Anais XVI Simpósio Brasileiro de Sensoriamento Remoto - SBSR, Foz do Iguaçu, PR, Brasil, 13 a 18 de abril de 2013, INPE



Similar documents
Analyzing temporal and spatial dynamics of deforestation in the Amazon: a case study in the Calha Norte region, State of Pará, Brazil

Land Use/Land Cover Map of the Central Facility of ARM in the Southern Great Plains Site Using DOE s Multi-Spectral Thermal Imager Satellite Images

A KNOWLEDGE-BASED APPROACH FOR REDUCING CLOUD AND SHADOW ABSTRACT

Changes in forest cover applying object-oriented classification and GIS in Amapa- French Guyana border, Amapa State Forest, Module 4

Texas Prairie Wetlands Project (TPWP) Performance Monitoring

TerraColor White Paper

Data Processing Flow Chart

Digital Remote Sensing Data Processing Digital Remote Sensing Data Processing and Analysis: An Introduction and Analysis: An Introduction

Remote Sensing in Natural Resources Mapping

Introduction to Imagery and Raster Data in ArcGIS

Open icon. The Select Layer To Add dialog opens. Click here to display

Policy Opportunities for Human Dimensions Science: The Challenge of Frontier Governance

INPE s Brazilian Amazon Deforestation and Forest Degradation Program. Dalton M. Valeriano (dalton@dsr.inpe.br) Program Manager

Remote Sensing Method in Implementing REDD+

Digital image processing

Landsat Monitoring our Earth s Condition for over 40 years

CIESIN Columbia University

Short technical report. Understanding the maps of risk assessment of deforestation and carbon dioxide emissions using scenarios for 2020

Using Remote Sensing Imagery to Evaluate Post-Wildfire Damage in Southern California

RESOLUTION MERGE OF 1: SCALE AERIAL PHOTOGRAPHS WITH LANDSAT 7 ETM IMAGERY

WATER BODY EXTRACTION FROM MULTI SPECTRAL IMAGE BY SPECTRAL PATTERN ANALYSIS

SAMPLE MIDTERM QUESTIONS

Partitioning the Conterminous United States into Mapping Zones for Landsat TM Land Cover Mapping

Advanced Image Management using the Mosaic Dataset

Selecting the appropriate band combination for an RGB image using Landsat imagery

Andrea Bondì, Irene D Urso, Matteo Ombrelli e Paolo Telaroli (Thetis S.p.A.) Luisa Sterponi e Cesar Urrutia (Spacedat S.r.l.) Water Calesso (Marco

HIGH SPATIAL RESOLUTION IMAGES - A NEW WAY TO BRAZILIAN'S CARTOGRAPHIC MAPPING

2.3 Spatial Resolution, Pixel Size, and Scale

Remote Sensing and Land Use Classification: Supervised vs. Unsupervised Classification Glen Busch

CBERS Program Update Jacie Frederico dos Santos Liporace AMS Kepler

Mosaicking and Subsetting Images

A land cover map for the Brazilian Legal Amazon using SPOT-4 VEGETATION data and machine learning algorithms

DETECTING LANDUSE/LANDCOVER CHANGES ALONG THE RING ROAD IN PESHAWAR CITY USING SATELLITE REMOTE SENSING AND GIS TECHNIQUES

Lake Monitoring in Wisconsin using Satellite Remote Sensing

The Role of SPOT Satellite Images in Mapping Air Pollution Caused by Cement Factories

A remote sensing instrument collects information about an object or phenomenon within the

Introduction to GIS (Basics, Data, Analysis) & Case Studies. 13 th May Content. What is GIS?

ANALYSIS OF FOREST CHANGE IN FIRE DAMAGE AREA USING SATELLITE IMAGES

Modeling deforestation to REDD+ Project: a case study in Alto Mayo Protected Forest, San Martin Region, Peru

Analysis of Landsat ETM+ Image Enhancement for Lithological Classification Improvement in Eagle Plain Area, Northern Yukon

Estimating Central Amazon forest structure damage from fire using sub-pixel analysis

TerraAmazon - The Amazon Deforestation Monitoring System - Karine Reis Ferreira

Review for Introduction to Remote Sensing: Science Concepts and Technology

APPLICATION OF MULTITEMPORAL LANDSAT DATA TO MAP AND MONITOR LAND COVER AND LAND USE CHANGE IN THE CHESAPEAKE BAY WATERSHED

SESSION 8: GEOGRAPHIC INFORMATION SYSTEMS AND MAP PROJECTIONS

Remote Sensing and GIS Application In Change Detection Study In Urban Zone Using Multi Temporal Satellite

Remote sensing is the collection of data without directly measuring the object it relies on the

DETERring Deforestation in the Brazilian Amazon: Environmental Monitoring and Law Enforcement

Supporting Online Material for Achard (RE ) scheduled for 8/9/02 issue of Science

IMAGINES_VALIDATIONSITESNETWORK ISSUE EC Proposal Reference N FP Name of lead partner for this deliverable: EOLAB

Accuracy Assessment of Land Use Land Cover Classification using Google Earth

How to calculate reflectance and temperature using ASTER data

Resolutions of Remote Sensing


Generation of Cloud-free Imagery Using Landsat-8

by César I. Delgado Dr. Jennifer Swenson, Advisor April 2008 Masters project final draft submitted in partial fulfillment of the

MULTIPURPOSE USE OF ORTHOPHOTO MAPS FORMING BASIS TO DIGITAL CADASTRE DATA AND THE VISION OF THE GENERAL DIRECTORATE OF LAND REGISTRY AND CADASTRE

Application of Remotely Sensed Data and Technology to Monitor Land Change in Massachusetts

APPLYING SATELLITE IMAGES CLASSIFICATION ALGORITHMS FOR SOIL COVER AND GEORESOURCES IDENTIFICATION IN NOVA LIMA, MINAS GERAIS - BRAZIL

The USGS Landsat Big Data Challenge

Temporal characterization of the diffuse attenuation coefficient in Abrolhos Coral Reef Bank, Brazil

APPLICATION OF GOOGLE EARTH FOR THE DEVELOPMENT OF BASE MAP IN THE CASE OF GISH ABBAY SEKELA, AMHARA STATE, ETHIOPIA

Development of Method for LST (Land Surface Temperature) Detection Using Big Data of Landsat TM Images and AWS

RESULT, ANALYSIS AND DISCUSSION

The Idiots Guide to GIS and Remote Sensing

UPPER COLUMBIA BASIN NETWORK VEGETATION CLASSIFICATION AND MAPPING PROGRAM

Information Contents of High Resolution Satellite Images

Methods for Monitoring Forest and Land Cover Changes and Unchanged Areas from Long Time Series

Pixel-based and object-oriented change detection analysis using high-resolution imagery

Obtaining and Processing MODIS Data

REDUCING UNCERTAINTY IN SOLAR ENERGY ESTIMATES

How To Update A Vegetation And Land Cover Map For Florida

Managing Imagery and Raster Data in ArcGIS

Some elements of photo. interpretation

REGIONAL SEDIMENT MANAGEMENT: A GIS APPROACH TO SPATIAL DATA ANALYSIS. Lynn Copeland Hardegree, Jennifer M. Wozencraft 1, Rose Dopsovic 2 INTRODUCTION

A PHOTOGRAMMETRIC APPRAOCH FOR AUTOMATIC TRAFFIC ASSESSMENT USING CONVENTIONAL CCTV CAMERA

A GIS helps you answer questions and solve problems by looking at your data in a way that is quickly understood and easily shared.

Remote Sensing an Introduction

T.A. Tarasova, and C.A.Nobre

Summary. Deforestation report for the Brazilian Amazon (February 2015) SAD

Land Use/ Land Cover Mapping Initiative for Kansas and the Kansas River Watershed

Global environmental information Examples of EIS Data sets and applications

Imagery. 1:50,000 Basemap Generation From Satellite. 1 Introduction. 2 Input Data

AAFC Medium-Resolution EO Data Activities for Agricultural Risk Assessment

Extraction of Satellite Image using Particle Swarm Optimization

Objectives. Raster Data Discrete Classes. Spatial Information in Natural Resources FANR Review the raster data model

VCS REDD Methodology Module. Methods for monitoring forest cover changes in REDD project activities

Ane Alencar CONDESSA BR163

ENVI Classic Tutorial: Atmospherically Correcting Multispectral Data Using FLAASH 2

MODIS IMAGES RESTORATION FOR VNIR BANDS ON FIRE SMOKE AFFECTED AREA

Lectures Remote Sensing

LANDSAT 8 Level 1 Product Performance

Remote Sensing, GPS and GIS Technique to Produce a Bathymetric Map

Impact of water harvesting dam on the Wadi s morphology using digital elevation model Study case: Wadi Al-kanger, Sudan

Mapping Land Cover Patterns of Gunma Prefecture, Japan, by Using Remote Sensing ABSTRACT

ENVIRONMENTAL MONITORING Vol. I - Remote Sensing (Satellite) System Technologies - Michael A. Okoye and Greg T. Koeln

High Resolution Information from Seven Years of ASTER Data

Aneeqa Syed [Hatfield Consultants] Vancouver GIS Users Group Meeting December 8, 2010

Monitoring alterations in vegetation cover and land use in the Upper Paraguay River Basin Brazilian Portion Period of Analysis: 2002 to 2008

New challenges of water resources management: Title the future role of CHy

Transcription:

Image processing and land-cover change analysis in the tri-national frontier of Madre de Dios (Peru), Acre (Brazil), and Pando (Bolivia) - MAP: an increasing demand for data standardization Karla da Silva Rocha 1,2 Andrea Chavez 2 Matt Marsik 2 Stephen George Perz 2 1 Universidade Federal do Acre - UFAC Caixa Postal 500-69920-900 Rio Branco - AC, Brasil rochakarla@uol.com.br 2 University of Florida - UF P.O. Box 117330-32611-7330- Gainesville - FL, USA {rocha, achavez, Marsik, sperz}@ ufl.edu Abstract A remote sensing database was constructed to analyze land cover change in the MAP region of Southwestern Amazonia. This database provides baseline observations to measure land cover changes in response to new road construction and paving in the region. Landsat (TM and ETM+) data for the MAP region across a period of 20 years (1986-2006) was obtained to evaluate land cover change. The image season was the austral winter, specifically July, with cloud-free images as temporally close as possible given the large geographic area. Climatic variability among image dates determined the exact image month and day to acquire each scene. The Landsat data were radiometrically calibrated, geometrically registered, normalized for precipitation differences (when necessary), and mosaicked. This database was constructed with in collaboration with the Universidad Amazonica de Pando (UAP); Universidade Federal do Acre (UFAC); Universidad Nacional Amazonica de Madre de Dios (UNAMAD) and University of Florida (UF). This paper highlights the need for institutional collaboration, data and method of remote sensing standardization across the tri-national boundary. Key Words: image processing, land-cover change, Amazon, processamento de imagem, mudanças de cobertura da terra, Amazônia. 1. Introduction Changes in land cover, in particular in Southwest Amazonia, have increased due to new road infrastructure (Arima 2005, Brown, et al. 2002). These changes are expected to accelerate socio-economic, environmental, and political processes. Recent political events indicate that the Southwestern portion of this region, comprised by the department of Madre de Dios (Peru), state of Acre (Brazil) and department of Pando (Bolivia) MAP and linked by the Inter-Oceanic Highway represents hope for biodiversity conservation. Therefore, MAP represents a priority site for innovative scientific research to monitor, understand, and plan for the complex changes that are already occurring. Simulations of Soares-Filho et al (2004) indicate that this region will be heavily impacted in the coming years, demanding thus, precisely more research to better understand the future changes and impacts brought by the paving of the inter-oceanic highway. The analysis is based on the NSF Human Social Dynamics Program, project: Infrastructure Change, Human Agency, and Social-Ecological Resilience which focuses on the impacts of the Inter-Oceanic Highway in the tri-national MAP frontier region. The analysis 8176

required processing of 8 Landsat images to evaluate land cover change in the region across a period of 20 years from 1986 to 2006. For each date the images underwent preprocessing for radiometric and geometric correction, climate normalization, and mosaicking. The length and complexity of the process reminded about the importance of data standardization. The time series analysis generated methods of data standardization for image processing and land cover change employing the combined techniques of remote sensing and geographic information system. Data standardization plays a very important role to understand land cover dynamics in regions where large-scale infrastructure projects need to be integrated. In addition, it is also important to minimize errors for land cover change analysis. Thus, this paper addresses the following question: How do we undertake large-scale remote sensing based analysis to document and quantify land cover change in the MAP region? How do we standardize methods and minimize errors and make decisions more applicable? 1.2 The Need for RS Analysis It s clear enough that there are important differences among the three countries that compose the MAP region, but quantifying those differences are major challenges in image processing standardization (radiometric calibration, geometric correction and image mosaicking). Therefore, an increase need for development of spatial data standardization for image processing is crucial. However, the challenge to develop efficient mechanisms by which such data could be understood, assessed, transferred and shared between different sources, applications, and systems remains. The digital spatial database produced fills up a gap of lack of standardization and need for a tri-national remote sensing database that facilitates comparisons of land-use and land-cover change (LULCC across borders). It also allows direct comparisons of three distinct countries in regard to socio-economic and political differences and with similar ecological conditions. Moreover, it is important to note that standardization techniques are well-established, but they are time-consuming and more difficult if a large number of images are involved for analyses of large-scale environmental changes of larger regions such as the case reported in this paper. Local universities within the MAP region are using remote sensing (Brow at al 2004, Maldonado at al 2007), however different types of digital data are being produced in different formats, projections, scale and different methods and software of image processing, thus reducing the accuracy of data produced. Not much has been done about image calibration and remote sensing standardization for comparison across national borders in MAP region. Researchers skip discussion and standardization of radiometric calibration, geometrical correction and mosaicking, but in this project where we are using multiple images, it is crucial that processing efforts be standardized in order to allow for a more accurate analysis. We standardized image processing for a large data base of images across space and time (radiometric calibration, geometrical correction and mosaicking) in order to get accurate data ready to be used for analysis. The main goal of data standardization is to minimize errors for land cover change analysis, especially in large region where multiple images are required for each time step. It also allows us to compare changes in ecologically similar forest ecosystems, even across national boundaries and provides decision makers with data to create and develop more effective polices for sustainable development of natural resources. This paper focuses on the importance of data standardization and integration for large scale remote sensing project in the MAP region. The standardized digital data base produced will help scientist and regional decision makers evaluate the impact of the road on LULCC). It will hopefully provide valuable information in order to 8177 2

create more effective policies for sustainable development of natural resources in a region of rich biodiversity and social diversity which makes integrated research and monitoring very important tools. 2. The Study Area The MAP region covers the department of Madre de Dios/Peru, the state of Acre/Brazil and department of Pando/Bolivia, also known as the MAP region (Figure 1). The tri-national frontier covers an area of approximately 300,000 km 2, of this 84,000 km 2 in Madre de Dios (Peru), 153,000 km 2 in Acre (Brazil) and 63,000 km 2 in Pando (Bolivia). It is one of the most rich biodiversity regions of the world with 85 % of its forests preserved (Brown et al., 2004, Myers, et al. 2000). It also has a highly cultural diversity, containing extractive communities such as rubber tappers and riverine, colonists, indigenous groups, big ranchers and urban population. This region exhibits a complex mosaic of land tenure categories and management regimes, ranging from biological, indigenous and extractive reserves to national parks, agroextractive settlement, and private lands. However, this region is undergoing unprecedented change due to infrastructure projects, notably paving of the BR-317 federal highway, which reached through Brazilian to the Bolivian and Peruvian borders in 2002. Plans to continue paving this road, part of the Inter-Oceanic Highway, through Peru and Bolivia ultimately links southern Brazil to Pacific ports, thereby articulating MAP to both the Atlantic and Pacific arenas of the global economy. Figure 1: Study area 3. Methods of standardization Data were obtained with NSF HSD acquisition funds; a time series of Landsat (TM and ETM+) for the MAP region across a period of 20 years from 1986 to 2006, to evaluate land cover change. Data consisted of eight scene footprint from path 1 to 3, and row 67 to 69, excluding scene path 1 row 69. Image analysis was done in a time intervals of every five years from 1986 to 2001. Since the infrastructure projects in the region began in 2000, the time 8178 3

intervals (from 2000 to 2006) for image processing were collected every year in order to measure the impact caused by the highway paving. As a result, the temporal resolution of image processing consisted of 10 years Landsat time series of 8 images each and a total of 80 images. For each date, images underwent preprocessing for radiometric and geometric correction, climate normalization, and mosaicking. 3.1 Radiometric calibration Image calibration was performed for both the thermal and reflective bands to eliminate sources of variability such as noise, differences due to satellite instrumentation, Earth-Sun distance, solar elevation angle, solar curve and atmospheric effects (Jensen 2005). The process facilitates comparison over both space (mosaic) and time (1986-2006) and of socio-economic and political differences of these three distinct countries. It is especially important for timeseries analysis, but also for comparisons across different satellite sensors (Southworth 2005). This method was standardized using the CIPEC protocol which was developed by Glen Green (University of Indiana) for Landsat TM5 and Landsat ETM+ 7, available at http://ltpwww.gsfc.nasa.goc. It has pre-established standardized procedures for image registration and calibration, allowing for comparison of results across a larger region than this study alone, potentially the full CIPEC meta database. 3.2 Geometric Correction procedure Satellite image registration (Figure 2) is critical for applications such as mosaicking, change detection, cloud removal and digital elevation model generation. Reference or base images were selected for each path and row from the GLCF (University of Maryland) Landsat Geocover dataset. Subsequently, all other radiometrically calibrated images were georeferenced to their respective base image. For each image, approximately 45-60 Ground Control Points - GCP per image were selected, distributed as uniformly as possible across each image. These included the natural (river junctions), man-made (road intersections), and features readily identifiable. Preference was given to GCPs considered to be more constant over time (e.g. road intersections). There were images however, where other points had to be used, as they had little infrastructure. These included oxbow lakes and vertices of cleared areas (Figure 3). Once all GCPs were placed, the RMS error was evaluated (Jensen 2005). If it was greater than 0.5 pixels (15m), the input and reference GCPs were modified by adjusting their position, until an acceptable total RMS error was obtained. All 80 images were rectified and a RMS error obtained was lower than 0.5 pixels. Subsequently, a polynomial interpolation algorithm was used to resample the images using a Nearest Neighbor method and output cell size of 30m. For images with significant relief (i.e. Andes foothills), a Landsat geometric model was used (Erdas imagine) to correct for relief displacement. 8179 4

Figure 2. Landsat ETM + registration, path 02, row 67, 2004. Road Intersections River junction Vertices of cleared areas Figure 3. Example of areas selected as GCPs. 3.3 Mosaicking It is a common knowledge that a study area may be larger in size than any one satellite image or aerial photo; straddle the overlap region of two images. In such situations, it is necessary to create single, larger images by stitching the smaller pieces together or combining multiple images into a single seamless composite image (Jensen 2005). As the study site has approximately 300,000 km², it was necessary to mosaick in order to have a better understanding of the land cover pattern and change in the region (Figure 4). Data for a period of 20 years (1986-2006) were mosaicked. 8180 5

Figure 4. Pre-mosaicking of the study area showing 8 footprint images form path 1 to 3 and rows 67 to 69. For this paper, each time step (10), 8 pre-processed images (preprocessing included radiometric and geometric corrections, as well as normalization for precipitation differences) were mosaiked. There are a number of functions and tools included with Erdas Imagine to aid to create a better mosaicking image from many separate images. In this project images mosaicking were standardized by using the following functions into Erdas Imagine Mosaicking tool: Image Dodging, Color Balancing and Histogram Matching. The Image Dodging feature of the Mosaic Tool applied a filter and global statistics across each image that was mosaicking in order to smooth out light imbalance over the image. On the statistics collection function, the images were partitioned in 15 blocks for both X and Y and a grid size of 15m were applied, which means that the images were cut into 15 sections in both X and Y, generating, therefore 225 blocks. We applied a skip factor of 1 (one) for both x and y which means that no pixels were skipped, and the image dodging was most accurate. The mosaicking color balancing tools was used to balance color disparities between images, it removed the brightness variations in images before they were mosaicked together into one large image. There are four surface options to balance any color disparities in the images, which is parabolic, linear, conic, and exponential surface. It can be selected depending on the patterned color difference observed in the image (a patterned color difference is the shape of a particular color difference in an image). We also applied a Histogram Matching to match data of the same or adjacent scenes that was captured on different days, or data that is slightly different because of sun or atmospheric effects. These processing methods used in this analyses (radiometric calibration, geometric correction and mosaicking) are all well-known, but when using them on a large scale project it turns very complex, demanding for a standardized methodology in order to produce accurate data for study of road impacts on LULCC through remote sensing methods. 8181 6

4. Results and conclusion The methods of image processing standardization is very important to understand land cover dynamics in regions that are being integrated by large-scale infrastructure projects and also to compare changes in ecologically similar forest ecosystems across multiple countries. Therefore, the standardized Landsat image data produced provided baseline observations to measure land cover changes in response to new road construction and paving in the region. The 80 images were radiometric calibrated, georeferenced and mosaicked, see (Figure) 5 for results obtained by the standardized methods produced by this research. 1986 Mosaic 1996 Mosaic 2005 Mosaic Mosaic 1986 Mid Infrared Index Mosaic 1996 Mid Infrared Index 2005 Mosaic Figure 5. Temporal mosaic produced The data base produced has already being used for a variety of applications such as to evaluate land use and land cover change - LULCC and analyze the impacts of the new road infrastructure in the region. Some analysis using the standardized data base produced has already been undertaken by the HSD project RS component. Besides the HSD project there is a great demand for using this database from other researchers and projects in the MAP region, an example is the G-MAP consortium and the regional universities. Data products from this research are also of potential interest to MAP initiative organizers, who include many decisionmakers around the region. The standardized data base will allow combining spatial data into a GIS workspace with multiple layers for administrative boundaries, various land tenure parcels, roads and rivers, social and agricultural data, vegetation, and satellite data allowing thus integration of data. MAP thus represents a priority site for innovative scientific research to monitor, understand, and plan for the complex upcoming changes. Acknowledgements The close collaboration of the Infrastructure Change, Human Agency, and Resilience in Social-Ecological Systems project, RS component participants. Many contributed to this research; therefore we would like to acknowledge the crucial contribution from all our colleagues, who directly or indirectly collaborated to this work, especially Natalia Hoyos, Forrest 8182 7

Stevens, Frank Barra, Jane Southworth. I also thank NSF s "Human and Social Dynamics" program, grant # 0527511. References Arima, Eugenio Y., Robert T. Walker, Stephen G. Perz, and Marcellus Caldas. Loggers and Forest Fragmentation: Behavioral Models of Road Building in the Amazon Basin. Annals of the Association of American Geographers. 2005. Brown, I.F., Selhorst, D., Pantoja, N.V., Mendoza, E.R.H., Vasconcelos, S.S. de, Rocha, K. da S. Os desafios do monitoramento de desmatamento, queimadas e atividade madeireira na região MAP área fronteiriça de Bolívia, Peru e Brasil. In: Aplicações de Geotecnologia na Engenharia Florestal, eds. A.A. Disperati & J.R. dos Santos, Curitiba, Brazil: Copiadora Gabardo Ltda. 2004, pp. 70-77. Brown, I.F., S.H.C. Brilhante, E. Mendoza, and I. Ribeiro de Oliveira. Estrada de Rio Branco, Acre, Brasil aos Portos do Pacífico: Como Maximizar os Benefícios e Minimizar os Prejuízos para o Desenvolvimento Sustentable da Amazônia Sul-Ocidental. In CEPEI, La Integración Regional Entre Bolivia, Brasil y Peru. Lima: CEPEI. 2002. Jensen, J.R. Introductory Digital Image Processing.Prentice-Hall, Englewood Cliffs, NJ. 2005. Maldonado, M. J. De Los Rios, I. F. Brown, D. Valeriano, V. Duarte. Modificações no método do PRODES para estimar a mudança da cobertura florestal na Bacia Trinacional do Rio Acre na região de fronteira entre Bolívia, Brasil e Peru na Amazonia Sul-ocidental. Anais XIII Simpósio Brasileiro de Sensoriamento Remoto, Florianópolis, Brasil, 21-26 abril 2007, INPE, p. 5903-5910. Soares-Filho, B., A. Alencar, D. Nepstad, G. Cerqueira, M. del C. Vera Diaz, S. Rivero, L. Solórzano, and E. Voll. Simulating the Response of Land-Cover Changes to Road Paving and Governance along a Major Amazon Highway: The Santarém-Cuiabá Corridor. Global Change Biology 10: 745-764. 2004. Southworth, J. An assessment of Landsat TM Band 6 thermal data for analyzing land cover in tropical dry forests. International Journal of Remote Sensing, 25: 689-706. 2004. 8183 8