RCMRD PROJECT IMPLEMENTATION GUIDE: MALAWI LAND COVER MAPPING FOR GREEN HOUSE GAS INVENTORIES DEVELOPMENT PROJECT IN EAST AND SOUTHERN AFRICA REGION

Size: px
Start display at page:

Download "RCMRD PROJECT IMPLEMENTATION GUIDE: MALAWI LAND COVER MAPPING FOR GREEN HOUSE GAS INVENTORIES DEVELOPMENT PROJECT IN EAST AND SOUTHERN AFRICA REGION"

Transcription

1 RCMRD PROJECT IMPLEMENTATION GUIDE: MALAWI LAND COVER MAPPING FOR GREEN HOUSE GAS INVENTORIES DEVELOPMENT PROJECT IN EAST AND SOUTHERN AFRICA REGION Drafted by RCMRD-SERVIR Africa November 2014

2 DOCUMENT VERSION CONTROL The following changes have been made to this document: Version Date Change / Addition Assumptions Authorised Version 0 26/05/201 2 Version 2 03/10/201 2 Version 3 21/03/201 2 Zero Draft Page 1 of 110

3 ACRONYMS AFOLU ALU Tool EPA ESA Region ETM ETM GDA GHG GIS GPS IPCC LC LULUCF NASA QA QC RCMRD ROIs TM UNFCC USAID USEPA Agriculture, Forestry and Land Use Agriculture and Land Use (ALU) Tool Environmental Protection Agency East and Southern Africa Region Enhanced Thematic Mapper Enhanced Thematic Mapper Geospatial Data Analysis Corporation Green House Gas Geographical Information Systems Global Positioning System Intergovernmental Panel on Climate Change Land Cover Land use, land-use change and forestry National Aeronautics and Space Administration Quality Assessment Quality Control Regional Centre for Mapping of Resources fro Development Regions of interest Thematic Mapper United Nations Framework Convention on Climate Change United States Agency for International Development US Environmental Protection Agency Page 2 of 110

4 USGS UTM US Geological Surveys Universal Transverse Mercator WGS84 World Geodetic System 1984 Page 3 of 110

5 PURPOSE AND SCOPE This document is designed to be a guide for the country teams in developing and implementing land cover mapping efforts for Agriculture, Forestry and Land Use (AFOLU) sector in the development of Green House Gas (GHG) Inventory that meet IPCC requirements. The document introduces users to the IPCC guidelines and illustrates processes, requirements, and provides methods and procedures to produce accurate and precise estimates of land areas. This guide is not meant to be an instruction, but should be used in conjunction with the IPCC Good Practice Guidance on Land Use, Land-Use Change and Forestry (2003) and or AFOLU (2006). However, it is intended as an addition to, providing explanation, clarification and enhanced methodologies for collecting ancillary data, processing and classifying Landsat imagery, carrying out ground referencing and validation of the land cover maps generated. Page 4 of 110

6 ACKNOWLEDGEMENT Dr. Tesfaye Korme, Dr. Ashutosh Limaye, and Andre Kooiman played a central role in developing the methods and procedures presented here. We also wish to thank the organizations and agencies that have supported and funded this work, in particular, US Agency for International Development (USAID), NASA, USEPA, UNFCCC and ICFI. Finally, we acknowledge the following: Fredrick Mokua, Viola Kirui, Hilda Manzi, Phoebe Oduor, Benson Kemboi, Ababu Wakayanga, Susan Kotikot and Eric Ng ang a, who worked innovatively to generate the land cover maps and immensely contributed to development of the procedures covered in this guide. Content on ground referencing techniques and considerations for choice of sampling methodology were largely derived from presentations by Zac Andereck, (ICFI) during the project kick-off workshop held in Mauritius in March 2011, under discussions on Central America land use land cover mapping lessons Page 5 of 110

7 PROJECT SUMMARY In response to the request made by eight ESA countries, the UNFCCC in collaboration with other partners has initiated a Capacity Building Project for the development of Sustainable National GHG Inventory Management Systems in the Eastern and Southern African (ESA) region. Although the countries in the region have made some progress in preparing greenhouse gas (GHG) inventories, there are significant constraints that still exist in the development of quality and sustainable national GHG inventory systems. Following the project kick-off workshop held from March 7-9, 2011 in Mauritius; it has been identified that the countries are challenged in quality and sustainable accounting of the LULUCF component of the GHG inventory development. Regional Centre for Mapping of Resources for Development (RCMRD) an intergovernmental organization in Africa mandated for provision of capacity building services to the member states; is building capacity for land cover mapping in the participating countries which are also RCMRD member states. The main deliverables for the project are land cover maps for the years 1990, 2000 and 2010, and land cover mapping training services offered to the countries. The activities of the project include capacity needs assessment, collection of auxiliary (existing/historical) land use land cover and ground reference data, data accuracy checks, validation, and land cover classification using Landsat imagery. The project is intended for the development of adequate, consistent and replicable procedures for creating complete, transparent and comparable land cover data compilations and map products for the countries. Page 6 of 110

8 TABLE OF CONTENTS DOCUMENT VERSION CONTROL... 1 ACRONYMS... 2 PURPOSE AND SCOPE... 4 ACKNOWLEDGEMENT... 5 PROJECT SUMMARY... 6 TABLE OF CONTENTS... 7 LIST OF FIGURES SECTION 1: PROJECT BACKGROUND, SCOPE AND PRIORITIES SECTION 2: ANCILLARY DATA COLLECTION AND CLASSIFICATION SCHEME DESIGN Introduction Ancillary Data Collection Workshop Objectives Results for Ancillary Data Collection Activity: Quality check Workshop Proceedings Day One Workshop Proceedings Day Two a) IPCC Definitions b) National Descriptions Day Three and Four SECTION 3: DATA ACQUISITION, QUALITY ASSESSMENT AND ARCHIVING Data Acquisition Malawi Landsat Imagery Acquisition Ancillary Data Acquisition Data Quality Assessment Page 7 of 110

9 Landsat Image Quality Assessment a) Radiometric QA b) Atmospheric QA c) Geometric QA Quality Assessment of Landsat Imagery Ancillary Datasets Quality Assessment a) Lineage b) Positional accuracy c) Logical consistency d) Attribute accuracy e) Completeness Malawi Data Management SECTION 4: IMAGE PROCESSING AND CLASSIFICATION PROCEDURES Image processing Layer stacking Cloud Masking Computing Image statistics Sub-Setting Imagery to Country Boundary Band compositing and selection Image Classification Classification Scheme Selection Important Classification Factors Classification methods a) Image Interpretation using Ancillary Data and Google Earth c) Separability Analysis Page 8 of 110

10 d) Classification Using Maximum Likelihood Post classification processing Quality Assessment of Classification Using Landsat Imagery Quality Assessment of Classification Using Google Earth ROI Merge Decision tree Class coding and Color Matching Classification of Settlements Edge matching procedure SECTION 5: GROUND REFERENCING AND ACCURACY ASSESSMENT Sample design for Ground Referencing Applied Sampling Methodology Proportionate Stratified Random Sampling Procedure; Field Data Collection Form Collection of Ground reference data Field Preparation Prerequisites Before leaving fieldwork Materials and Tools needed Tasks at the Field Trimble Juno SB fieldwork methods Collecting points of interest (POI) Independent Validation Point Interpretation for Accuracy Assessment Accuracy Assessment Procedures SECTION 6: RESULTS Page 9 of 110

11 6.1. Accuracy Statistics for 2010 Classification Scheme I Accuracy Statistics for 2010 Classification Scheme II Accuracy Statistics for 2000 Classification Scheme I Accuracy Statistics for 2000 Classification Scheme II Overall Accuracy = (928/1197) % Kappa Coefficient = Accuracy Statistics for 1990 Classification Scheme I Reference Data... Error! Bookmark not defined. Classified Data... Error! Bookmark not defined. Ground Truth (Percent)... Error! Bookmark not defined. Classified Data... Error! Bookmark not defined Accuracy Statistics for 1990 Classification Scheme II REFERENCES Ground Referencing Pictures Showing Major Land Cover Categories Page 10 of 110

12 LIST OF FIGURES Figure 1: Section Image Quality Report Figure 2: GHG Geo-database tool description Figure 3: GHG GDB graphic model Figure 4: Raster Catalog files containing various datasets Figure 5 Main Land Cover Mapping steps Figure 6: The Compute Statistics Parameters Dialog Figure 7: Sample of computed results Figure 8: Work flow for Image Classification Figure 9: Vegetation Delineation Tool Figure 10: Separability Analysis Window Figure 11: Selected training areas for water bodies Figure 12: Check of classification against Landsat Imagery Figure 13: Illustrating ROI Merge Figure 14: Land Cover Coding Tree Figure 15: Land cover map with settlements in grey Figure 16: Multiple ring buffer tools in Arc Tool box Figure 17: District Headquarter sampling zone Figure 18: Probable Sampling Zone along Major Roads Figure 19: Defined sampling zones (in pink) Figure 20: Defined sampling zones (in green) Figure 21 Sub-setting vector file parameters Figure 22: Sampling zone ROIs overlaid on classified image Figure 23: Subset Data via ROIs Tool Figure 24: Sampling zones clip Figure 25: Generate Random Sample tool Figure 26: Random Generated Validation points for Accuracy Assessment Figure 27: Malawi Land Cover Map for 2010 Scheme I Figure 28: Malawi Land Cover Map for 2010 Scheme II Figure 29: Malawi Land Cover Map for 2000 Scheme I Figure 30: Malawi Land Cover Map for 2000 Scheme II Figure 31: Malawi Land Cover Map for 1990 Scheme I Figure 32: Malawi Land Cover Map for 1990 Scheme II Figure 33 Land cover mapping workflow Page 11 of 110

13 SECTION 1: PROJECT BACKGROUND, SCOPE AND PRIORITIES The land use sector; forestry and agriculture, is an important source of anthropogenic GHG emissions. LULC change, mainly due to deforestation, has been found to contribute to about 20% of the GHG emissions from anthropogenic sources (IPCC, 2000 and 2007c). Land use, landuse change and forestry (LULUCF) sector in general has an aggregate share of over 30% of the gross global emissions. This makes LULUCF a critical component in accounting for GHG emissions. In accounting for emissions from this sector; quality land cover maps are an important requirement. The Eastern and Southern Africa project to improve regional capacity for conducting national greenhouse gas inventories was initiated in March 2011 with the goal of improving inventories for the Agriculture and Land Use, Land-Use Change and Forestry (LULUCF) Sectors. The two sectors require land cover maps as a key component of the national activity data. Considering the importance of consistent, reliable, accurate and relevant land cover information required for GHG Inventories; most countries have made an effort to develop land cover maps. In addition the countries have limited resources available for land cover mapping leading to the development of incomplete and incomparable land cover maps. This guide provides the required consistent and replicable procedures to come up with harmonized compilation of the land cover data at national level. RCMRD working closely with NASA through the SERVIR Africa project is building capacity of the countries in developing quality and sustainable GHG Inventories by assisting in the development of land cover maps and offering training and data dissemination services. SERVIR Africa is a regional visualization and monitoring system for Africa that integrates the use of satellite imagery and other geospatial data for improved scientific knowledge and decision making. This project builds upon RCMRD s rich experience acquired over the many years in the region, a team of skilled staff, rich databases of spatial data, global linkages with earth observation agencies and supporting partners and stakeholders. Page 12 of 110

14 SECTION 2: ANCILLARY DATA COLLECTION AND CLASSIFICATION SCHEME DESIGN 2.1. Introduction This section is drawn to guide the initial in-country workshops for ancillary data collection for the land cover mapping effort in the GHG Inventories. GHG inventory includes the estimation of carbon stock, emissions and removal of greenhouse gas resulting from Agriculture, Forestry and Land Use (AFOLU) sector activities. This requires spatial information on the extent of major land cover and land use categories and their variation over time making ancillary data collection one of the major activities in the effort of mapping land cover (LC) for the development of GHG Inventories. The objective for collecting this data is to ensure that consistent and accurate information is made available for proper definition of national land cover classification schemes. It is also needed to guide and verify the classification of digital satellite imagery. The data includes existing land cover and land use maps, annual agricultural census, periodic land use surveys, existing forest maps, relevant reports and publications, ground reference locations, high spatial resolution imagery and any other remote sensing data. The availability of this data is a critical starting point for laying a baseline for GHG Inventories. The ancillary data collection workshop serves as a platform for involvement of key national stakeholders and exercising openness in data collection, processing, application, sharing and archiving as discussed and agreed upon in the workshop. During these sessions, the national GHG Inventory development team is required to provide information on the availability of the required data and identify key national stakeholders in land use land cover mapping. Finally an assessment of the capacities of the national institutions in land cover mapping is ascertained and in consultation with country team preferred classification scheme is decided upon. The land use maps considered in this guide include countrydefined categories for land cover including forest types, grassland types, cropland types, wetlands and other lands whose sub-categories are informed by the national policies and mapping goals, definitions and descriptions. Page 13 of 110

15 2.2. Ancillary Data Collection The collection, reworking and capturing of quality information, forms one of the major activities of the GHG project. The data collection starts with a workshop on the assessment of the past and ongoing efforts on land cover mapping in a country. The workshop session is meant for introducing the mapping component of the GHG inventory and explanation and development of the standards, guidelines and methodologies for land cover mapping, adapting these to national preferences and practices. The National GHG inventory team is charged with identifying key stakeholders and with making ancillary data available to the project. The specific workshop tasks include: i) Gathering existing and or historical land use maps and previously collected ground reference data ii) Identifying the classification scheme to be used within each country iii) Reworking and documenting the metadata of the existing land use land cover and related products iv) Ensuring that enough relevant data is made available for classification of satellite imagery to the required classes/categories v) Documenting national description of land use land cover categories in reference to IPCC guidelines and subcategories vi) Identifying data gaps with regard to land cover categories Workshop Objectives The workshop objectives include the following; To gather existing and or historical land use maps and previously collected ground reference data To identify the classification scheme to be used within Participating country To document national descriptions of land use land cover categories and sub categories To ensure that enough relevant ancillary datasets are available for classification of Landsat imagery to the required classes/categories Page 14 of 110

16 To assess capacity and training needs with respect to land cover mapping for GHG Inventory Results for Ancillary Data Collection Activity: The results include; A list of ancillary data collected Set of metadata Classification scheme and Legend, description of classes or land cover categories and discrimination criteria Quality check After collection, the data is checked for accuracy, consistency and completeness as shown in the workflow diagram in Annex I. This is done before data is adopted for use in image interpretation. The main tasks include; receiving data, performing initial quality assessment and requesting for clarification from the data providers if required, verifying, reworking, validating the data and carrying out super overlays on Google Earth to check consistency Workshop Proceedings Day One a) The inaugural session i) This session begins with a welcoming note from the national coordinator for the GHG Inventory Development project and introduction of the participants, resource persons and the invited guests. ii) Introduce the overall GHG Inventory Development Program and other presentations by RCMRD team. b) Presentations i) Status and Plan for the GHG Land Cover Mapping Project (Please refer to Annex I and workshop report for details) o Covers the background of the Land Cover mapping for GHG Inventory Development project o The objectives of the Land Cover mapping for GHG Inventory Development project o The methodology and anticipated results. Page 15 of 110

17 ii) ii) Thematic Needs and Land cover Mapping Challenges: Ancillary Data Availability (Details shown in the Annex I and workshop report) o Introduce the participants to land cover mapping processes and the guiding principles for production of suitable maps for the GHG Inventory maps. o Introduce the participants to the factors that determine the usefulness of the maps generated namely; purpose, consistency and prior classification scheming. iii) Initiatives and or Status Land Cover Mapping in participating country: Min./depts. of Agriculture, Forestry, Lands and surveys, Natural Resources, Environment, other institutions represented in the workshop iv) IPCC Definitions of Land Cover Categories for GHG Inventories v) c) Available Data at Departmental Level i) During session iii) above, the workshop participants should indicate the kind of datasets available at their respective departments and or possible data provider s preferably governmental bodies Workshop Proceedings Day Two a) IPCC Definitions The following land-use categories are defined for greenhouse gas inventory reporting as provided by the IPCC guidelines and discussed during the workshop sessions: 1. Forestland This category includes all land with woody vegetation consistent with thresholds used to define Forest Land in the national greenhouse gas inventory. It also includes systems with a vegetation structure that currently fall below, but in situ could potentially reach the threshold values used by a country to define the Forest Land category. 2. Cropland This category includes cropped land, including rice fields, and agro-forestry systems where the vegetation structure falls below the thresholds used for the Forest Land category. 3. Grassland This category includes rangelands and pasture land that are not considered Cropland. It also includes systems with woody vegetation and other non-grass vegetation such as Page 16 of 110

18 herbs and brushes that fall below the threshold values used in the Forest Land category. The category also includes all grassland from wild lands to recreational areas as well as agricultural and silvi-pastural systems, consistent with national definitions. 4. Wetlands This category includes areas of peat extraction and land that is covered or saturated by water for all or part of the year (e.g., peat-lands) and that does not fall into the Forest Land, Cropland, Grassland or Settlements categories. It includes reservoirs as a managed sub-division and natural rivers and lakes as unmanaged sub-divisions. 5. Settlements This category includes all developed land, including transportation infrastructure and human settlements of any size, unless they are already included under other categories. This should be consistent with national definitions. 6. Other Land This category includes bare soil, rock, ice, and all land areas that do not fall into any of the other five categories. It allows the total of identified land areas to match the national area, where data are available. If data are available, countries are encouraged to classify unmanaged lands by the above land-use categories (e.g., into Unmanaged Forest Land, Unmanaged Grassland, and Unmanaged Wetlands). This will improve transparency and enhance the ability to track land-use conversions from specific types of unmanaged lands into the categories above. b) National Descriptions i) The participants are engaged in a discussion to highlight their country specific descriptions of the land cover and land use categories. ii) The descriptions should conform to the existing country policies i.e. forest with regards to the forest land cover category. iii) If descriptions do not exist for the rest of the land cover types, the participants should indicate the official national specific categories descriptions. iv) The participants should come up with the required sub-categories for all the main IPCC classes Page 17 of 110

19 Considerations for National Category Definitions IPCC (2006) provides general non-prescriptive definitions for the six main land-use categories: Forest Land, Cropland, Grassland, Wetlands, Settlements and Other Land. In order to reflect national circumstances, country-specific definitions are developed, based predominantly on criteria used in the land-use land cover mapping and or surveys in the countries. However, the definitions for Other Lands and Wetlands are based on the IPCC (2006) definitions for these categories. 1. Forest Land: Considerations a) Area: A land-use category that includes areas at least 36.6 m wide and 0.4 ha in size 1 with at least 10 percent cover 2 (or equivalent stocking) by live trees of any size, including land that formerly had such tree cover and that will be naturally or artificially regenerated. b) Forest land includes transition zones, such as areas between forest and non-forest lands that have at least 10 percent cover (or equivalent stocking) with live trees and forest areas adjacent to urban and built-up lands. c) Roadside, streamside, and shelterbelt strips of trees must have a crown width of at about 36m 3 and continuous length of at about 120 m 4 to qualify as forest land. d) Unimproved roads and trails, streams, and clearings in forest areas are classified as forest if they are less than about 36 m wide or 0.4 ha in size 5 ; otherwise they are excluded from Forest Land and classified as Settlements. 1 The minimum land area of a group of tree to qualify as forest, according to the national forest policy if existing 2 The minimum percent canopy cover of a group of tree to qualify as a forest, according to the national forest policy statements 3 The minimum strip width of a group of tree to qualify as a forest, according to the national forest policy statements 4 The minimum strip length of a group of tree to qualify as a forest, according to the national forest policy statements 5 The maximum size of areas of clearings or roads to be ignored for a group of tree to qualify as a forest, according to the national forest policy statements Page 18 of 110

20 e) Tree-covered areas in agricultural production settings, such as fruit orchards, or tree-covered areas in urban settings, such as city parks, are not considered forest land (Smith et al. 2009). 2. Cropland: Considerations a) A land-use category that includes areas used for the production of adapted crops for harvest; this category includes both cultivated and non-cultivated lands. b) Cultivated crops include row crops or close-grown crops and also hay or pasture in rotation with cultivated crops. c) Non-cultivated cropland includes continuous hay, perennial crops (e.g., orchards) and horticultural cropland. d) Cropland also includes land with alley cropping and windbreaks, as well as lands in temporary fallow or enrolled in conservation reserve programs (i.e., set-asides170). Roads through Cropland, including interstate highways, state highways, other paved roads, gravel roads, dirt roads, and railroads are excluded from Cropland area estimates and are, instead, classified as Settlements. 3. Grassland: Considerations a) A land-use category on which the plant cover is composed principally of grasses, grass-like plants, forbs, or shrubs suitable for grazing and browsing, and includes both pastures and native rangelands. This includes areas where practices such as clearing, burning, chaining, and/or chemicals are applied to maintain the grass vegetation. b) Savannas, some wetlands and deserts, low woody plant communities and shrubs, such as mesquite, mountain shrub, etc. are also classified as Grassland if they do not meet the criteria for Forest Land. Page 19 of 110

21 c) Grassland includes land managed with agro forestry practices such as silver-pasture and windbreaks, assuming the stand or woodlot does not meet the criteria for Forest Land. d) Roads through Grassland, including highways, other paved roads, gravel roads, dirt roads, and railroads are excluded from Grassland area estimates and are, instead, classified as Settlements. 4. Wetlands: Considerations a) A land-use category that includes land covered or saturated by water for all or part of the year. b) Managed Wetlands are those where the water level is artificially changed, or were created by human activity. c) Certain areas that fall under the managed Wetlands definition are covered in other areas of the IPCC guidance and/or the inventory, including Cropland (e.g., rice cultivation), Grassland, and Forest Land (including drained or untrained forested wetlands). 5. Settlements: Considerations a) A land-use category representing developed areas consisting of units of 0.25 acres (0.1 ha) or more that includes residential, industrial, commercial, and institutional land; construction sites; public administrative sites; railroad yards; cemeteries; airports; golf courses; sanitary landfills; sewage treatment plants; water control structures and spillways; parks within urban and built-up areas; and highways, railroads, and other transportation facilities. b) Also included are tracts of less than 10 acres (4.05 ha) that may meet the definitions for Forest Land, Cropland, Grassland, or Other Land but are completely surrounded by urban or built-up land, and so are included in the settlement category. Page 20 of 110

22 c) Rural transportation corridors located within other land uses (e.g., Forest Land, Cropland) are also included in Settlements. 6. Other Land: Considerations a) A land-use category that includes bare soil, rock, ice, and all land areas that do not fall into any of the other five land-use categories. b) It allows the total of identified land areas to match the managed national area. Data Collection Appointments i) All appointments for data collection are confirmed at the end of the workshop just before the closing remarks are made. ii) The offices to be visited and the expected data are clearly documented Day Three and Four These days are spent for departmental office visits i) The departments should be visited at the appointed time for data collection ii) The data should be collected using a hard drive and all documentation/metadata included. iii) All rights on data sharing must be inquired and documented iv) All documentations and policy documents should be collected from the respective offices to support decisions made during category descriptions. Page 21 of 110

23 SECTION 3: DATA ACQUISITION, QUALITY ASSESSMENT AND ARCHIVING This section provides information on data acquisition and guidelines for quality checking of Landsat images imagery and ancillary data used for land cover mapping for the development of GHG inventories Data Acquisition Malawi Landsat Imagery Acquisition Landsat TM and ETM scenes were acquired from the USGS site and preprocessed by Geospatial Data Analysis Corporation (GDA) before delivery to the RCMRD through ICFI. The scenes were delivered as ERDAS Imagine compressed file format of single bands of surface reflectance, Ortho-rectified and Gap-filled products for 2000, 2005 and 2010 epochs. RCMRD-SERVIR Africa downloaded images from the USGS site for gap filling cloud cover masked areas and replacing striped images in the 2010 epoch and images with over 20% cloud in all the epochs. The 1990 imagery was entirely acquired from the USGS site and processed by RCMRD-SERVIR Africa Ancillary Data Acquisition Ancillary data collection process begins with a workshop held in the countries. During the workshop sessions, the participants discuss current and past projects, and relevant data and reports generated that can be availed for purposes of GHG Inventories. After the workshop sessions, ancillary data committed to the GHG land cover mapping by the participants is collected through office visits Data Quality Assessment Landsat Image Quality Assessment Common quality checks in remotely sensed images are based on: radiometric, atmospheric and geometric integrity. Using both Visual checks and statistical methods the quality levels for the imagery are established and used as a basis for image selection for classification. This can be achieved using both manual and automatic means. Radiometric quality is influenced by sensor characteristics and detector responses and Page 22 of 110

24 includes striping, drop lines, other noise elements and missing bands in image data sets. Atmospheric quality is dependent on image acquisition conditions and includes haze, cloud cover etc. Geometric quality is either dependent on sensor characteristics and also satellite situation such as attitude, position, velocity and perturbations. Earth s surface relief is another important factor affecting Geometric Quality of the images. Compromise in geometric quality is illustrated by quality elements/indicators as; lack of band to band co-registration and image to map co-registration. The quality elements vary in different sensors due to different observation techniques. Therefore, QA and subsequent QC of images acquired by different sensors should be processed independently. Guidelines on QA with respect to the QC are discussed in the following sections. a) Radiometric QA Radiometric ability of an image acquired by a sensor generally describes the detail of an image which is represented by grey levels. Radiometric QA guidelines are summarized in Table 1. Table 1: Radiometric quality defects Error Visual Checking Statistical Checking Stripping Drop line Noise (systematic and/or random) Different overall brightness of adjacent lines/rows of the image Null rows in the image; basically a row of zeros. Dark and bright points at the background Significantly different variance and mean of adjacent rows in a band. Zero variance of row in a band Radiometric anomalies (unsteady variances of rows in case of random noise and systematic variance pattern for systematic noise). Missing bands Lack of data in a band Zero variance of a band The above quality defects can be improved as shown in the table below for Radiometric QC Page 23 of 110

25 Error Manual QC Automatic QC Striping Problem was significant in the 2005 and 2010 epochs Drop line Problem was not significant Missing bands Problem was not significant Equalize mean and variance of adjacent rows of the image. Alternatively, duplicate a striped row with an adjacent better row. Replacement of adjacent rows affected by their neighbours. _ Filtering Adopt an average of two neighbouring scan lines/rows. Adopt alternative bands with the same information. b) Atmospheric QA Tables below show procedures for QA/QC of atmospheric effects such as haze and cloud cover. Atmospheric effect Visual Checking Cloud cover White cotton shaped aggregated see figure below. Haze Ambiguous and unusually bright image see figure below. Statistical checking High visible, reflectance and low brightness Temperature (in Thermal band). Unusually high variance in other bands and Compressed and shifted histogram. Cloud cover Haze Striped Imagery Page 24 of 110

26 In the case that alternative images of the area of interest are not available, the QC guidelines can be considered: Atmospheric effect Manual QC Automatic QC Cloud cover Problem was significant Haze Problem was not significant Cloud masking. Mask the affected area with an alternate clear image. This was carried out on a number images Apply Enhancements techniques. The best alternative is to mask using an alternative image of the same season. Was not carried out Cloud masking by clustering and using thresholds i.e. standard deviation and mean Was not carried out Apply histogram transformations. Haze directly affects the histogram of an image (shift and compression), histogram transformation can eliminate it. Was not carried out c) Geometric QA The geometric QA procedure is used to determine and validate the relationship between terrain points and their positions in Landsat images. It provides the necessary information on the possibilities of integrating images and other spatially referenced data sources. Therefore, geometric QA are essential in verifying that no distortions or shifts exist. Common geometric QC is summarized below: Geometric QA Geometric effects Visual Checking Statistical checking Band to band registration Problem was not significant Image to map registration Problem was not significant Geometric QC Image overlay mismatch Page 25 of 110 High matching residuals Geometric effects Visual Checking Statistical checking Band to band registration Image to map registration Rectification/georeferencing using manually selected Ground Control Points (GCPs) Rectification by template matching method.

27 Quality Assessment of Landsat Imagery The figure below shows a section of the compiled image quality report. Figure 1: Section Image Quality Report Ancillary Datasets Quality Assessment Ancillary datasets collected from the countries were assessed with respect to various quality standards. The data was used to aid image interpretation and description of classes and to populate metadata. Since this kind of data are gathered from different institutional bodies, QA is necessary to evaluate the fitness for use of the data. QA of GIS ancillary datasets involves checking for consistency/defects using five elements (qualitative and quantitative): lineage, geometric or positional accuracy, semantic precision or accuracy of attributes, completeness and logical consistency. These elements serve as a basis to determine the fitness of the data for this particular application in image interpretation (Morrison, 1995). Page 26 of 110

28 a) Lineage Lineage is the history of a geographic data set. This is a description of the source material from which the data were derived, and the methods of derivation, including all transformations involved in the production process. This information is important in assessing whether the dataset is fit for with respect to all transformations it has gone through. Most of the datasets collected from the countries did not have metadata enough to describe the lineage properly. b) Positional accuracy Positional accuracy is the accuracy of coordinate values of the dataset. It gives the level of conformity of data with respect to the nominal terrain (perceived true positions on the ground). It thus defines the deviation in the values of the respective positions between the database data and the nominal terrain. Such information assists in deciding whether datasets are within tolerable accuracy limits of the application. Most of the data collected was in vector format. c) Logical consistency Logical consistency is defined as the fidelity of relationships encoded in the data structure. It describes the correspondences of the dataset with the characteristics of the structure of the model used (respecting specified integrity constraints). Though this was not a problem in most data sets collected from the country, topological consistency can be evaluated based on the following: i. Valid values, graphic data, topological, date ii. Geometric-, semantic- and topological consistency. iii. Conceptual-, domain-, format- and topological consistency. Normally a QA can be conducted using GIS software like Arc Map using topology tool as illustrated below: Page 27 of 110

29 Topology tool in Arc Map d) Attribute accuracy Attribute accuracy provides information on the difference between the values of nonspatial attributes and their real value and thus gives us the deviations of measurements of qualitative attributes or quantitative attributes (classification). The data collected from the country did not have significant attribute accuracy problems e) Completeness Completeness is a measure of the absence of data and the presence of excess data. Though exhaustiveness concerning spatial and attribute properties is important, of much concern was the spatial aspect in regard to country coverage. The data was divided to three levels of coverage; Level I (National), Level II (Regional) and Level III (local). The local level data sets were only useful in interpreting minor and very fragmented categories Malawi Data Management The data is managed using a folder system. The hierarchal folder system starts with Country Name, Epoch/Time slice, Footprint Number, Data type (Surface Reflectance, Gap-filled, Ortho-rectified) and year in that order. Every footprint folder contains a selected folder with Page 28 of 110

30 selected images, classifications, Regions of interest (ROIs) and masks. In the effort of enhancing data storage and archiving, SERVIR Africa Geospatial Data Specialist developed a geo-database (GHG GBD Tool) for the GHG project data in ArcGIS model. This tool facilitates structuring and filling of GHG image processing products and improves documentation, search and sharing of GHG outputs with other interested parties. The tool is used to create file geo-databases based on the four main outputs from GHG image processing: Gap-filled, Ortho-rectified, Surface reflectance and selected. Different raster catalog files are added to the file Geo-databases based on the imagery dates (Years) and eras e.g. Yr_1999, Yr_2000 and Yr_2001 representing 2000 epoch (era). The raster catalog files are thereafter populated with imagery datasets from specific source folder files. The file Geo-databases are finally compressed and stored within their respective folders files. The figures below show the design of the Geo-database. Figure 2: GHG Geo-database tool description Page 29 of 110

31 Figure 3: GHG GDB graphic model. Figure 4: Raster Catalog files containing various datasets Page 30 of 110

32 SECTION 4: IMAGE PROCESSING AND CLASSIFICATION PROCEDURES Accurate and reliable information about land areas and area changes is critically important for developing inventories that are consistent with good practices as defined in the IPCC Guidelines. In this context remote sensing imagery represents a cost effective tool for representing areas for the GHG inventory compilations, especially with the global coverage of satellite imagery obtained from Landsat. This section presents a series of technical guidelines for land cover classification from the Landsat imagery data as part of the GHG project analysis and capacity building of the project countries. In this section, there is brief discussion of the underlying theory of image processing, classification and description of the operational details of the land cover mapping. The processes of integrating remote sensing techniques and field data to accurately map land use and land cover is described. Satellite images of the epochs 1990, 2000 and 2010 are used in this guide to illustrate image processing and classification steps followed as carried out in ENVI software. This section is intended to provide technical guidance on image processing and automatic (digital) classification for GHG inventory development projects and for the purposes of producing GHG inventory reports every two years. Land cover in the ESA countries is greatly fragmented unlike the developed world where proper zoning of various land uses is effectively implemented. There are slight modifications expected in the procedures to suit specific country conditions, interests, classification schemes and/or criteria and choice of categories Image processing After image acquisition, the following processing steps are carried out before image classification; layer stacking, selection of cloud free scene, computation of image statistics, image selection, band combination and finally image enhancement as discussed below. Figure 5 below is a summary of all the steps carried out from image acquisition to the final classification product. Page 31 of 110

33 Step I: Image Acquisition Dry season Un-stacked Bands Gap-filled, Ortho-rectified /Surface Reflectance products Step VII: Complete Product Compile metadata Produce documentations and manuals Distribute results via database custodians Step II: Image Processing Layer stacking Image Quality Assessment Image Enhancement Cloud Masking Step V: Accuracy Assessment Decide sampling intensity by class Use ancillary Data Publish Accuracy Assessment Step III: Image classification Spectral training Spectral Integrity Checks Class evaluation classification Step VI: Field Checking Verify draft land cover classes Supply recommended edits to draft classification Get additional ground reference data Step IV: Classification Editing Classification iterations to minimize errors Combine classes Code classes Mosaicking of Classified scenes Figure 5 Main Land Cover Mapping steps Layer stacking Using ERDAS Imagine remote sensing software, the single bands are layer stacked. The process of layer stacking involves selecting individual bands and compositing them to form a multi band image. In the GHG project both ERDAS Imagine and ENVI software s were used in layer stacking. The significance of layer stacking is to have a multi-spectral image that can be classified since DN values of various bands increase variability between features thus enhancing discrimination of specific classes or categories. Layer stacking is done for six bands excluding the thermal and panchromatic bands since a Page 32 of 110

34 wide range of features is being classified. The following steps illustrate the procedure for layer stacking. a) Identify the single bands for the footprint to be layer stacked. Single bands in folders ready for layer stacking b) Start layer stacking procedure in ERDAS Imagine as shown below; Layer stacking procedure in ERDAS Imagine Page 33 of 110

35 Images in individual bands are layer stacked and ready for conversion into a geo-tiff (displayed in band 4, 3, and 2 below) Single bands before classification Layer stacked image c) The layer stacked image file is converted to a Geo-tiff file that can be used for visualization in most GIS software including open source for quality assessments Cloud Masking The major challenge encountered is the high cloud cover on the images selected for classification even on the gap filled product. In an effort to reduce the cloud cover percentage, Landsat data from the USGS site was used for filling gaps in the cloud masked images. The figure below shows an Ortho-rectified scene used to produce a gap filled product. Page 34 of 110

36 Ortho rectified image with high cloud cover Gap filled image with less cloud cover The remaining cloud cover in the gap filled images is masked out using ENVI software and mosaicked with another image that is cloud free in the same area. The Process of masking out and filling in cloud covered areas involves 3-steps and requires two images; Image-1 with cloud areas to be masked out and Image 2 with corresponding cloud-free areas for gap filling on the selected image for classification. The figures below show an example of a cloud masking and gap filling effort used in the project to further reduce cloud cover in the classified images. Image with cloud cover and image showing areas of interest for gap filing Page 35 of 110

37 Cloud masked image and cloud free image subset for gap filling image Cloud gap-filled Computing Image statistics The Ortho-rectified images with little or no cloud cover are selected for classification and subjected to general statistical computation using ENVI software. The Statistics option is used to generate statistical reports and display plots of histograms, mean spectra, eigenvalues, and other statistic information for image files. Calculation of basic statistics and/or tabulated histogram information (frequency distributions) for singleband or multi-band images is carried out to check the spectral quality of imagery. The minimum, maximum, and mean spectra can only be calculated for multi-band images. Similarly, covariance statistics, which include eigenvectors and a correlation matrix, can be calculated for multi-band images. The statistics are calculated in double-precision as shown below. Page 36 of 110

38 Image without clouds and with ROI for masking Zeros Page 37 of 110

39 . The statistics are significant in enabling one to ensure that there is sufficient variability among spectral bands. If there is no variability between the bands the image is rejected. The significance of band variability in classification is to give indication of ability to discriminate various classes during classification. If an image also has 2 or 3 bands that are highly correlated, one of them is selected for classification; this reduces on image size and redundancy during classification. Figure 6: The Compute Statistics Parameters Dialog The figure below shows a sample of computed results. Page 38 of 110

40 Figure 7: Sample of computed results Sub-Setting Imagery to Country Boundary The sub-setting of the image to the specific area of interests ensures that the ROI selected are picked from within the boundary of the country of interest. This is important as it improves the accuracy of the image during classification and saves on time. Using the country boundary with a 10 km buffer the image is reduced to the area of interest specific to that country whose ancillary data is available for interpretation Band compositing and selection Remote sensing sensors (Landsat) record the relative brightness of an area over specific portions of the electromagnetic spectrum. All sensors have spectral sensitivity limitations; this is referred to as spectral resolution. No single sensor is sensitive to all wavelengths of the electromagnetic spectrum. Recorded wavelengths are referred to as bands. Displaying of a remote sensing image on a computer monitor is limited to 3 bands. Selected bands are shown consecutively through the three color monitor guns (red, green, and blue). Band combinations are dependent upon the type of feature Page 39 of 110

41 analysis being performed. The table below shows Landsat band combinations. The highlighted composites were used in identifying and selecting feature training areas for supervised classification for the GHG project; Landsat TM Band Combinations Red en Blue Feature Screen color Bare Soil Magenta/Lavender/Pink Cropland Green Urban Areas Lavender For crop lands and Shrub lands Wetland Vegetation Green Trees Green Bare Soil White/Light Grey Cropland Wetland Vegetation Medium-Light Green Dark Green/Black Trees Olive Green Bare Soil Blue/Grey Cropland Pink/Red For forest density l Wetland etation Dark Red Trees Red Bare Soil Green/Dark Blue Cropland Urban Areas Wetland Vegetation Trees Yellow/Tan White/Blue Brown Tan/Orange Brown Extraction of LULUCF Using Landsat 4.2. Image Classification Image classification is the categorization of all the pixels in a digital image into one of several land cover classes, or themes using algorithms. This categorized data can then be used to produce thematic maps of the land cover present in an image. The term land cover originally referred to the kind and state of vegetation, such as forest or grass cover, but it has broadened in subsequent usage to include human structures such as buildings or pavement and other aspects of the natural environment, such as soil type, biodiversity, and Page 40 of 110

42 surface and groundwater (Meyer, 2008). Generally, land cover is what covers the surface of the earth (Nerd, 2004). Classification in remote sensing involves clustering the pixels of an image to a (relatively small) set of classes, such that pixels in the same class are having similar properties. The majority of image classification is based on the detection of the spectral response patterns of land cover classes. Classification depends on distinctive signatures for the land cover classes in the band set being used, and the ability to reliably distinguish these signatures from other spectral response patterns that may be present (Eastman, 2003). There are many different approaches to classifying remotely sensed data. However, in common they all fall under two main topics: unsupervised and supervised classification technique. The primary distinction between these two approaches is the amount of information the operator has about the images. In a supervised classification, the user generally has some verification or training data, meaning that the user generally knows what the features on the image represent. In unsupervised classification, an algorithm is chosen that will take a remotely sensed data set and find a pre-specified number of statistical clusters in multispectral space. Although these clusters are not always equivalent to actual classes of land cover, this method is used without having prior knowledge of the ground cover in the study site (Nie et al., 2001). In supervised classification, the multispectral data from the pixels in the sample area or spectral signatures from spectral library are used to train a classification algorithm (Kamaruzaman et al., 2009). Once trained, the algorithm can then be applied to the entire image and a final classification image is obtained. In this project both unsupervised and supervised classification approaches were used Classification Scheme Selection One of the most important and difficult steps in planning a land cover classification project is selection of the categories to be discriminated in the mapping effort. The classification scheme should be compatible with the existing systems and yet represent the local land cover characteristics. Selecting the appropriate levels of categorical detail is also important. Choosing an over-abundance of categories can lead to considerable Page 41 of 110

43 confusion among cover types, whereas selecting too few classes may not meet the user needs. Numerous existing classification schemes were studied to help guide the structure and categorical detail of the GHG scheme and most importantly, the classification scheme used was informed by country specific interest, definitions, descriptions, mapping goals and policy statements. Globally acceptable legends including FAO LCCS were studied (Anderson et al 1972). As discussed earlier the IPCC guidelines played an overarching role in developing the classification scheme that meets the country specific mapping standards and that can be rolled back to the IPCC categories. The following are the IPCC categories with their respective codes as shown. a) Schema I: Six IPCC Classes as the bare minimum tier 1, (All sub-categories gets rolled up to the bare minimum) i) 10 - Forestland ii) 30 Grassland iii) 40 Cropland iv) 50 Wetland v) 60 Settlement vi) 70 0therland vii) 255 No Data The sub-categories are agreed upon by the countries in the consultative forums in the workshop sessions and are based on either the government policies or on previous classification in the event of lack or inadequate policy provisions. The following are the most common sub-categories used to create level II discriminations; b) Schema II: national sub classification categories agreed upon with countries; 12 to 15 classes, and keeping them consistent with IPCC guidelines (as required). This level II discrimination creates a reference classification that Page 42 of 110

44 can be subdivided further to level III or IV and beyond or as required; using good ancillary data (shape files) with application of binary classifiers. i) 10 - Forestland 1) 01 Dense Forest 2) 02 Moderate Forest 3) 03 Sparse Forest 4) 04 Planted Forest ii) 30 Grassland 5) 31 Closed Shrubland 6) 32 Open Shrubland 7) 33 Closed Grassland 8) 34 Open Grassland iii) 40 Cropland 9) 41 Perennial Cropland 10) 42 Annual Cropland iv) 50 Wetland 11) 51 Vegetated Wetland 12) 52 Water Bodies v) 60 Settlement 13) 61 Settlement vi) 70 Otherland 14) 71 Otherland vii) 255 No data viii) 81 Clouds ix) 82 Shadow Important Classification Factors A land use and land cover classification system which can effectively employ orbital and high-altitude remote sensor data should meet the following criteria as considered in this guide (Anderson, 1971): Page 43 of 110

45 i) The minimum level of interpretation accuracy in the identification of land use and land cover categories from remote sensor data should be at least 75 percent. ii) The accuracy of interpretation for the several categories should be about equal. iii) Repeatable or repetitive results should be obtainable from one interpreter to another and from one time of sensing to another. iv) The classification system should be applicable over extensive areas. v) The classification system should be suitable for use with remote sensor data obtained at different times of the year. vi) Effective use of subcategories that can be obtained from ground surveys or from the use of larger scale or enhanced remote sensor data should be possible. vii) Aggregation of categories must be possible. viii) Comparison with future land use data should be possible. ix) Some of these criteria should apply to land use and land cover classification in general x) Criteria apply primarily to land use and land cover data interpreted from remote sensor data. xi) Multiple uses of land should be recognized when possible Classification methods Landsat scenes are classified individually and merged after classification. A hybrid of supervised and post classification methods are employed used in the classification as discussed below. The figure below provides a workflow of image classification Malawi scenes (classification workflow for processing Landsat TM scenes). Page 44 of 110

46 Figure 8: Work flow for Image Classification a) Image Interpretation using Ancillary Data and Google Earth The ancillary data is used in image interpretation to help in selection of training areas. This is done by performing super-overlays of the selected ancillary data and the Landsat data on Google Earth. The ancillary data is also overlain on the imagery in the GIS software to interpret the various features on the images. Using the very high resolution imagery on Google earth, it is possible to check various features that are not clear on the Landsat imagery for interpretation b) Vegetation Delineation Tool The Vegetation Delineation tool enables quick identification of the presence of forest in its sub categories of Dense, Moderate and Sparse forest and to visualize its levels of vigor. This tool was used for delineating forest sub-categories in terms of dense, moderate and sparse categories. Page 45 of 110

47 Figure 9: Vegetation Delineation Tool Delineation of the different forest classes uses NDVI, and Blue and Green bands signature analysis. The tool analyses spectral signature to produce an image as shown below for the three types of forest; dense (above 70%), moderate (40 less than 70%) and sparse (10 less than 40); This vegetation delineation tool is used to standardize selection of training areas for various forest density classes without reducing wide variation among image interpreters. Page 46 of 110

48 c) Separability Analysis Jeffries Matusita distance for the separation of the six IPCC and 15 class schema categories. Separability analysis is carried out to ensure that all classes are separable from each other during classification. This process involves loading an image in ENVI Software, picking training areas for each class and computing a separability test, displayed in text file. Figure 10: Separability Analysis Window Separability becomes difficult where gross similarities in spectral properties exist. If the Separability test gives a factor below 1 then, Separability of the two classes is insufficient. d) Classification Using Maximum Likelihood Maximum likelihood is a statistical decision rule for examining the probability function of a pixel for each of the classes, and assigning the pixel to the class with the highest probability. The classifier assumes that the training data for each class has a normal or 'Gaussian' distribution. The classifier uses training data statistics to compute a probability value of whether it belongs to a particular land cover category/class. The Maximum Likelihood algorithm is the most commonly used supervised classification technique in remotely sensed image classification (Landgrebe, 1980; Michelson et al., 2000; Richards, 1999; Tso and Mather, 1999). Page 47 of 110

49 Maximum likelihood classifier was run using the selected training polygons. This technique is a per-pixel classifier, and therefore assigns a class value to each pixel based on its individual spectral response pattern. In the supervised classification method as mentioned earlier, images are classified using spectral signatures or training area/aoi/roi (i.e., reflectance values) obtained from training samples (polygons that represent distinct sample areas of the different land cover types to be classified). These samples are collected by the image analyst and maximum likelihood is used to build the reference land-cover classification in ENVI Software. e) Regions of Interest for Image Training These are created by selection of the various training sites depending on the class identified on the image. The figure below shows ROI selected for the class water. Figure 11: Selected training areas for water bodies Page 48 of 110

50 f) Chain Classification Chain classification involves selecting common ROIs on two overlapping scenes for classification to reduce edge difference between them. Alternatively ROIs of the two images can be harmonized and applied for classification preceding image in order to ensure uniformity and flow among the scenes (chain classification). g) Samples of Training Areas 1) very dense 2) Dense 432 RGB 432RGB 3) Moderately Dense 4) Sparse forest 5) Water RGB 6) Wetlands 1 432RGB RGB742 RGB742 7) wetlands 2 8) Wetland Vegetation1 Page 49 of 110

51 RGB742 RGB742 9) Wetland Vegetation 2 10) Cropland 1 RGB742 RGB742 11) Cropland 2 12) Cropland 3 RGB742 RGB742 Page 50 of 110

52 13) Annual Crops 13) Cropland Perennial 1 14) Cropland Perennial 2 RGB742 RGB742 15) Cropland Perennial 3 RGB742 RGB742 16) Cropland Perennial 4 17) Cropland Perennial 5 RGB742 RGB742 18) Cropland Perennial 6 19) Shrub lands (Closed) RGB742 RGB742 Page 51 of 110

53 20) Open shrub land 21) Other lands 1 RGB742 RGB742 22) Grasslands Closed 23) Grasslands Open RGB742 RGB742 24) Clouds clouds 1 25) clouds 2 RGB742 26) Shadow RGB742 27) Woodlands RGB742 RGB742 Page 52 of 110

54 28) Water a. Water 1 29) Water 2 RGB742 31) Water 3 RGB742 30) 32) Water 4 RGB742 33) Irrigated Land 1 RGB742 34) Water 6 RGB742 RGB742 Page 53 of 110

55 After the ROI are selected, the classification is run based on the maximum likelihood classification statistic as shown in the figure below. Page 54 of 110

56 Correction of misclassified classes involves re-running the classification several times and/or moving to the post classification processing where various processes are used to improve the classification accuracy Post classification processing Quality Assessment of Classification Using Landsat Imagery This is an intermediary step that is performed to ensure that all features are well represented in the classification output. It guides in determining the need for further Iterations. The classification is overlaid on the Landsat Image and using the swipe to the assessment is made. The figure below illustrates a check of this nature using ERDAS software. A similar check can be applied using effects tool in ArcGIS. Figure 12: Check of classification against Landsat Imagery Quality Assessment of Classification Using Google Earth Classified imagery is converted to a.kmz file and opened in Google Earth. Using the Adjust Opacity bar the classification is assessed against high resolution imagery. Checks on Google Earth take into account temporal differences between the Landsat imagery and a contemporary high resolution imagery being used. Page 55 of 110

57 ROI Merge Upon satisfaction with quality of classification then all ROI s of a particular Land cover are merged. The figure below illustrates this process. Figure 13: Illustrating ROI Merge The classes of Annual Cropland based on the varying reflectances of this cover category are now being merged to one class Decision tree Coding of the classified images was done using the decision tree. A decision tree is a type of multistage classifier that can be applied to a single image or a stack of images. It is made up of a series of binary decisions that are used to determine the correct category for each pixel. Decision tree also enables the correction of classification results. The classifier was used to delineate between the classes and features that had similar spectral signature where MLC identified them as one class e.g. opens shrubs and closed shrubs Class coding and Color Matching Coding land cover and color matching of land cover categories is achieved using a decision tree. To interactively execute and edit the decision tree, ENVI software is used. Page 56 of 110

58 This is possible if a decision tree is already created and the image or bands are georeferenced. The classified image or band is specified as a variable and various classes are assigned values. The figure below shows the decision tree used for coding land cover classes. Figure 14: Land Cover Coding Tree Classification of Settlements There is very limited separability between bare areas and settlements when using Landsat imagery. This makes it difficult to run automatic classification using training sites for settlements. Using ancillary data for towns and Google Earth, the settlements were delineated to create shape files that show major settlements in Malawi, excluding the major forest areas within the urban areas. The shape files are rasterized and built into the land cover using decision tree as shown in the figure below. The use of the decision tree ensures exclusion of the major forest areas, wetlands and water bodies within the outline of the Settlement area. Page 57 of 110

59 Decision Tree for Intergrating settlements to Land Cover Classification Figure 15: Land cover map with settlements in grey Edge matching procedure This involved using feathering distance to reduce edges along the overlap areas between the two classes. The feathering distance was determined by measuring the overlap area between the two classifications of different images. Once this is done, the feathering distance is applied during mosaicking in order to remove edges. Page 58 of 110

60 Finally; once the post classification processes have been done, the classification is ready for accuracy assessment. Page 59 of 110

61 SECTION 5: GROUND REFERENCING AND ACCURACY ASSESSMENT Sampling in conjunction with ground reference surveys is one of the most popular approaches to data collection. Sampling is choosing which subjects to measure in a research project, while ground referencing is finding and measuring the subjects in question (Garson, 2012). Ground reference or ground truth data is collected to train the computer to recognize the various land cover categories latent in the imagery and to assess the categorical accuracy of the resulting classification. Ground reference data generally cannot be collected for large portions of the entire project area therefore; representative samples are frequently used (Lille sand and Kiefer 1994). Our ground referencing activity was limited in time and resources and this led us into considering an alternative approach for accuracy assessment. The second approach considers independent interpretation of Landsat image (Stehman 2001). This unbiased approach takes care of temporal differences between the images used in classification and time the field work is carried out Sample design for Ground Referencing The sampling design is a procedure for selecting the locations at which the reference data are obtained. A probability sampling design is the preferred approach and typically combines random or systematic stratified sampling with cluster sampling (depending on the spatial correlation and the cost of the observations). Estimators should be constructed following the principle of consistent estimation, and the sampling strategy should produce accuracy estimators with adequate precision. The design-based sample is used to define the sample size, sample locations and the reference assessment units (i.e. pixels or image blocks). Stratification should be applied in case of rare classes (i.e. for change categories) and to reflect and account for relevant gradients (i.e. eco-regions) or known factors influencing the accuracy of the mapping process. In this Guide ground reference data is collected from the field on randomly generated points at selected zones for image classification accuracy assessment. According to Lille Page 60 of 110

62 sand, 1998, several criteria must be considered when evaluating the suitability of any ground reference data set for land cover classification; The data collection method should be systematic, that is representative of the entire area that has been classified The method must have an element of randomness to avoid selection bias A sufficient number of reference samples must be utilized to provide an appropriate sample density and ensure that the classification accuracy is known within a specified confidence level The reference data must be reasonably contemporary with respect to the acquisition date of the imagery The level of accuracy of the reference data must be high To meet the above criteria, therefore, the specific tasks for this activity explicitly discussed herein include; i) Definition of a representative sampling zone ii) Develop a sampling design and Generate the Sampling frame iii) Collect ground reference data iv) Implement simple QA/QC procedures on ground reference data 5.2. Applied Sampling Methodology Two important factors are taken into consideration when delineating sampling zone areas; a) Proximity of the sampling zone to potential stop-over sites This entails first, locating the potential stop over sites for accommodation for the field team as well as provision of vital utilities like electricity for charging field equipment. District Headquarters are used for this. The sampling zone is then defined to fall in the area not less than 5km from the District Headquarters and also not further than 15km from the same towns. This is to avoid data collection contamination and the effects of urbanization on the Landscape, while at the same time avoiding excessive traveling. NB. The Country s capital, major cities and their vicinities are completely left out from this. Procedure: Page 61 of 110

63 i. District Headquarters shape file is loaded onto ESRI ArcGIS 10 ii. Using the 'multiple ring buffer' tool in Arc Toolbox > Analysis tools, a buffer is done to cover the areas not less than 5km from the District Headquarters and also not further than 15km from the same towns. Figure 16: Multiple ring buffer tools in Arc Tool box Figure 15 below shows a zoomed in part of Malawi showing areas (in purple) around the District Headquarters that qualify to be part of the sampling zone based on this factor. Page 62 of 110

64 Figure 17: District Headquarter sampling zone b) Accessibility of the Sampling Zone The available official road layer is used. Bearing in mind the limited time frame allocated for the field exercise, the sampling zone is defined not less than 200m from the main roads to avoid contamination of data and also not further than 3km from the same roads, to minimize travelling time. Procedure: i. Select only the main roads from the available official road network shape file using the 'select by attribute' tool in the selection tab in ArcGIS and convert it into a new shape file. Page 63 of 110

65 Major Roads ii. Applying the 'multiple ring buffer' tool again, a buffer is done to cover areas not less than 200m from the main road also not further than 3km for the same roads. Fig.18. shows a zoomed in part of Malawi showing areas (in blue) along the major roads that qualify to be part of the sampling zone based on factor 2. Page 64 of 110

66 Figure 18: Probable Sampling Zone along Major Roads iii. Finally, using the 'intersect tool' in the Arc Toolbox > Analysis tools (Fig.18. below), an intersection is done on the two final defined sampling zones based on factor 1 (Towns) and factor 2 (roads), to only remain with areas that meet these two important considerations. Intersect tool in Arc Toolbox Fig.19. shows a zoomed in part of Malawi showing the final sampling zone (areas in pink), while Fig.20, shows the sampling zone (areas in deep green) for the entire country of Malawi Page 65 of 110

67 Figure 19: Defined sampling zones (in pink) Page 66 of 110

68 Figure 20: Defined sampling zones (in green) Page 67 of 110

69 Proportionate Stratified Random Sampling Procedure a) Clipping/sub-setting of classified image; i) An already classified image for 2010 epoch of the area/country in question is loaded onto ENVI 4.7 ii) Import the generated sampling zone in shape file (.shp) format defined in Task 1 above as a vector file using the tool, file > import vector file. Set the vector file parameters to suit your area/country of study, e.g. for Malawi; File projection - UTM, Datum WGS84, Units Meters, Zone 36 S (See Fig.9. below) Figure 21 Sub-setting vector file parameters iii) Convert the vector layer to ROI format. In the vector parameters window that appears, Go to, File > Export Active Layer to ROIs. Page 68 of 110

70 Convert Vector to ROI Figure 22: Sampling zone ROIs overlaid on classified image i) The resulting ROIs can now be used to clip the classified image. In ENVI go to, Basic Tools > Subset Data via ROIs. You ll be prompted set the Input File to subset via ROI. Choose the Classified image that you want to clip. Page 69 of 110

71 Figure 23: Subset Data via ROIs Tool ii) Select the input ROIs and set the Mask pixels outside of ROI to Yes. This discards all the areas in the input image that are outside the area covered by the input ROIs. Remember to give the output file a name. Fig.14. below shows a zoomed in part of Malawi showing the clipped out classified image. Figure 24: Sampling zones clip This clip out is used to generate the sampling frame whereby each 30m by 30 m pixel centroid from the clipped classified image is converted into a point that constitutes the sampling population. From this population the sampling frame/size can now be established. Page 70 of 110

72 b) Generating Random Sample According to Congalton, 1991, it is a good rule of thumb to collect a minimum of 50 samples for each vegetation or land use category in the error matrix especially if the land use categories are more than 12. Congalton, 1991, also points out that the number of samples for each category can be adjusted based on the inherent variability within each of the categories. It is useful to take fewer samples in categories that show little variability or cover less area such as water or forest plantations and increase the sampling in the categories that are more variable or cover large areas such as cropped lands or riparian areas (Fitzpatrick-Lins, 1981). Proportionate stratified random sampling method is recommended and chosen to establish the sampling frame to be used for the GHG Inventory Development project. Essentially, random sampling is a data collection method in which every entity in the population has a chance of being selected (Garson, 2012). When a sampling frame is partitioned into groups prior to selecting a sample and a sample is taken from within each group, the groups are called strata and the sampling design is called stratified sampling. A stratified random sample therefore is a sampling plan in which a population is divided into L mutually exclusive and exhaustive strata, and a simple random sample of n h elements is taken within each stratum h. The sampling is then performed independently within each stratum (Cochran, 2007). There are two major subtypes of stratified sampling: proportionate stratified sampling and disproportionate stratified sampling. In proportionate stratified random sampling, the sample size of each stratum is proportionate to the population size of the stratum when viewed against the entire population. This means that each stratum has the same sampling fraction that is; the size of the sample drawn from each stratum is proportional to the relative size of that stratum in the target population. As such, it is a self-weighting and equal probability selection method (EPSEM) sampling procedure. The same sampling fraction is applied to Page 71 of 110

73 each stratum, giving every element in the population an equal chance to be selected (Daniel, 2011). Working with a value of 50 points per land use category generated a total of approximately 750 points in the sampling frame. The proportionate stratified sampling technique was applied which randomly distributed these points across the sampling zone and across each land cover category in relation to their area coverage in the sampling zone. Procedure; i) From the ENVI main menu bar go to, Classification > Post Classification > Generate Random Sample > Using Ground Truth Image. Figure 25: Generate Random Sample tool ii) Select the input file as the clipped classified image generated previously from which to draw the sample from. Page 72 of 110

74 Select input file iii) The class distributions are computed and the Generate Random Sample from Ground Truth dialog appears. NB. It may take several seconds to compute the class distributions especially when sampling from a large classification image. A status reporting dialog will be displayed iv) Select the ground truth classes or ROIs to include in the sampling. Page 73 of 110

75 v) Click OK. The Generate Random Sample Input Parameters dialog appears. Random sample input parameters dialogue vi) Select the sampling type from the drop-down list and set the Stratification type radio button to proportionate. vii) Enter the Minimum Sample Size in percent or pixels by clicking the toggle button. Entering a value for one will automatically update the value for the other, making it easy to see the relationship between the percentage sample size and the pixel sample size. Set the minimum sample size to 1, to ensure that at least one pixel is included from the smallest class e.g. planted forests land cover category viii) To view the proportionate class sample sizes for the current Minimum Sample Size Setting, click view class sample sizes. The Proportionate Sample Sizes dialog must be closed before the Generate Random Sample Input Parameters dialog becomes active again. Page 74 of 110

76 Proportionate sample sizes ix) The total sample size displays to the left of the view class sample sizes button and will update dynamically as new settings are entered. x) The resulting proportionate sample for each land cover class is then proportionately reduced to an approximate predetermined total sample size of approximately 750 points Field Data Collection Form The table below shows the attribute form used for field data collection. This was filled for every point. Table 2: Field data collection Form Field Attribute LC Code UNIQUE_ID SAMPLED_LC POINT_X POINT_Y BENCHMARK OBS_LC CANOPY_H1 CANOPY_H2 CANOPY_H3 CANOPY%_H1 Description Land cover type unique code Tracked Field point unique ID Classified land cover type Longitude Latitude Distance of observed point from tracked point Observed land cover type Height of topmost canopy Height of middle canopy Height of lowest canopy Percentage of Canopy H1 Page 75 of 110

77 CANOPY%_H2 CANOPY%_H3 NOTES N_PICTURE E_PICTURE S_PICTURE W_PICTURE Percentage of Canopy H2 Percentage of Canopy H3 Any relevant site information both current and historical Picture taken to the North of tracked point Picture taken to the East of tracked point Picture taken to the South of tracked point Picture taken to the West of tracked point 5.4. Collection of Ground reference data Field Preparation Preparing for field work is well worth the time used and is crucial to the project's success. It is important to recognize that field work is expensive, and time in the field is limited. In most cases it also may not be possible to go back and get more data, i.e. it can be a one-shot deal (Maher, 2004). If appropriate preparation has been made prior to the field trip, location work in the field will proceed with greater effectiveness and efficiency (McCoy, 2005) Prerequisites Before leaving fieldwork The following are the prerequisites for fieldwork i. Ensure that you do a thorough check on the field points you are to track in the field. Check them against the pre-classified image and make sure the points fall in a homogenous area at least within a 90m x 90m area NB. A homogenous site is an area that has only one type of land cover on it. ii. iii. iv. You are well versed with and can recognize the different land cover types and their descriptions Ability to explicitly use necessary field equipment; the Trimble /GPS, Camera, Compass Ability to interpret and apply the concepts and techniques of the canopy charts Page 76 of 110

78 v. Most importantly, ability to walk long distances if necessary, work in nonconducive weather etc. which necessitates the carrying of appropriate field gear (shoes, raincoats, umbrellas etc.) vi. vii. You have necessary field equipment. Frequent use of the GPS/Trimble receiver instrument before going to the field will familiarize one with its rate of energy consumption. Above all, be sure to carry a good supply of extra new batteries for the GPS/Trimble receiver. Plan for a dummy field exercise near the office a few days before the actual field exercise and standardize the land cover field parameters with colleagues and other experienced staff Materials and Tools needed a) A Trimble kit i) Trimble Charger ii) Trimble batteries iii) Memory cards iv) Important shape files loaded; sampled points, all roads, etc. b) Make copies of necessary maps and documents i) Sampling zone maps ii) Canopy charts iii) classified image map iv) road maps c) Necessary Stationery i) Markers ii) Pencils iii) Erasers iv) rough papers v) clip boards d) Compass Page 77 of 110

79 Tasks at the Field a) Before leaving for the sample site i) Initialize the Trimble GPS Receiver, at the day s first start-up, the Trimble/GPS instrument must determine its current location. This will in most cases take quite some time. Ensure to initialize it well in good time before leaving the Hotel/Hostel ii) Be sure the GPS antenna is properly oriented and unobstructed at all times iii) Create a new Arc pad file and name it after the field work day number i.e. Day 1, Day 2,.Day n) iv) Load the necessary shape files onto the initialized Arcpad v) Track the nearest point in the sampling zone of interest vi) Instruct the driver about the route and direction to take as you monitor the distance to the tracked point of interest vii) Try and navigate the point to the zero mark if accessible. b) At the sample site i) Once you get to the zero mark, and find that it s not a homogenous area (i.e. the 90m x 90m square surrounding the zero mark should have the same vegetation type), move to a representative homogenous area within the area and pick the point as a waypoint. ii) Complete the sample point s shape file attribute table, with the necessary data to be collected as explained in the below steps. Remember time is of the essence so be quick in this. iii) Indicate the observed land cover at the point of interest iv) If the area has vegetation above 5m and is not cropland or a grass land, describe the canopy heights to indicate the layers of canopy observable. Use the canopy charts for this (see figures below) v) Indicate the percentages of the canopies described above and indicate the observed land cover. Indicate any other relevant information in the notes column Page 78 of 110

80 vi) If the area is a cropland indicate if it s a perennial or an annual cropland. Indicate the crop type and any other relevant information in the notes column vii) If the point of interest is a water body, indicate the benchmark distance and pick a waypoint and pictures at the bank. viii) In the notes column note any unusual or helpful metadata. Metadata can provide insight about an area that may not be clear in the image scientists are looking at. Get help from local experts accompanying you in history, plant identification, land cover mapping etc. ix) Take four pictures in clockwise direction starting with the North, East, South and West (Cardinal directions). Make sure your camera is held upright and focus on a wide observable area. Avoid any obstructions at your line of focus e.g. fingers x) Save the collected information xi) Track the next point of interest Trimble Juno SB fieldwork methods The Trimble calculates its exact position is based on triangulation of radio signals received from different satellites in the orbit. The Trimble Juno unit is therefore simply a radio receiver set to listen to signals coming from GPS satellites. To calculate an exact position, the GPS unit needs to be receiving a clear signal from at least 4 well positioned satellites at once. The Trimble Juno SB works very well also under tree canopy and other obstructions such as within buildings and cars. However, for best reception one should make sure there is as little obstructions between the device and the sky as possible. It is very easy to confirm the position accuracy from the Trimble. The satellite window will show the position accuracy on the screen. It has been empirically noted that accuracies of ±3 meters (meaning a circle with a diameter of 6 meters) are possible with this unit. If you cannot get a better signal: i) Try moving slightly ii) Point the antenna upwards, in a 45 angle iii) Point the device away from yourself as you might be the obstruction iv) If in a car or indoors, move closer to a window Page 79 of 110

81 v) Wait for some time the satellites are constantly moving and a better constellation take a while to achieve. vi) Use the average function when taking a point (takes several points and calculates the average, which will usually result in a better accuracy) Collecting points of interest (POI) Hold the GPS at the point of interest (POI) and activate form for editing and carefully complete the attribute form. If not at the tracked point take waypoint. The waypoint attribute form will open for editing. This page allows you to insert data about the point. Please note: that before saving the point information the error should be ± 3m Independent Validation Point Interpretation for Accuracy Assessment Ground referencing efforts was limited both in time and resources leading to collection of sample field points based on the criteria used around towns: between 5-15 km from towns and not more than 3 Km from the main roads. The deficit in this approach of data collection is that features not falling within our criteria were not sampled leading to biasedness and skewedness. This led us to considering a different approach for accuracy assessment that had distribution points across the entire country. Classification methodology employed, supervised classification, being an automatic classification requires a mechanism for giving feedback on the level/quality of output. The approach used required that stratified random points are generated within the boundary of the country; 1,000 points were generated for Botswana. A buffer of 30m is generated from the points; the polygons are then overlaid on Landsat images and interpreted independently. Polygons falling within overlapping land covers are moved to more homogenous locations. Centroids of these polygons are combined with selected ground referencing points (550 points) for accuracy assessment. Figure below shows points generated for Botswana for this purpose. The land Cover Column gives the land cover interpretation of the point. Page 80 of 110

82 Figure 26: Random Generated Validation points for Accuracy Assessment 5.7. Accuracy Assessment Procedures Ground Referencing points are split two ways. Some of the points are used in refining the classification and the remaining points combined with validation points from independent interpretation for accuracy assessment. The accuracy assessment of the developed map will be done through the construction and analysis of an error matrix and an overall accuracy of not less than 75% (according to USGS classification) will be accepted. Page 81 of 110

83 i) The classified image and the points for Accuracy checks are opened in ENVI. Open the vector file with the ground referencing sample points and load them to the display with the classified image ii) The vector file is then exported as an ROI Dialogue for exporting vector file to ROI Page 82 of 110

84 i) ROIs of the same Land Cover are merged as shown below. Confusion Matrix Dialogue for Merging ROI s of the same Land Cover ROI Confusion matrix is developed from the points to be used in accuracy assessment. The figures below illustrate the stepwise procedure for this. From ENVI main menu open Classification Post Classification Confusion Matrix Using Ground Truth ROIs, open the confusion matrix tool. Match the class parameters on the ground truth ROI and the classified image and run the Confusion Matrix. Page 83 of 110

85 Page 84 of 110

86 SECTION 6: RESULTS The assessment indicates that the overall accuracy of the 2010 land cover data is 84.01% with a Kappa coefficient of for Scheme I and 77.0% with a Kappa coefficient of for Scheme II. On the Other hand that for the 2000 epoch is % with a Kappa coefficient of for Scheme I and % with a Kappa coefficient of Scheme II. The overall accuracy for 1990 scheme I is 87.76% with a Kappa coefficient of while that of scheme II is % with a Kappa coefficient of Accuracy Statistics for 2010 Classification Scheme I Overall Classification Accuracy = (1461/1739) 84.01% Overall Kappa Statistics = Accuracy Summary Class Name Forestland Grassland Cropland Wetland Settlement Otherland Reference Totals Classified Totals Number Correct Producers Accuracy Users Accuracy % 81.69% % 61.35% % 85.87% % 98.71% % 88.61% % 69.23% Totals Overall Classification Accuracy = 84.01% Page 85 of 110

87 Error Matrix Classified Data Forestland Grassland Cropland Wetland Settlement Otherland Column Total Reference Data Total Forestland Grassland Cropland Wetland Settlement Otherland Accuracy Statistics for 2010 Classification Scheme II Overall Classification Accuracy = (1340/1739) 77.0% Overall Kappa Statistics = Accuracy Summary Reference Class Name Total Dense Forest 72 Moderate Forest 160 Sparse Forest 229 Open Grassland 17 Open Shrubland 107 Closed Grassland 14 Closed Shrubland 57 Perennial Cropland 49 Annual Cropland 619 Water Body 68 Classified Totals Number Correct Producers Accuracy Users Accuracy 79.17% 73.08% 55.63% 69.53% 63.76% 52.14% 70.59% 48.00% 51.4% 48.67% 64.29% 60.00% 73.68% 77.78% 75.51% 82.22% 82.07% 84.67% 64.71% 91.67% Page 86 of 110

88 Wetland 263 Settlement 73 Otherland % 100.0% 95.89% 88.61% 81.82% 69.23% Totals Overall Classification Accuracy = 77.0% Page 87 of 110

89 Classified Data Error Matrix Dense Forest Reference Data Moderate Sparse Open Open Closed Closed Perennial Annual Water Total Forest Forest Grassland Shrubland Grassland Shrubland Cropland Cropland Body Wetland Settlement Otherland Dense Forest Moderate Forest Sparse Forest Open Grassland Open Shrubland Closed Grassland Closed Shrubland Perennial Cropland Annual Cropland Water Body Wetland Settlement Otherland Column Total Page 88 of 110

90 1.3. Accuracy Statistics for 2000 Classification Scheme I Overall Classification Accuracy = (1016/1197) % Overall Kappa Statistics = Accuracy Summary Class Name Forestland Grassland Cropland Wetland Settlement Otherland Reference Totals Classified Totals Number Correct Producers Accuracy Users Accuracy 91.58% 84.26% 52.53% 77.03% 88.69% 80.56% 97.74% 96.30% % % 25.00% 33.33% Totals Overall Classification Accuracy = % Error Matrix Classified Data Forestland Grassland Cropland Wetland Settlement Otherland Column Total Reference Data Total Forestland Grassland Cropland Wetland Settlement Otherland Page 89 of 110

91 1.4. Accuracy Statistics for 2000 Classification Scheme II Overall Classification Accuracy = (928/1197) % Overall Kappa Statistics = Accuracy Summary Reference Class Name Total Classified Totals Number Correct Dense Forest Moderate Forest Sparse Forest Closed Grassland Open Grassland Closed Shrubland Open Shrubland Perennial Cropland Annual Cropland Water Body Wetland Settlement Otherland Producers Accuracy Users Accuracy 85.71% 50.00% 90.57% 55.81% 75.16% 79.04% 28.57% 28.57% 41.67% 35.71% 56.06% 52.11% 25.76% 60.71% % % 88.54% 80.34% % % 82.86% 99.62% % % 25.00% 33.33% Totals Overall Classification Accuracy = 77.52% Page 90 of 110

92 Error Matrix Classified Data Dense Forest Moderate Forest Sparse Forest Closed Grassland Open Grassland Closed Shrubland Open Shrubland Perennial Cropland Annual Cropland Water Body Wetland Settlement Otherland Column Total Reference Data Dense Moderate Sparse Closed Open Closed Open Perennial Annual Water Total Forest Forest Forest Grassland Grassland Shrubland Shrubland Cropland Cropland Body Wetland Settlement Otherland Page 91 of 110

93 1.5. Accuracy Statistics for 1990 Classification Scheme I Overall Classification Accuracy = (1047/1193) 87.76% Overall Kappa Statistics = Accuracy Summary Class Name Reference Totals Classified Totals Number Correct Producers Accuracy Users Accuracy Forestland % 90.56% Grassland % 80.37% Cropland % 78.66% Wetland % 92.47% Settlement % 100% Otherland % 100% Totals Overall Classification Accuracy = 85.81% Error Matrix Reference Data Classified Data Forestland Grassland Cropland Wetland Settlement Otherland Totals Forestland Grassland Cropland Wetland Settlement Otherland Column Total P a g e

94 1.6. Accuracy Statistics for 1990 Classification Scheme II Overall Classification Accuracy = (936/1193) % Overall Kappa Statistics = Accuracy Summary Class Name Reference Total Classified Totals Number Correct Producers Accuracy Users Accuracy 53.66% 56.41% Dense Forest Moderate Forest % 75.97% % 75.2% Sparse Forest Closed 11 Grassland % 90.91% % 50% Open Grassland 7 2 Closed 47 Shrubland % 74.47% Open % Shrubland Perennial 5 100% 100% Cropland 5 5 Annual Cropland % 78.21% % 63.33% Water Body % 99.14% Wetland Settlement % 100% % 100% Otherland 7 2 Totals Overall Classification Accuracy = 73.52% P a g e

95 4. Error Matrix Classified Data Dense Forest Moderate Forest Sparse Forest Closed Grassland Open Grassland Closed Shrubland Open Shrubland Perennial Cropland Annual Cropland Water Body Wetland Settlement Otherland Column Total Dens e Forest Moderate Forest Sparse Forest Closed Grassland Open Grassland Closed Shrubland Reference Total Open Shrubland Perennial Cropland Annual Cropla nd Water Body Wetland Settlement Otherland Total 94 P a g e

96 Figure 27: Malawi Land Cover Map for 2010 Scheme I 95 P a g e

97 Figure 28: Malawi Land Cover Map for 2010 Scheme II 96 P a g e

98 Figure 29: Malawi Land Cover Map for 2000 Scheme I 97 P a g e

99 Figure 30: Malawi Land Cover Map for 2000 Scheme II 98 P a g e

100 Figure 31: Malawi Land Cover Map for 1990 Scheme I 99 P a g e

101 Figure 32: Malawi Land Cover Map for 1990 Scheme II 100 P a g e

102 REFERENCES ISO (International Standardisation Organisation). (2002). Geographic information Quality principles, ISO 19113(2002). Morrison, J. L. (1995). Spatial data quality. In J. L. Guptill, S. C. and Morrison (Ed.), Elements of spatial data quality (pp. 1 12). Tokyo. Cochran, W. G. (2007). Sampling techniques. Congalton, R. G. (1991). A Review of Assessing the Accuracy of Classifications of Remotely Sensed Data, 46(October 1990), Daniel, J. (n.d.). Sampling Essentials (pp ). Fitzpatrick-Lins, K. (1981). Comparison of sampling procedures and data analysis for a landuse and land-cover map. Photogrammetric Engineering and Remote Sensing, 47(3), Garson, D. (2012). Table_of_Contents. International immunology, 24(8), NP. doi: /intimm/dxs035 Harmon D. Maher Jr., Dept. of Geography and Geology, University of Nebraska at Omaha, 2004 McCoy, Roger M Field methods in remote sensing, Guilford press Lille sand, T.M., and R.W. Kiefer Remote sensing and image interpretation, 3 rd edition. Wiley, New York. 750pp. Lille sand, T.M., and Chipman, J. 1998, Upper Midwest Gap Analysis Program Image Processing Protocol. Environmental Remote Sensing Center, University of Wisconsin- Madison Meyer WB (1995) Past and present land-use and land-cover in the USA consequences. The Nature and Implications of Environmental Change 1(1), Nerd H (2004) Remote sensing resources: Remote sensing & Geographic Information System facility. Center for Biodiversity and Conservation. American Museum of Natural History. Jensen, J. R., Introductory Digital Image Processing, Prentice-Hall, New Jersey, p P a g e

103 Crist, E.P. and R.C. Cicone, 1984 "Application of the Tasseled Cap Concept to Simulated Thematic Mapper Data," Photogrammetric Engineering and Remote Sensing, Vol. 50, pp Kauth, R.J., P.F. Lambeck, W. Richardson, G.S. Thomas, and A.P. Pentland, "Feature Extraction Applied to Agricultural Crops as Seen by Landsat," Proceedings, LACIE Symposium, Houston TX, NASA, pp Huang, C., B. Wylie, L. Yang, C. Homer, and G. Zylstra. "Derivation of a Tasseled Cap Transformation Based on Landsat 7 At-Satellite Reflectance". USGS EROS Data Center White Paper ( Meyer, R. (2008). Houston-Galveston Area Council 2008 Land Cover Image Processing Protocol, (December). Rosenqvist, Å., Milne, A., Lucas, R., Imhoff, M., & Dobson, C. (2003). A review of remote sensing technology in support of the Kyoto Protocol. Environmental Science & Policy, 6(5), doi: /s (03) Anderson, B. J. R., Hardy, E. E., Roach, J. T., & Witmer, R. E. (2001). A Land Use And Land Cover Classification System For Use With Remote Sensor Data, Goodenough, D. G., Bhogal, A. S. P., Chen, H., & Dyk, A. (2001). Comparison of Methods for Estimation of Kyoto Protocol Products of Forests From Multitemporal Landsat, 00(C), Meyer, R. (2008). Houston-Galveston Area Council 2008 Land Cover Image Processing Protocol, (December). Berberoglu, S., & Akin, a. (2009). Assessing different remote sensing techniques to detect land use/cover changes in the eastern Mediterranean. International Journal of Applied Earth Observation and Geoinformation, 11(1), doi: /j.jag // P a g e

104 ANNEX I Methodology and work flow The chart shown below indicates the work flow and methodology to be applied in land cover mapping for Participating country Figure 33 Land cover mapping workflow 103 P a g e

105 Ancillary and Ground Referencing Data 104 P a g e

106 105 P a g e

107 Ground Referencing Pictures Showing Major Land Cover Categories Part of a wetland cultivated A wetland used for rice cultivation. Cultivated wetland Perennial Cropland (Tea) Perennial Cropland (Coffee) Perennial Cropland (Macadamia Fruits) Annual cropland (cotton) Annual cropland (Maize) 106 P a g e

108 Cropland left fallow Annual cropland (Pigeon peas) Macadamia crops looks like a grassland Dense Forest Village forest (moderate) Moderate Forest (used as graveyards) Bamboo Forest Planted Forest (pine) 107 P a g e

109 Miombo Forest Sparse forest Open Water (lake Malawi) Open grassland Closed grassland Woodlands Vegetated wetland Settlement (Quarry) i) 108 P a g e

110 Chart for 0 to 50% Canopy Chart for 40% to 70% Canopy 109 P a g e

Module 2.1 Monitoring activity data for forests using remote sensing

Module 2.1 Monitoring activity data for forests using remote sensing Module 2.1 Monitoring activity data for forests using remote sensing Module developers: Frédéric Achard, European Commission (EC) Joint Research Centre (JRC) Jukka Miettinen, EC JRC Brice Mora, Wageningen

More information

CHAPTER 1 THE UNITED STATES 2001 NATIONAL LAND COVER DATABASE

CHAPTER 1 THE UNITED STATES 2001 NATIONAL LAND COVER DATABASE CHAPTER 1 THE UNITED STATES 2001 NATIONAL LAND COVER DATABASE Collin Homer*, Jon Dewitz, Joyce Fry, and Nazmul Hossain *U.S. Geological Survey (USGS) Center for Earth Resources Observation and Science

More information

An Internet-based Agricultural Land Use Trends Visualization System (AgLuT)

An Internet-based Agricultural Land Use Trends Visualization System (AgLuT) An Internet-based Agricultural Land Use Trends Visualization System (AgLuT) Second half yearly report 01-01-2001-06-30-2001 Prepared for Missouri Department of Natural Resources Missouri Department of

More information

7.1 INTRODUCTION 7.2 OBJECTIVE

7.1 INTRODUCTION 7.2 OBJECTIVE 7 LAND USE AND LAND COVER 7.1 INTRODUCTION The knowledge of land use and land cover is important for many planning and management activities as it is considered as an essential element for modeling and

More information

Rio Santa Geodatabase Project

Rio Santa Geodatabase Project Rio Santa Geodatabase Project Amanda Cuellar December 7, 2012 Introduction The McKinney research group (of which I am a part) collaborates with international and onsite researchers to evaluate the risks

More information

Abstract. TECHNOFAME- A Journal of Multidisciplinary Advance Research. Vol.2 No. 2, (2013) Received: Feb.2013; Accepted Oct.

Abstract. TECHNOFAME- A Journal of Multidisciplinary Advance Research. Vol.2 No. 2, (2013) Received: Feb.2013; Accepted Oct. Vol.2 No. 2, 83-87 (2013) Received: Feb.2013; Accepted Oct. 2013 Landuse Pattern Analysis Using Remote Sensing: A Case Study of Morar Block, of Gwalior District, M.P. Subhash Thakur 1 Akhilesh Singh 2

More information

Programme. MC : Byron Anangwe. Morning Session

Programme. MC : Byron Anangwe. Morning Session Programme MC : Byron Anangwe Morning Session DAY 2 Summary MC : Byron Anangwe The USGS global land cover mapping initiative Land use and land cover mapping at the Joint Research Centre Alan Belward FAO

More information

Land Use / Land Cover Mapping in

Land Use / Land Cover Mapping in Land Use / Land Cover Mapping in Eastern and Southern African Regions RCMRD Experience by 6/24/2013, Nairobi Kenya Dr. Tesfaye Korme Director of RS, GIS and Mapping, RCMRD I. About RCMRD, Its Vision and

More information

COMBINING ENUMERATION AREA MAPS AND SATELITE IMAGES (LAND COVER) FOR THE DEVELOPMENT OF AREA FRAME (MULTIPLE FRAMES) IN AN AFRICAN COUNTRY:

COMBINING ENUMERATION AREA MAPS AND SATELITE IMAGES (LAND COVER) FOR THE DEVELOPMENT OF AREA FRAME (MULTIPLE FRAMES) IN AN AFRICAN COUNTRY: COMBINING ENUMERATION AREA MAPS AND SATELITE IMAGES (LAND COVER) FOR THE DEVELOPMENT OF AREA FRAME (MULTIPLE FRAMES) IN AN AFRICAN COUNTRY: PRELIMINARY LESSONS FROM THE EXPERIENCE OF ETHIOPIA BY ABERASH

More information

Environmental Impact Assessment Land Use and Land Cover CISMHE 7.1 INTRODUCTION

Environmental Impact Assessment Land Use and Land Cover CISMHE 7.1 INTRODUCTION 7 LAND USE AND LAND COVER 7.1 INTRODUCTION The knowledge of land use and land cover is important for many planning and management activities as it is considered an essential element for modeling and understanding

More information

Geospatial technology for land cover analysis

Geospatial technology for land cover analysis Home Articles Application Environment & Climate Conservation & monitoring Published in : Middle East & Africa Geospatial Digest November 2013 Lemenkova Polina Charles University in Prague, Faculty of Science,

More information

Regional Centre for Mapping of Resources for Development (RCMRD), Nairobi, Kenya

Regional Centre for Mapping of Resources for Development (RCMRD), Nairobi, Kenya Regional Centre for Mapping of Resources for Development (RCMRD), Nairobi, Kenya Introduction to GIS (2 weeks: 10 days) Intakes: 8 th January, 6 th February, 5th March, 3 rd. April 9 th, May 7 th, June

More information

ISO Land Cover for Agricultural Regions of Canada, Circa 2000 Data Product Specification. Revision: A

ISO Land Cover for Agricultural Regions of Canada, Circa 2000 Data Product Specification. Revision: A ISO 19131 Land Cover for Agricultural Regions of Canada, Circa 2000 Data Product Specification Revision: A Data specification: Land Cover for Agricultural Regions, circa 2000 Table of Contents 1. OVERVIEW...

More information

Principals and Elements of Image Interpretation

Principals and Elements of Image Interpretation Principals and Elements of Image Interpretation 1 Fundamentals of Photographic Interpretation Observation and inference depend on interpreter s training, experience, bias, natural visual and analytical

More information

WGIA7 9th July, Noriko KISHIMOTO

WGIA7 9th July, Noriko KISHIMOTO Utilizing Global Map for addressing Climate Change WGIA7 9th July, 2009 Seoul, Republic of Korea Noriko KISHIMOTO n-kishimoto@gsi.go.jp Geographic Survey Institute, JAPAN Outline of the Global Map What

More information

MANUAL ON THE BSES: LAND USE/LAND COVER

MANUAL ON THE BSES: LAND USE/LAND COVER 6. Environment Protection, Management and Engagement 2. Environmental Resources and their Use 5. Human Habitat and Environmental Health 1. Environmental Conditions and Quality 4. Disasters and Extreme

More information

NATIONAL MAPPING EFFORTS: THE PHILIPPINES

NATIONAL MAPPING EFFORTS: THE PHILIPPINES NATIONAL MAPPING EFFORTS: THE PHILIPPINES Dr. RIJALDIA N. SANTOS DENR National Mapping and Resource Information Authority (NAMRIA) May 30, 2018 Land Cover/Land Use Changes (LC/LUC) and Its Impacts on Environment

More information

Quality and Coverage of Data Sources

Quality and Coverage of Data Sources Quality and Coverage of Data Sources Objectives Selecting an appropriate source for each item of information to be stored in the GIS database is very important for GIS Data Capture. Selection of quality

More information

ISO INTERNATIONAL STANDARD. Geographic information Metadata Part 2: Extensions for imagery and gridded data

ISO INTERNATIONAL STANDARD. Geographic information Metadata Part 2: Extensions for imagery and gridded data INTERNATIONAL STANDARD ISO 19115-2 First edition 2009-02-15 Geographic information Metadata Part 2: Extensions for imagery and gridded data Information géographique Métadonnées Partie 2: Extensions pour

More information

Summary Description Municipality of Anchorage. Anchorage Coastal Resource Atlas Project

Summary Description Municipality of Anchorage. Anchorage Coastal Resource Atlas Project Summary Description Municipality of Anchorage Anchorage Coastal Resource Atlas Project By: Thede Tobish, MOA Planner; and Charlie Barnwell, MOA GIS Manager Introduction Local governments often struggle

More information

The Combination of Geospatial Data with Statistical Data for SDG Indicators

The Combination of Geospatial Data with Statistical Data for SDG Indicators Session x: Sustainable Development Goals, SDG indicators The Combination of Geospatial Data with Statistical Data for SDG Indicators Pier-Giorgio Zaccheddu Fabio Volpe 5-8 December2018, Nairobi IAEG SDG

More information

Supplementary material: Methodological annex

Supplementary material: Methodological annex 1 Supplementary material: Methodological annex Correcting the spatial representation bias: the grid sample approach Our land-use time series used non-ideal data sources, which differed in spatial and thematic

More information

Brazil Paper for the. Second Preparatory Meeting of the Proposed United Nations Committee of Experts on Global Geographic Information Management

Brazil Paper for the. Second Preparatory Meeting of the Proposed United Nations Committee of Experts on Global Geographic Information Management Brazil Paper for the Second Preparatory Meeting of the Proposed United Nations Committee of Experts on Global Geographic Information Management on Data Integration Introduction The quick development of

More information

Regional Centre for Mapping of Resources for Development (RCMRD), Nairobi, Kenya

Regional Centre for Mapping of Resources for Development (RCMRD), Nairobi, Kenya Regional Centre for Mapping of Resources for Development (RCMRD), Nairobi, Kenya Introduction GIS ( 2 weeks: 10 days) Intakes: 7 th Jan, 4 th Feb,4 th March, 1 st April 6 th May, 3 rd June, 1 st July,

More information

Urban Growth Analysis: Calculating Metrics to Quantify Urban Sprawl

Urban Growth Analysis: Calculating Metrics to Quantify Urban Sprawl Urban Growth Analysis: Calculating Metrics to Quantify Urban Sprawl Jason Parent jason.parent@uconn.edu Academic Assistant GIS Analyst Daniel Civco Professor of Geomatics Center for Land Use Education

More information

An Update on Land Use & Land Cover Mapping in Ireland

An Update on Land Use & Land Cover Mapping in Ireland An Update on Land Use & Land Cover Mapping in Ireland Progress Towards a National Programme Kevin Lydon k.lydon@epa.ie Office of Environmental Assessment, Environmental Protection Agency, Johnstown Castle,

More information

Development of a Regional Land Cover Monitoring System In the Lower Mekong Region a Joint Effort Between SERVIR-Mekong and Partners -

Development of a Regional Land Cover Monitoring System In the Lower Mekong Region a Joint Effort Between SERVIR-Mekong and Partners - Mekong Development of a Regional Land Cover Monitoring System In the Lower Mekong Region a Joint Effort Between SERVIR-Mekong and Partners - Aekkapol Aekakkararungroj SERVIR-Mekong Asian Disaster Preparedness

More information

Display data in a map-like format so that geographic patterns and interrelationships are visible

Display data in a map-like format so that geographic patterns and interrelationships are visible Vilmaliz Rodríguez Guzmán M.S. Student, Department of Geology University of Puerto Rico at Mayagüez Remote Sensing and Geographic Information Systems (GIS) Reference: James B. Campbell. Introduction to

More information

Regional Centre for Mapping of Resources for Development (RCMRD), Nairobi, Kenya. Introduction GIS ( 2 weeks: 10 days)

Regional Centre for Mapping of Resources for Development (RCMRD), Nairobi, Kenya. Introduction GIS ( 2 weeks: 10 days) Regional Centre for Mapping of Resources for Development (RCMRD), Nairobi, Kenya Introduction GIS ( 2 weeks: 10 days) Intakes: 8 th Jan, 6 th Feb,5 th March, 3 rd April 9 th, May 7 th, June 4 th, July

More information

Regional Centre for Mapping of Resources for Development (RCMRD), Nairobi, Kenya. Introduction GIS (2 weeks: 10 days)

Regional Centre for Mapping of Resources for Development (RCMRD), Nairobi, Kenya. Introduction GIS (2 weeks: 10 days) Regional Centre for Mapping of Resources for Development (RCMRD), Nairobi, Kenya Introduction GIS (: 10 days) Intake Dates: 9 th Jan, 6 th Feb, 6 th Mar, 3 rd April, 8 th May, 5 th June, 3 rd July, 2017

More information

Dynamic Land Cover Dataset Product Description

Dynamic Land Cover Dataset Product Description Dynamic Land Cover Dataset Product Description V1.0 27 May 2014 D2014-40362 Unclassified Table of Contents Document History... 3 A Summary Description... 4 Sheet A.1 Definition and Usage... 4 Sheet A.2

More information

International Journal of Intellectual Advancements and Research in Engineering Computations

International Journal of Intellectual Advancements and Research in Engineering Computations ISSN:2348-2079 Volume-5 Issue-2 International Journal of Intellectual Advancements and Research in Engineering Computations Agricultural land investigation and change detection in Coimbatore district by

More information

This is trial version

This is trial version Journal of Rangeland Science, 2012, Vol. 2, No. 2 J. Barkhordari and T. Vardanian/ 459 Contents available at ISC and SID Journal homepage: www.rangeland.ir Full Paper Article: Using Post-Classification

More information

VISUALIZATION URBAN SPATIAL GROWTH OF DESERT CITIES FROM SATELLITE IMAGERY: A PRELIMINARY STUDY

VISUALIZATION URBAN SPATIAL GROWTH OF DESERT CITIES FROM SATELLITE IMAGERY: A PRELIMINARY STUDY CO-439 VISUALIZATION URBAN SPATIAL GROWTH OF DESERT CITIES FROM SATELLITE IMAGERY: A PRELIMINARY STUDY YANG X. Florida State University, TALLAHASSEE, FLORIDA, UNITED STATES ABSTRACT Desert cities, particularly

More information

ISO Land Use 1990, 2000, 2010 Data Product Specifications. Revision: A

ISO Land Use 1990, 2000, 2010 Data Product Specifications. Revision: A ISO 19131 Land Use 1990, 2000, 2010 Data Product Specifications Revision: A Contents 1 OVERVIEW... 3 1.1 Informal description... 3 1.2 Data product specifications Metadata... 3 1.3 Terms and Definitions...

More information

Vegetation Change Detection of Central part of Nepal using Landsat TM

Vegetation Change Detection of Central part of Nepal using Landsat TM Vegetation Change Detection of Central part of Nepal using Landsat TM Kalpana G. Bastakoti Department of Geography, University of Calgary, kalpanagb@gmail.com Abstract This paper presents a study of detecting

More information

Louisiana Transportation Engineering Conference. Monday, February 12, 2007

Louisiana Transportation Engineering Conference. Monday, February 12, 2007 Louisiana Transportation Engineering Conference Monday, February 12, 2007 Agenda Project Background Goal of EIS Why Use GIS? What is GIS? How used on this Project Other site selection tools I-69 Corridor

More information

Spanish national plan for land observation: new collaborative production system in Europe

Spanish national plan for land observation: new collaborative production system in Europe ADVANCE UNEDITED VERSION UNITED NATIONS E/CONF.103/5/Add.1 Economic and Social Affairs 9 July 2013 Tenth United Nations Regional Cartographic Conference for the Americas New York, 19-23, August 2013 Item

More information

STEREO ANALYST FOR ERDAS IMAGINE Stereo Feature Collection for the GIS Professional

STEREO ANALYST FOR ERDAS IMAGINE Stereo Feature Collection for the GIS Professional STEREO ANALYST FOR ERDAS IMAGINE Stereo Feature Collection for the GIS Professional STEREO ANALYST FOR ERDAS IMAGINE Has Your GIS Gone Flat? Hexagon Geospatial takes three-dimensional geographic imaging

More information

Yanbo Huang and Guy Fipps, P.E. 2. August 25, 2006

Yanbo Huang and Guy Fipps, P.E. 2. August 25, 2006 Landsat Satellite Multi-Spectral Image Classification of Land Cover Change for GIS-Based Urbanization Analysis in Irrigation Districts: Evaluation in Low Rio Grande Valley 1 by Yanbo Huang and Guy Fipps,

More information

Urban Tree Canopy Assessment Purcellville, Virginia

Urban Tree Canopy Assessment Purcellville, Virginia GLOBAL ECOSYSTEM CENTER www.systemecology.org Urban Tree Canopy Assessment Purcellville, Virginia Table of Contents 1. Project Background 2. Project Goal 3. Assessment Procedure 4. Economic Benefits 5.

More information

Landsat-based Global Urban Area Map

Landsat-based Global Urban Area Map 1. TITLE Landsat-based Global Urban Area Map Abbreviation Metadata Identifier Landsat-based Global Urban Area Map LaGURAM LaGURAM20160711142456-DIAS20160706142617-en 2. CONTACT 2.1 CONTACT on DATASET Hiroyuki

More information

An Internet-based Agricultural Land Use Trends Visualization System (AgLuT)

An Internet-based Agricultural Land Use Trends Visualization System (AgLuT) An Internet-based Agricultural Land Use Trends Visualization System (AgLuT) Prepared for Missouri Department of Natural Resources Missouri Department of Conservation 07-01-2000-12-31-2001 Submitted by

More information

International Journal of Scientific & Engineering Research, Volume 6, Issue 7, July ISSN

International Journal of Scientific & Engineering Research, Volume 6, Issue 7, July ISSN International Journal of Scientific & Engineering Research, Volume 6, Issue 7, July-2015 1428 Accuracy Assessment of Land Cover /Land Use Mapping Using Medium Resolution Satellite Imagery Paliwal M.C &.

More information

Yaneev Golombek, GISP. Merrick/McLaughlin. ESRI International User. July 9, Engineering Architecture Design-Build Surveying GeoSpatial Solutions

Yaneev Golombek, GISP. Merrick/McLaughlin. ESRI International User. July 9, Engineering Architecture Design-Build Surveying GeoSpatial Solutions Yaneev Golombek, GISP GIS July Presentation 9, 2013 for Merrick/McLaughlin Conference Water ESRI International User July 9, 2013 Engineering Architecture Design-Build Surveying GeoSpatial Solutions Purpose

More information

IMPROVING REMOTE SENSING-DERIVED LAND USE/LAND COVER CLASSIFICATION WITH THE AID OF SPATIAL INFORMATION

IMPROVING REMOTE SENSING-DERIVED LAND USE/LAND COVER CLASSIFICATION WITH THE AID OF SPATIAL INFORMATION IMPROVING REMOTE SENSING-DERIVED LAND USE/LAND COVER CLASSIFICATION WITH THE AID OF SPATIAL INFORMATION Yingchun Zhou1, Sunil Narumalani1, Dennis E. Jelinski2 Department of Geography, University of Nebraska,

More information

Using Geographic Information Systems and Remote Sensing Technology to Analyze Land Use Change in Harbin, China from 2005 to 2015

Using Geographic Information Systems and Remote Sensing Technology to Analyze Land Use Change in Harbin, China from 2005 to 2015 Using Geographic Information Systems and Remote Sensing Technology to Analyze Land Use Change in Harbin, China from 2005 to 2015 Yi Zhu Department of Resource Analysis, Saint Mary s University of Minnesota,

More information

Large Scale Mapping Policy for the Province of Nova Scotia

Large Scale Mapping Policy for the Province of Nova Scotia Large Scale Mapping Policy for the Province of Nova Scotia December, 2005 Version 1.0 TABLE OF CONTENTS PAGE BACKGROUND...3 POLICY...5 Policy 1.0 Large Scale Mapping Program...5 Policy 2.0 Service Offering...5

More information

Research on Topographic Map Updating

Research on Topographic Map Updating Research on Topographic Map Updating Ivana Javorovic Remote Sensing Laboratory Ilica 242, 10000 Zagreb, Croatia Miljenko Lapaine University of Zagreb, Faculty of Geodesy Kaciceva 26, 10000 Zagreb, Croatia

More information

Monitoring of Forest Cover Change in Sundarban mangrove forest using Remote sensing and GIS

Monitoring of Forest Cover Change in Sundarban mangrove forest using Remote sensing and GIS Monitoring of Forest Cover Change in Sundarban mangrove forest using Remote sensing and GIS By Mohammed Monirul Alam April 2008 Content 1: INTRODUCTION 2: OBJECTIVES 3: METHODOLOGY 4: RESULTS & DISCUSSION

More information

Land cover/land use mapping and cha Mongolian plateau using remote sens. Title. Author(s) Bagan, Hasi; Yamagata, Yoshiki. Citation Japan.

Land cover/land use mapping and cha Mongolian plateau using remote sens. Title. Author(s) Bagan, Hasi; Yamagata, Yoshiki. Citation Japan. Title Land cover/land use mapping and cha Mongolian plateau using remote sens Author(s) Bagan, Hasi; Yamagata, Yoshiki International Symposium on "The Imp Citation Region Specific Systems". 6 Nove Japan.

More information

New Land Cover & Land Use Data for the Chesapeake Bay Watershed

New Land Cover & Land Use Data for the Chesapeake Bay Watershed New Land Cover & Land Use Data for the Chesapeake Bay Watershed Why? The Chesapeake Bay Program (CBP) partnership is in the process of improving and refining the Phase 6 suite of models used to inform

More information

Abstract: About the Author:

Abstract: About the Author: REMOTE SENSING AND GIS IN LAND USE PLANNING Sathees kumar P 1, Nisha Radhakrishnan 2 1 1 Ph.D Research Scholar, Department of Civil Engineering, National Institute of Technology, Tiruchirappalli- 620015,

More information

The Icelandic geographic Land Use database (IGLUD)

The Icelandic geographic Land Use database (IGLUD) Page 1 of 7 The Icelandic geographic Land Use database (IGLUD) Jón Kilde: Norsk institutt for skog og landskap Adresse: http://skogoglandskap.pdc.no/utskrift.php? seks_id=21176&sid=19698&t=v Guðmundsson

More information

BIODIVERSITY CONSERVATION HABITAT ANALYSIS

BIODIVERSITY CONSERVATION HABITAT ANALYSIS BIODIVERSITY CONSERVATION HABITAT ANALYSIS A GIS Comparison of Greater Vancouver Regional Habitat Mapping with Township of Langley Local Habitat Mapping Preface This report was made possible through the

More information

Land Accounts - The Canadian Experience

Land Accounts - The Canadian Experience Land Accounts - The Canadian Experience Development of a Geospatial database to measure the effect of human activity on the environment Who is doing Land Accounts Statistics Canada (national) Component

More information

2015 Nigerian National Settlement Dataset (including Population Estimates)

2015 Nigerian National Settlement Dataset (including Population Estimates) 2015 Nigerian National Settlement Dataset (including Population Estimates) Data Users Report and Meta Data (version 2.0) A Commercial Spatial Data Product Developed by GeoTerraImage (Pty) Ltd, South Africa

More information

Evaluating Urban Vegetation Cover Using LiDAR and High Resolution Imagery

Evaluating Urban Vegetation Cover Using LiDAR and High Resolution Imagery Evaluating Urban Vegetation Cover Using LiDAR and High Resolution Imagery Y.A. Ayad and D. C. Mendez Clarion University of Pennsylvania Abstract One of the key planning factors in urban and built up environments

More information

DEVELOPMENT OF DIGITAL CARTOGRAPHIC DATABASE FOR MANAGING THE ENVIRONMENT AND NATURAL RESOURCES IN THE REPUBLIC OF SERBIA

DEVELOPMENT OF DIGITAL CARTOGRAPHIC DATABASE FOR MANAGING THE ENVIRONMENT AND NATURAL RESOURCES IN THE REPUBLIC OF SERBIA DEVELOPMENT OF DIGITAL CARTOGRAPHIC BASE FOR MANAGING THE ENVIRONMENT AND NATURAL RESOURCES IN THE REPUBLIC OF SERBIA Dragutin Protic, Ivan Nestorov Institute for Geodesy, Faculty of Civil Engineering,

More information

Training on national land cover classification systems. Toward the integration of forest and other land use mapping activities.

Training on national land cover classification systems. Toward the integration of forest and other land use mapping activities. Training on national land cover classification systems Toward the integration of forest and other land use mapping activities. Guiana Shield 9 to 13 March 2015, Paramaribo, Suriname Background Sustainable

More information

Open Source Software Education in Texas

Open Source Software Education in Texas Open Source Software Education in Texas PHILLIP DAVIS / RICHARD SMITH GEOACADEMY The Challenge for Open Source Adoption OPEN SOURCE Less the 5% of US colleges and universities offer training in Free and

More information

Preparation of LULC map from GE images for GIS based Urban Hydrological Modeling

Preparation of LULC map from GE images for GIS based Urban Hydrological Modeling International Conference on Modeling Tools for Sustainable Water Resources Management Department of Civil Engineering, Indian Institute of Technology Hyderabad: 28-29 December 2014 Abstract Preparation

More information

Digital Change Detection Using Remotely Sensed Data for Monitoring Green Space Destruction in Tabriz

Digital Change Detection Using Remotely Sensed Data for Monitoring Green Space Destruction in Tabriz Int. J. Environ. Res. 1 (1): 35-41, Winter 2007 ISSN:1735-6865 Graduate Faculty of Environment University of Tehran Digital Change Detection Using Remotely Sensed Data for Monitoring Green Space Destruction

More information

GEOGRAPHIC INFORMATION SYSTEMS Session 8

GEOGRAPHIC INFORMATION SYSTEMS Session 8 GEOGRAPHIC INFORMATION SYSTEMS Session 8 Introduction Geography underpins all activities associated with a census Census geography is essential to plan and manage fieldwork as well as to report results

More information

RWANDA LAND COVER MAPPING FOR Report on the Inception Meeting

RWANDA LAND COVER MAPPING FOR Report on the Inception Meeting RWANDA LAND COVER MAPPING FOR 2015 Report on the Inception Meeting Prepared By: Regional Centre for Mapping of Resources for Development October 2016 Phone: +254 020 2680748/2680722 Mobile: +254 723 786161

More information

Lecture 9: Reference Maps & Aerial Photography

Lecture 9: Reference Maps & Aerial Photography Lecture 9: Reference Maps & Aerial Photography I. Overview of Reference and Topographic Maps There are two basic types of maps? Reference Maps - General purpose maps & Thematic Maps - maps made for a specific

More information

Progress and Land-Use Characteristics of Urban Sprawl in Busan Metropolitan City using Remote sensing and GIS

Progress and Land-Use Characteristics of Urban Sprawl in Busan Metropolitan City using Remote sensing and GIS Progress and Land-Use Characteristics of Urban Sprawl in Busan Metropolitan City using Remote sensing and GIS Homyung Park, Taekyung Baek, Yongeun Shin, Hungkwan Kim ABSTRACT Satellite image is very usefully

More information

1. Introduction. S.S. Patil 1, Sachidananda 1, U.B. Angadi 2, and D.K. Prabhuraj 3

1. Introduction. S.S. Patil 1, Sachidananda 1, U.B. Angadi 2, and D.K. Prabhuraj 3 Cloud Publications International Journal of Advanced Remote Sensing and GIS 2014, Volume 3, Issue 1, pp. 525-531, Article ID Tech-249 ISSN 2320-0243 Research Article Open Access Machine Learning Technique

More information

The Road to Data in Baltimore

The Road to Data in Baltimore Creating a parcel level database from high resolution imagery By Austin Troy and Weiqi Zhou University of Vermont, Rubenstein School of Natural Resources State and local planning agencies are increasingly

More information

Quick Response Report #126 Hurricane Floyd Flood Mapping Integrating Landsat 7 TM Satellite Imagery and DEM Data

Quick Response Report #126 Hurricane Floyd Flood Mapping Integrating Landsat 7 TM Satellite Imagery and DEM Data Quick Response Report #126 Hurricane Floyd Flood Mapping Integrating Landsat 7 TM Satellite Imagery and DEM Data Jeffrey D. Colby Yong Wang Karen Mulcahy Department of Geography East Carolina University

More information

Extent. Level 1 and 2. October 2017

Extent. Level 1 and 2. October 2017 Extent Level 1 and 2 October 2017 Overview: Extent account 1. Learning objectives 2. Review of Level 0 (5m) 3. Level 1 (Compilers): Concepts (15m) Group exercise and discussion (15m) 4. Level 2 (Data Providers)

More information

Challenges and Successes in Sharing Geospatial Data in Africa

Challenges and Successes in Sharing Geospatial Data in Africa Challenges and Successes in Sharing Geospatial Data in Africa 2018 GeoNode Summit Torino, Italy March 26-28, 2018 Bernard Justus Muhwezi Manager, Geo-Information Services Uganda Bureau of Statistics, Kampala,

More information

HIRES 2017 Syllabus. Instructors:

HIRES 2017 Syllabus. Instructors: HIRES 2017 Syllabus Instructors: Dr. Brian Vant-Hull: Steinman 185, 212-650-8514, brianvh@ce.ccny.cuny.edu Ms. Hannah Aizenman: NAC 7/311, 212-650-6295, haizenman@ccny.cuny.edu Dr. Tarendra Lakhankar:

More information

MAPPING LAND USE/ LAND COVER OF WEST GODAVARI DISTRICT USING NDVI TECHNIQUES AND GIS Anusha. B 1, Sridhar. P 2

MAPPING LAND USE/ LAND COVER OF WEST GODAVARI DISTRICT USING NDVI TECHNIQUES AND GIS Anusha. B 1, Sridhar. P 2 MAPPING LAND USE/ LAND COVER OF WEST GODAVARI DISTRICT USING NDVI TECHNIQUES AND GIS Anusha. B 1, Sridhar. P 2 1 M. Tech. Student, Department of Geoinformatics, SVECW, Bhimavaram, A.P, India 2 Assistant

More information

Submitted to: Central Coalfields Limited Ranchi, Jharkhand. Ashoka & Piparwar OCPs, CCL

Submitted to: Central Coalfields Limited Ranchi, Jharkhand. Ashoka & Piparwar OCPs, CCL Land Restoration / Reclamation Monitoring of more than 5 million cu. m. (Coal + OB) Capacity Open Cast Coal Mines of Central Coalfields Limited Based on Satellite Data for the Year 2013 Ashoka & Piparwar

More information

Monitoring Vegetation Growth of Spectrally Landsat Satellite Imagery ETM+ 7 & TM 5 for Western Region of Iraq by Using Remote Sensing Techniques.

Monitoring Vegetation Growth of Spectrally Landsat Satellite Imagery ETM+ 7 & TM 5 for Western Region of Iraq by Using Remote Sensing Techniques. Monitoring Vegetation Growth of Spectrally Landsat Satellite Imagery ETM+ 7 & TM 5 for Western Region of Iraq by Using Remote Sensing Techniques. Fouad K. Mashee, Ahmed A. Zaeen & Gheidaa S. Hadi Remote

More information

SECTION 1: Identification Information

SECTION 1: Identification Information Page 1 of 6 Home Data Catalog Download Data Data Status Web Services About FAQ Contact Us SECTION 1: Identification Information Originator: Minnesota DNR - Division of Forestry Title: GAP Land Cover -

More information

1. Introduction. Chaithanya, V.V. 1, Binoy, B.V. 2, Vinod, T.R. 2. Publication Date: 8 April DOI: https://doi.org/ /cloud.ijarsg.

1. Introduction. Chaithanya, V.V. 1, Binoy, B.V. 2, Vinod, T.R. 2. Publication Date: 8 April DOI: https://doi.org/ /cloud.ijarsg. Cloud Publications International Journal of Advanced Remote Sensing and GIS 2017, Volume 6, Issue 1, pp. 2088-2096 ISSN 2320 0243, Crossref: 10.23953/cloud.ijarsg.112 Research Article Open Access Estimation

More information

Southern African Land Cover ( )

Southern African Land Cover ( ) Southern African Land Cover (2013-14) (Hill-shaded 3D version) Data Users Report and Meta Data (version 1.2) A Commercial Spatial Data Product Developed by GeoTerraImage (Pty) Ltd, South Africa (www.geoterraimage.com)

More information

Landuse and Landcover change analysis in Selaiyur village, Tambaram taluk, Chennai

Landuse and Landcover change analysis in Selaiyur village, Tambaram taluk, Chennai Landuse and Landcover change analysis in Selaiyur village, Tambaram taluk, Chennai K. Ilayaraja Department of Civil Engineering BIST, Bharath University Selaiyur, Chennai 73 ABSTRACT The synoptic picture

More information

ENVI Tutorial: Vegetation Analysis

ENVI Tutorial: Vegetation Analysis ENVI Tutorial: Vegetation Analysis Vegetation Analysis 2 Files Used in this Tutorial 2 About Vegetation Analysis in ENVI Classic 2 Opening the Input Image 3 Working with the Vegetation Index Calculator

More information

Economic and Social Council

Economic and Social Council United Nations Economic and Social Council Distr.: General 2 July 2012 E/C.20/2012/10/Add.1 Original: English Committee of Experts on Global Geospatial Information Management Second session New York, 13-15

More information

SATELLITE REMOTE SENSING

SATELLITE REMOTE SENSING SATELLITE REMOTE SENSING of NATURAL RESOURCES David L. Verbyla LEWIS PUBLISHERS Boca Raton New York London Tokyo Contents CHAPTER 1. SATELLITE IMAGES 1 Raster Image Data 2 Remote Sensing Detectors 2 Analog

More information

Indicator: Proportion of the rural population who live within 2 km of an all-season road

Indicator: Proportion of the rural population who live within 2 km of an all-season road Goal: 9 Build resilient infrastructure, promote inclusive and sustainable industrialization and foster innovation Target: 9.1 Develop quality, reliable, sustainable and resilient infrastructure, including

More information

APPLICATION OF LAND CHANGE MODELER FOR PREDICTION OF FUTURE LAND USE LAND COVER A CASE STUDY OF VIJAYAWADA CITY

APPLICATION OF LAND CHANGE MODELER FOR PREDICTION OF FUTURE LAND USE LAND COVER A CASE STUDY OF VIJAYAWADA CITY APPLICATION OF LAND CHANGE MODELER FOR PREDICTION OF FUTURE LAND USE LAND COVER A CASE STUDY OF VIJAYAWADA CITY K. Sundara Kumar 1, Dr. P. Udaya Bhaskar 2, Dr. K. Padmakumari 3 1 Research Scholar, 2,3

More information

GIS = Geographic Information Systems;

GIS = Geographic Information Systems; What is GIS GIS = Geographic Information Systems; What Information are we talking about? Information about anything that has a place (e.g. locations of features, address of people) on Earth s surface,

More information

USE OF LANDSAT IMAGERY FOR EVALUATION OF LAND COVER / LAND USE CHANGES FOR A 30 YEAR PERIOD FOR THE LAKE ERIE WATERSHED

USE OF LANDSAT IMAGERY FOR EVALUATION OF LAND COVER / LAND USE CHANGES FOR A 30 YEAR PERIOD FOR THE LAKE ERIE WATERSHED USE OF LANDSAT IMAGERY FOR EVALUATION OF LAND COVER / LAND USE CHANGES FOR A 30 YEAR PERIOD FOR THE LAKE ERIE WATERSHED Mark E. Seidelmann Carolyn J. Merry Dept. of Civil and Environmental Engineering

More information

DATA SOURCES AND INPUT IN GIS. By Prof. A. Balasubramanian Centre for Advanced Studies in Earth Science, University of Mysore, Mysore

DATA SOURCES AND INPUT IN GIS. By Prof. A. Balasubramanian Centre for Advanced Studies in Earth Science, University of Mysore, Mysore DATA SOURCES AND INPUT IN GIS By Prof. A. Balasubramanian Centre for Advanced Studies in Earth Science, University of Mysore, Mysore 1 1. GIS stands for 'Geographic Information System'. It is a computer-based

More information

DETECTION AND ANALYSIS OF LAND-USE/LAND-COVER CHANGES IN NAY PYI TAW, MYANMAR USING SATELLITE REMOTE SENSING IMAGES

DETECTION AND ANALYSIS OF LAND-USE/LAND-COVER CHANGES IN NAY PYI TAW, MYANMAR USING SATELLITE REMOTE SENSING IMAGES DETECTION AND ANALYSIS OF LAND-USE/LAND-COVER CHANGES IN NAY PYI TAW, MYANMAR USING SATELLITE REMOTE SENSING IMAGES Kay Khaing Oo 1, Eiji Nawata 1, Kiyoshi Torii 2 and Ke-Sheng Cheng 3 1 Division of Environmental

More information

Wetland Mapping. Wetland Mapping in the United States. State Wetland Losses 53% in Lower US. Matthew J. Gray University of Tennessee

Wetland Mapping. Wetland Mapping in the United States. State Wetland Losses 53% in Lower US. Matthew J. Gray University of Tennessee Wetland Mapping Caribbean Matthew J. Gray University of Tennessee Wetland Mapping in the United States Shaw and Fredine (1956) National Wetlands Inventory U.S. Fish and Wildlife Service is the principle

More information

A Method for Mapping Settlement Area Boundaries in the Greater Golden Horseshoe

A Method for Mapping Settlement Area Boundaries in the Greater Golden Horseshoe A Method for Mapping Settlement Area Boundaries in the Greater Golden Horseshoe Purpose This paper describes a method for mapping and measuring the lands designated for growth and urban expansion in the

More information

A Comprehensive Inventory of the Number of Modified Stream Channels in the State of Minnesota. Data, Information and Knowledge Management.

A Comprehensive Inventory of the Number of Modified Stream Channels in the State of Minnesota. Data, Information and Knowledge Management. A Comprehensive Inventory of the Number of Modified Stream Channels in the State of Minnesota Data, Information and Knowledge Management Glenn Skuta Environmental Analysis and Outcomes Division Minnesota

More information

ESBN. Working Group on INSPIRE

ESBN. Working Group on INSPIRE ESBN Working Group on INSPIRE by Marc Van Liedekerke, Endre Dobos and Paul Smits behalf of the WG members WG participants Marc Van Liedekerke Panos Panagos Borut Vrščaj Ivana Kovacikova Erik Obersteiner

More information

Martin MENSA, Eli SABLAH, Emmanuel AMAMOO-OTCHERE and Foster MENSAH, Ghana. Key words: Feeder Roads Condition Survey, Database Development

Martin MENSA, Eli SABLAH, Emmanuel AMAMOO-OTCHERE and Foster MENSAH, Ghana. Key words: Feeder Roads Condition Survey, Database Development Digital Mapping and GIS-Driven Feeder Road Network Database Management System for Road Project Planning and Implementation Monitoring in the Feeder Road Sector Martin MENSA, Eli SABLAH, Emmanuel AMAMOO-OTCHERE

More information

ISO MODIS NDVI Weekly Composites for Canada South of 60 N Data Product Specification

ISO MODIS NDVI Weekly Composites for Canada South of 60 N Data Product Specification ISO 19131 MODIS NDVI Weekly Composites for South of 60 N Data Product Specification Revision: A Data specification: MODIS NDVI Composites for South of 60 N - Table of Contents - 1. OVERVIEW... 3 1.1. Informal

More information

KENYA NATIONAL BUREAU OF STATISTICS Workshop on

KENYA NATIONAL BUREAU OF STATISTICS Workshop on KENYA NATIONAL BUREAU OF STATISTICS Workshop on Capacity Building in Environment Statistics: the Framework for the Development of Environment Statistics (FDES 2013) Coordination with Sector Ministries

More information

LAND COVER CATEGORY DEFINITION BY IMAGE INVARIANTS FOR AUTOMATED CLASSIFICATION

LAND COVER CATEGORY DEFINITION BY IMAGE INVARIANTS FOR AUTOMATED CLASSIFICATION LAND COVER CATEGORY DEFINITION BY IMAGE INVARIANTS FOR AUTOMATED CLASSIFICATION Nguyen Dinh Duong Environmental Remote Sensing Laboratory Institute of Geography Hoang Quoc Viet Rd., Cau Giay, Hanoi, Vietnam

More information

Investigation of the Effect of Transportation Network on Urban Growth by Using Satellite Images and Geographic Information Systems

Investigation of the Effect of Transportation Network on Urban Growth by Using Satellite Images and Geographic Information Systems Presented at the FIG Congress 2018, May 6-11, 2018 in Istanbul, Turkey Investigation of the Effect of Transportation Network on Urban Growth by Using Satellite Images and Geographic Information Systems

More information

Chesapeake Bay Remote Sensing Pilot Executive Briefing

Chesapeake Bay Remote Sensing Pilot Executive Briefing Chesapeake Bay Remote Sensing Pilot Executive Briefing Introduction In his Executive Order 13506 in May 2009, President Obama stated The Chesapeake Bay is a national treasure constituting the largest estuary

More information