIENCY MODEL FOR THE BRAZILIAN LIGHTNING DETECTION NETWORK (BRASILDAT) K. P. Naccarato; O. Pinto Jr. Instituto Nacional de Pesquisas Espaciais (INPE) Sao Jose
More informationIntroduction to Radar
National Aeronautics and Space Administration ARSET Applied Remote Sensing Training http://arset.gsfc.nasa.gov @NASAARSET Introduction to Radar Jul. 16, 2016 www.nasa.gov Objective The objective of this
More informationGE 113 REMOTE SENSING
GE 113 REMOTE SENSING Topic 8. Image Classification and Accuracy Assessment Lecturer: Engr. Jojene R. Santillan jrsantillan@carsu.edu.ph Division of Geodetic Engineering College of Engineering and Information
More informationGIS Data Collection. Remote Sensing
GIS Data Collection Remote Sensing Data Collection Remote sensing Introduction Concepts Spectral signatures Resolutions: spectral, spatial, temporal Digital image processing (classification) Other systems
More informationFig Color spectrum seen by passing white light through a prism.
1. Explain about color fundamentals. Color of an object is determined by the nature of the light reflected from it. When a beam of sunlight passes through a glass prism, the emerging beam of light is not
More informationIntroduction to Remote Sensing
Introduction to Remote Sensing Daniel McInerney Urban Institute Ireland, University College Dublin, Richview Campus, Clonskeagh Drive, Dublin 14. 16th June 2009 Presentation Outline 1 2 Spaceborne Sensors
More informationEvaluation of FLAASH atmospheric correction. Note. Note no SAMBA/10/12. Authors. Øystein Rudjord and Øivind Due Trier
Evaluation of FLAASH atmospheric correction Note Note no Authors SAMBA/10/12 Øystein Rudjord and Øivind Due Trier Date 16 February 2012 Norsk Regnesentral Norsk Regnesentral (Norwegian Computing Center,
More informationInterpreting land surface features. SWAC module 3
Interpreting land surface features SWAC module 3 Interpreting land surface features SWAC module 3 Different kinds of image Panchromatic image True-color image False-color image EMR : NASA Echo the bat
More informationIntroduction to Remote Sensing Fundamentals of Satellite Remote Sensing. Mads Olander Rasmussen
Introduction to Remote Sensing Fundamentals of Satellite Remote Sensing Mads Olander Rasmussen (mora@dhi-gras.com) 01. Introduction to Remote Sensing DHI What is remote sensing? the art, science, and technology
More informationBasic Hyperspectral Analysis Tutorial
Basic Hyperspectral Analysis Tutorial This tutorial introduces you to visualization and interactive analysis tools for working with hyperspectral data. In this tutorial, you will: Analyze spectral profiles
More informationTextural analysis of coca plantations using 1-meter-resolution remotely-sensed data
UNODC Workshop, 25-28 November, Bogota, Colombia 1 Textural analysis of coca plantations using 1-meter-resolution remotely-sensed data Workshop on Measurement of Cultivation and Production of Coca Leaves
More informationGeo/SAT 2 INTRODUCTION TO REMOTE SENSING
Geo/SAT 2 INTRODUCTION TO REMOTE SENSING Paul R. Baumann, Professor Emeritus State University of New York College at Oneonta Oneonta, New York 13820 USA COPYRIGHT 2008 Paul R. Baumann Introduction Remote
More informationNASA Missions and Products: Update. Garik Gutman, LCLUC Program Manager NASA Headquarters Washington, DC
NASA Missions and Products: Update Garik Gutman, LCLUC Program Manager NASA Headquarters Washington, DC 1 JPSS-2 (NOAA) SLI-TBD Formulation in 2015 RBI OMPS-Limb [[TSIS-2]] [[TCTE]] Land Monitoring at
More informationHigh Resolution Multi-spectral Imagery
High Resolution Multi-spectral Imagery Jim Baily, AirAgronomics AIRAGRONOMICS Having been involved in broadacre agriculture until 2000 I perceived a need for a high resolution remote sensing service to
More informationHYPERSPECTRAL IMAGERY FOR SAFEGUARDS APPLICATIONS. International Atomic Energy Agency, Vienna, Austria
HYPERSPECTRAL IMAGERY FOR SAFEGUARDS APPLICATIONS G. A. Borstad 1, Leslie N. Brown 1, Q.S. Bob Truong 2, R. Kelley, 3 G. Healey, 3 J.-P. Paquette, 3 K. Staenz 4, and R. Neville 4 1 Borstad Associates Ltd.,
More informationRemote Sensing. in Agriculture. Dr. Baqer Ramadhan CRP 514 Geographic Information System. Adel M. Al-Rebh G Term Paper.
Remote Sensing in Agriculture Term Paper to Dr. Baqer Ramadhan CRP 514 Geographic Information System By Adel M. Al-Rebh G199325390 May 2012 Table of Contents 1.0 Introduction... 4 2.0 Objective... 4 3.0
More informationCOMPARISON ON URBAN CLASSIFICATIONS USING LANDSAT-TM AND LINEAR SPECTRAL MIXTURE ANALYSIS EXTRACTED IMAGES: NAKHON RATCHASIMA MUNICIPAL AREA, THAILAND
Suranaree J. Sci. Technol. Vol. 17 No. 4; Oct - Dec 2010 401 COMPARISON ON URBAN CLASSIFICATIONS USING LANDSAT-TM AND LINEAR SPECTRAL MIXTURE ANALYSIS EXTRACTED IMAGES: NAKHON RATCHASIMA MUNICIPAL AREA,
More informationTextbook, Chapter 15 Textbook, Chapter 10 (only 10.6)
AGOG 484/584/ APLN 551 Fall 2018 Concept definition Applications Instruments and platforms Techniques to process hyperspectral data A problem of mixed pixels and spectral unmixing Reading Textbook, Chapter
More informationSatellite Imagery and Remote Sensing. DeeDee Whitaker SW Guilford High EES & Chemistry
Satellite Imagery and Remote Sensing DeeDee Whitaker SW Guilford High EES & Chemistry whitakd@gcsnc.com Outline What is remote sensing? How does remote sensing work? What role does the electromagnetic
More informationRADAR (RAdio Detection And Ranging)
RADAR (RAdio Detection And Ranging) CLASSIFICATION OF NONPHOTOGRAPHIC REMOTE SENSORS PASSIVE ACTIVE DIGITAL CAMERA THERMAL (e.g. TIMS) VIDEO CAMERA MULTI- SPECTRAL SCANNERS VISIBLE & NIR MICROWAVE Real
More informationFires, Flares and Lights: Mapping Anthropogenic Emission Sources with Nighttime Low light Imaging Satellite Data
Fires, Flares and Lights: Mapping Anthropogenic Emission Sources with Nighttime Low light Imaging Satellite Data Christopher D. Elvidge, Ph.D. Earth Observation Group NOAA National Geophysical Data Center
More information