Source: UNIVERSITY OF NEW HAMPSHIRE submitted to NRP
VALIDATING REMOTELY SENSED FOREST AND OTHER LAND COVER MAPS GENERATED USING OBJECT-BASED IMAGE ANALYSIS AND OVER LARGE AREAS
Sponsoring Institution
National Institute of Food and Agriculture
Project Status
COMPLETE
Funding Source
MCINTIRE-STENNIS
Reporting Frequency
Annual
Accession No.
1002519
Grant No.
(N/A)
Cumulative Award Amt.
(N/A)
Proposal No.
(N/A)
Multistate No.
(N/A)
Project Start Date
Oct 1, 2014
Project End Date
Sep 30, 2018
Grant Year
(N/A)
Program Code
[(N/A)]- (N/A)
Recipient Organization
UNIVERSITY OF NEW HAMPSHIRE
51 COLLEGE RD SERVICE BLDG 107
DURHAM,NH 03824
Performing Department
Natural Resources and the Environment
Non Technical Summary
Forest/land cover maps derived from remotely sensed imagery are routinely used for a variety of resource management and policy decisions. These maps must be of the highest accuracy possible and new methods for creating better maps are continually being developed. Techniques for assessing the accuracy of these maps are absolutely vital to the continued development and effective use of these maps. The goal of this project is to investigate, develop, and evaluate some new methods for assessing the accuracy of maps that extend over large geographic areas and/or are derived using a new object-based image analysis method. The successful development of these assessment methods will lead to improved evaluation of these maps that in turn will lead to better management and policy decisions that will benefit everyone.
Animal Health Component
25%
Research Effort Categories
Basic
50%
Applied
25%
Developmental
25%
Classification

Knowledge Area (KA)Subject of Investigation (SOI)Field of Science (FOS)Percent
1237210209060%
1367210209040%
Knowledge Area
136 - Conservation of Biological Diversity; 123 - Management and Sustainability of Forest Resources;

Subject Of Investigation
7210 - Remote sensing equipment and technology;

Field Of Science
2090 - Statistics, econometrics, and biometrics;
Goals / Objectives
1. To evaluate the sampling issues related to labeling a pixel or cluster of pixels vs. an entire object (i.e., polygon) and determine the most effective method for object-based image classification. 2. To investigate the complexities of conducting a validation over large areas including consideration of appropriate sampling scheme, sample unit, and sample number. Multiple approached will be explored and recommendations made to most effectively and efficiently assess these maps.
Project Methods
The procedures for accomplishing this research can be outlined in the following steps:OBJECTIVE 1:1A. Create a "best" forest/land cover map using object-based image analysis.1B. Use the image objects to select reference sample units on the ground.1C. Perform an analysis to determine the necessary sampling strategies to accurately label the reference sample units.OBJECTIVE 2:2A. Select some large areas (multiple countries/continents) for validation of the remotely sensed maps. 2B. Compare traditional, small area sampling strategies with large area approaches such as hierarchical sampling and maplet approaches. 2C. Evaluate the results of each approach and determine best method.1A. Create a "best" forest/land cover map.The study area for objective 1 will be the Coastal Watershed of New Hampshire. New Landsat 8 satellite, launched in February 2013 with increased spectral and radiometric resolution, imagery is available for the entire study area. Other higher spatial resolution imagery is available for subsections of the study area. All the image processing will be performed using a combination of ERDAS Imagine 2013 and Definiens eCognition v.8.8 image analysis software. Object-based image analysis will be used to produce homogeneous vegetation polygons (i.e., objects). It is important to note that this automated creation of polygons in many ways mimics what we as humans would delineate with a pencil.1B. Use the image objects to select reference sample units on the groundOnce the forest/land cover map has been created from the remotely sensed data, polygons will be randomly selected for use in the validation process. Each of these polygons will be designated as a reference data sample unit and visited on the ground to determine the actual ground reference label. For traditional pixel-based accuracy assessment, a 3x3 pixel sample is used to determine the ground label for the area regardless of the size of the homogeneous area from which the sample was taken. However, in an object-based analysis, the entire homogeneous area has been grouped together as an object (i.e., polygon) and therefore, obtaining a single observation within the polygon may not be sufficient to accurately label that polygon. An analysis will be conducted to investigate the number and distribution of samples required within a reference data sample unit polygon.1C. Perform an analysis to determine the necessary sampling strategies to accurately lable the reference sample unitsReference data sample units will be field visited and observations made systematically throughout the polygon with the number of observations proportional to the size of the polygon. Where necessary, especially in the forest vegetation types, measurements will be used to augment the observations. Monte Carlo/Bootstrap methods will be used to analyze the data and determine the number needed to accurately label each polygon (Efron and Tibshirani 1993). It is anticipated that the number of observations in a polygon will be dependent on the land cover type with forests being the most complex and requiring the most observations. The results of this analysis will then be used to provide guidelines for labeling the reference data sample units used when assessing the accuracy of land cover maps generated from object-based image analysis of remotely sensed data.2A. Select some large areas (multiple countries/continents) for validation of the remotely sensed maps.A second objective in this research is to investigate the complexities of map validation (accuracy assessment) conducted over large geographic areas. Selection of areas to conduct this analysis will be performed in conjunction with a global mapping project underway by the project director of this proposal. This global mapping project is a five-year effort funded by NASA with a team of investigators from USGS Flagstaff, AZ; USGS EROS Data Center; NASA Ames Research Center; NASA Goddard Space Flight Center; University of Wisconsin; University of Northern Arizona; and UNH. Selection of large areas used in this proposal will be dependent upon at least preliminary mapping of large, continental regions from the global mapping project. North America is currently available for use and it is anticipated that most of Asia will be available by the fall of 2014.2B. Compare traditional, small area sampling strategies with large area approaches such as hierarchical sampling and maplet approaches.Once exact large areas have been identified, a comparison will be conducted between the merits of traditional small area sampling strategies applied to these larger areas with other methods includiing a hierarchical or nested approach (Stehman et al. 2012) and a maplet approach (Iiames and Lunetta In Press). The hierarchical approach involves stratifying the sampling based on different map scales and then selecting representative areas at the largest scale. The result is to apply the sampling to representative areas of the whole selected in a methodical fashion so as to include as much variability across the area as possible. The maplet approach is innovative in that it requires a total enumeration of smaller areas (perhaps 1 km x 1 km) where the entire area is labeled on the ground or from very high spatial resolution imagery and then the entire "maplet" area is compared to the remotely sensed map to conduct the validation. 2C. Evaluate the results of each approach and determine best method.The advantages, disadvantages, costs, and benefits of each of these approaches will be evaluated in this project. Recommendations will be made to provide guidance about when each method is appropriate. The results of this evaluation will not only be important to this research, but will also be used for the global mapping project and for future accuracy assessments of very large areas.

Progress 10/01/14 to 09/30/18

Outputs
Target Audience:The target audience for most of the work conducted during this reporting period is the remote sensing / geospatial analysis community. Sampling for assessing the accuracy of maps created from remotely sensed imagery is central to the effective use of these maps for resource management. New technologies that allow the creation of segments or objects (groupings of pixels) instead of mapping individual pixels has vastly improved our mapping capabilities and new sampling methods must be used to assess these maps. In addition, mapping projects have continued to expand from mapping small, local areas to mapping countries and entire continents. Assessing the accuracy of maps over large areas also requires special considerations not appropriate for local area mapping. Therefore, this work is also targeted to those in the natural resource and agricultural communities who use thematic maps generated from remotely sensed data. Most of the work performed this year has been presented at three conferences. One of these conferences was New Hampshire Farm and Forest Expo in Manchester, NH on February 1, 2018. A second conference was the American Society for Photogrammetry and Remote Sensing (ASPRS) Annual Conference in Denver, CO on February 5, 2018 and the third conference was the New England Region of the Society of American Foresters (SAF) Conference on March 28, 2018. Changes/Problems: Nothing Reported What opportunities for training and professional development has the project provided?A PhD graduate student worked on this project as part of her dissertation research. This student mentored an undergraduate student intern that also worked on this project. The undergraduate intern was funded as part of the project investigator's New Hampshire View project. NHView is part of the AmericaView Consortium. Two MS students were also part of this research. One used UAS to investigate collecting the reference data needed for accuracy assessment. The other is conducting an in-depth analysis of North America using a variety of different sampling approaches to gain understanding of the advantages and disadvantages of each and to select the most effective way to best assess map accuracy. How have the results been disseminated to communities of interest?Most of the work performed this year has been published in peer-reviewed journals and/or presented at three conferences. One of these conferences was New Hampshire Farm and Forest Expo in Manchester, NH on February 1, 2018. A second conference was the American Society for Photogrammetry and Remote Sensing (ASPRS) Annual Conference in Denver, CO on February 5, 2018 and the third conference was the New England Region of the Society of American Foresters (SAF) Conference on March 28, 2018. In addition, the research using UAS imagery has been incorporated into the Remote Sensing of Environment course at the University of New Hampshire. What do you plan to do during the next reporting period to accomplish the goals? Nothing Reported

Impacts
What was accomplished under these goals? The Basic and Applied Spatial Analysis Lab (BASAL) in the Department of Natural Resources & the Environment at the University of New Hampshire has a long history of investigating new techniques for assessing the thematic accuracy of maps created from remotely sensed data. Forest/land cover maps derived from remotely sensed imagery are routinely used for a variety of resource management and policy decisions including estimating crop and timber production, monitoring wetland degradation, determining habitat suitability, identifying invasive species, creating fire/fuel potential maps, etc. These maps must be of the highest accuracy possible to be effectively used and new methods for creating better maps are continually being developed. Therefore, techniques for assessing the accuracy of these maps are absolutely vital to the continued development and operational use of these maps. The goal of this project was to investigate, develop, and evaluate some new methods for assessing the accuracy of maps that extend over large geographic areas and/or are derived using a new object-based image analysis method. The successful development of these assessment methods will lead to improved evaluation of these maps that in turn will lead to better management and policy decisions that will benefit everyone. The research was conducted at both the local and large area levels during this project. A pilot study was conducted to evaluate using low altitude imagery as a source of reference data. Unmanned Aerial Systems (UAS), also called drones, were used to collect imagery to assess both the pixel-based and object-based classification approaches. Two graduate students obtained Remote Operator Certification from the Federal Aviation Administration (FAA) that allows for the collection of UAS imagery. Images were collected over the University of New Hampshire woodlands properties for which ground reference data were already available facilitating effective comparison with an evaluation of the UAS imagery. Research on the theoretical analysis and practical application of collecting reference data and conducting a valid accuracy assessment over large areas was also performed. This analysis was done over six continents (excluding Antarctica). Each continent offered unique challenges and problems related not only to reference data collection but also for stratification used in sampling and proper balancing of the sampling by land cover type. A paper was published on the specific issues explored using North America, Africa, and Australia and a follow-up paper has been submitted describing issues with every continent.

Publications

  • Type: Journal Articles Status: Published Year Published: 2018 Citation: Yadav, Kamini and Russell G. Congalton. 2018. Issues with large area thematic accuracy assessment for mapping cropland extent: a tale of three continents. Remote Sensing. 10, 53. DOI:10.3390/rs10010053.
  • Type: Journal Articles Status: Published Year Published: 2018 Citation: Sun, Peijun, Russell G. Congalton, and Yaozhong Pan. 2018. Improving the upscaling of land cover maps by fusing uncertainty and spatial structure information. Photogrammetric Engineering and Remote Sensing. Vol. 84, No. 2. pp. 87  100. DOI: 10.14358/PERS.84.2.87.
  • Type: Journal Articles Status: Published Year Published: 2018 Citation: Sun, Peijun and Russell G. Congalton. 2018. Using a similarity matrix approach to evaluate the accuracy of rescaled maps. Remote Sensing. 10, 487. DOI:10.3390/rs10030487.
  • Type: Journal Articles Status: Published Year Published: 2018 Citation: Fraser, Benjamin T. and Russell G. Congalton. 2018. Issues in Unmanned Aerial Systems (UAS) Data Collection of Complex Forest Environments. Remote Sensing. 10, 908. DOI:10.3390/rs10060908.
  • Type: Journal Articles Status: Published Year Published: 2018 Citation: Sun, Peijun, Russell G. Congalton, and Yaozhong Pan. 2018. Using a simulation analysis to evaluate the impact of crop mapping error on crop area estimation from stratified sampling. International Journal of Digital Earth. DOI:10.1080/17538947.2018.1499827.


Progress 10/01/16 to 09/30/17

Outputs
Target Audience:The target audience for most of the work conducted during this reporting period is the remote sensing / geospatial analysis community. Sampling for assessing the accuracy of maps created from remotely sensed imagery is central to the effective use of these maps for resource management. New technologies that allow the creation of segments or objects (groupings of pixels) instead of mapping individual pixels has vastly improved our mapping capabilities and new sampling methods must be used to assess these maps. In addition, mapping projects have continued to expand from mapping small, local areas to mapping countries and entire continents. Assessing the accuracy of maps over large areas also requires special considerations not appropriate for local area mapping. Most of the work performed this year has been presented at two workshops of the Global Food Security Support Analysis Data working group and has been used by this group for their global crop mapping efforts. In addition, a presentation was made on global accuracy assessment at the American Society for Photogrammetry and Remote Sensing (ASPRS) Annual Conference in Baltimore, MD on March 15, 2017. Some additional work was performed on local sampling strategies and mapping of the Coastal Watershed of NH. This local work was presented at the ASPRS Annual Conference in Baltimore, MD on March 16, 2017. Changes/Problems:No changes have been made to the project this year. It remains on-track for successful completion by the end of the project period. What opportunities for training and professional development has the project provided?A PhD graduate student worked on this project as part of her dissertation research. This student mentored an undergraduate student intern that also worked on this project. The undergraduate intern was funded as part of the project investigator's New Hampshire View project. NHView is part of the AmericaView Consortium. Two MS students were also part of this research. One used UAS to investigate collecting the reference data needed for accuracy assessment. The other is conducting an in-depth analysis of North America using a variety of different sampling approaches to gain understanding of the advantages and disadvantages of each and to select the most effective way to best assess map accuracy. How have the results been disseminated to communities of interest?Most of the work performed this year has been presented at two workshops of the Global Food Security Support Analysis Data working group (a part of the NASA MEaSUREs Program) and has been used by this group for their global crop mapping efforts. In addition, a presentation was made on global accuracy assessment at the American Society for Photogrammetry and Remote Sensing (ASPRS) Annual Conference in Baltimore, MD on March 15, 2017. Work that was performed on local sampling strategies and mapping of the Coastal Watershed of NH was presented at the ASPRS Annual Conference in Baltimore, MD on March 16, 2017. Finally, an MS Thesis has been completed and published through the University of New Hampshire Graduate School. What do you plan to do during the next reporting period to accomplish the goals?A lot of work was accomplished on both major project goals this year, especially in collecting most of the data needed for analysis. Work will continue both locally to analyze the data to evaluate the potential for using UASs for collecting reference data for pixel-based and object-based mapping projects. In addition, there is still considerable work to do on both analyzing the sampling strategies for large area accuracy assessments and for effectively collecting reference data needed in such assessments. More simulations will be performed in North America to explore the sampling strategies.

Impacts
What was accomplished under these goals? Forest/land cover maps derived from remotely sensed imagery are routinely used for a variety of resource management and policy decisions. These maps must be of the highest accuracy possible to be effectively used and new methods for creating better maps are continually being developed. Therefore, techniques for assessing the accuracy of these maps are absolutely vital to the continued development and operational use of these maps. The goal of this project is to investigate, develop, and evaluate some new methods for assessing the accuracy of maps that extend over large geographic areas and/or are derived using a new object-based image analysis method. The successful development of these assessment methods will lead to improved evaluation of these maps that in turn will lead to better management and policy decisions that will benefit everyone. Research was conducted at both the local and large area scales during this reporting period. Work that began last year to evaluate using low altitude imagery as a source of reference data was continued. Unmanned Aerial Systems (UAS), also called drones, were used to collect imagery to assess both the pixel-based and object-based classification approaches. Two graduate students obtained Remote Operator Certification from the Federal Aviation Administration (FAA) that allows for the collection of UAS imagery. Images were collected over University of New Hampshire woodlands properties for which ground reference data were already available facilitating effective comparison with UAS imagery. Work on the theoretical analysis and practical application of collecting reference data and conducting a valid accuracy assessment over large areas was conducted. The analysis was expanded from the three specific continents were used last year: North America, Africa, and Australia to include Asia, Europe, and South America. Each continent offered unique challenges and problems related not only to reference data collection, but also for stratification used in sampling and proper balancing of the sampling by land cover type. A paper was submitted on the specific issues explored using North America, Africa, and Australia and a follow-up paper is in preparation for issues with every continent.

Publications

  • Type: Theses/Dissertations Status: Published Year Published: 2017 Citation: Fraser, Benjamin T. 2017. Evaluating the Use of Unmanned Aerial Systems (UAS) for Collecting Thematic Mapping Accuracy Assessment Reference Data in New England Forest Communities. MS Thesis, University of New Hampshire, Durham, NH. 128p.
  • Type: Journal Articles Status: Published Year Published: 2017 Citation: Sun, Peijun, Jinshui Zhang, Russell G. Congalton, Yaozhong Pan, and Xiufang Zhu. 2017. A quantitative performance comparison of paddy rice acreage estimation using stratified sampling strategies with different auxiliary indicators. International Journal of Digital Earth. DOI: 10.1080/17538947.2017.1371256.
  • Type: Journal Articles Status: Published Year Published: 2017 Citation: Sun, Peijun, Russell G. Congalton, Heather Grybas, and Yaozhong Pan. 2017. The impact of crop mapping error on the performance of upscaling agricultural maps. Remote Sensing. 9. DOI:10.3390/rs9090901.


Progress 10/01/15 to 09/30/16

Outputs
Target Audience:The target audience for most of the work conducted during this reporting period is the remote sensing / geospatial analysis community. Sampling for assessing the accuracy of maps created from remotely sensed imagery is central to the effective use of these maps for resource management. New technologies that allow the creation of segments or objects (groupings of pixels) instead of mapping individual pixels has vastly improved our mapping capabilities and new sampling methods must be used to assess these maps. In addition, mapping projects have continued to expand from mapping small, local areas to mapping countries and entire continents. Assessing the accuracy of maps over large areas also requires special considerations not appropriate for local area mapping. Most of the work performed this year has been presented at two workshops of the Global Food Security Support Analysis Data working group and has been used by this group for planning reference data collections for their global crop mapping effort. Three continents were concentrated on: North America, Africa, and Australia. Some additional work was performed on local sampling strategies and mapping of the Coastal Watershed of NH. This local work was presented in a poster at the University of New Hampshire Graduate Research Conference to a general audience of students, faculty and staff. Changes/Problems:The only change in this project is really an addition. There are a number of sources of low attitude imagery that could be used for reference data collection. We have been able this year to include UAS imagery as another source here which greatly enhanced this work. What opportunities for training and professional development has the project provided?A PhD graduate student worked on this project as part of her dissertation research. This student mentored an undergraduate student intern that also worked on this project. The undergraduate intern was funded as part of the project investigator's New Hampshire View project. NHView is part of the AmericaView Consortium. Two new MS students also joined this research. One will be using UAS to investigate their use for collecting the reference data needed for accuracy assessment. The other will be conducting an in-depth analysis of North America using a variety of different sampling approaches to gain understanding of the advantages and disadvantages of each and to select the most effective way. How have the results been disseminated to communities of interest?Insights gained as part of the results of this work have been published in two Book Chapters. One is on global agricultural mapping and the other is on methods for assessing the accuracy of thematic maps with some emphasis on large areas. What do you plan to do during the next reporting period to accomplish the goals?A lot of work was accomplished on both major project goals this year, but there is much left to accomplish. Work will continue both locally to investigate the potential for using UASs for collecting reference data for pixel-based and object-based mapping projects and on issues and considerations for large area mapping. We will extend our work to other large areas and use the knowledge gained this year to perform the most effective accuracy assessments.

Impacts
What was accomplished under these goals? Forest/land cover maps derived from remotely sensed imagery are routinely used for a variety of resource management and policy decisions. These maps must be of the highest accuracy possible and new methods for creating better maps are continually being developed. Therefore, techniques for assessing the accuracy of these maps are absolutely vital to the continued development and effective use of these maps. The goal of this project is to investigate, develop, and evaluate some new methods for assessing the accuracy of maps that extend over large geographic areas and/or are derived using a new object-based image analysis method. The successful development of these assessment methods will lead to improved evaluation of these maps that in turn will lead to better management and policy decisions that will benefit everyone. Research was conducted at both the local and large area scales during this reporting period. After completing the land cover map last year, work began this year to evaluate using low altitude imagery as a source of reference data. Unmanned Aerial Systems (UAS), also called drones, were used to collect imagery to assess both the pixel-based and object-based classification approaches. Significant effort was required to operationally use the UAS for data collection including meeting FAA (Federal Aviation Administration) requirements as well as the many technical and practical challenges of flying the drone. Work on the theoretical analysis and practical application of collecting reference data and conducting a valid accuracy assessment over large areas was conducted. Three specific continents were used this year in the analysis: North America, Africa, and Australia. Each large area offers it own challenges and problems related not only to reference data collection, but also for stratification used in sampling and proper balancing of the sampling by land cover type. A paper is in preparation describing these issues and our solutions to these challenges.

Publications

  • Type: Book Chapters Status: Published Year Published: 2016 Citation: Teluguntla, P., P Thenkabail, J. Xiong, M. Krishna Gumma, C. Giri, C. Milesi, M. Ozdogan, R. Congalton, J. Tilton, T. Sankey, R. Massey, A. Phalke, and K.Yadav. 2016. Global Food Security Support Analysis Data (GFSAD) at Nominal 1-km (GCAD) derived from Remote Sensing in Support of Food Security in the Twenty-first Century: Current Achievements and Future Possibilities. IN: Remote Sensing Handbook; Vol. II: Land Resources Monitoring, Modeling, and Mapping with Remote Sensing. P. Thenkabail (Editor). CRC/Taylor & Francis, Boca Raton, FL. pp. 131-159.
  • Type: Book Chapters Status: Published Year Published: 2016 Citation: Congalton, R. 2016. Assessing Positional and Thematic Accuracies of Maps Generated from Remotely Sensed Data. IN: Remote Sensing Handbook; Vol. I: Data Characterization, Classification, and Accuracies P. Thenkabail (Editor). CRC/Taylor & Francis, Boca Raton, FL. pp. 583-601.


Progress 10/01/14 to 09/30/15

Outputs
Target Audience:The target audience for much of the work conducted during this reporting period is the remote sensing community. Sampling for assessing the accuracy of maps created from remotely sensed imagery is central to the effective use of these maps for resource management. New technologies that allow the creation of segments or objects (groupings of pixels) instead of mapping individual pixels has vastly improved our mapping capabilities and new sampling methods must be used to assess these maps. In addition, mapping projects have continued to expand from mapping small, local areas to mapping countries and entire continents. Assessing the accuracy of maps over large areas also requires special considerations not appropriate for local area mapping. Most of the work performed this year has been presented at two workshops of the Global Food Security Support Analysis Data working group and has been used by this group for planning reference data collections for their global crop mapping effort. In addition, work on local sampling strategies and mapping of the Coastal Watershed of NH was presented in a poster at the University of New Hampshire Graduate Research Conference. Changes/Problems: Nothing Reported What opportunities for training and professional development has the project provided?A PhD graduate student worked on this project as part of her dissertation research. This student mentored an undergraduate student intern that also worked on this project. The undergraduate intern was funded as part of the project investigator's New Hampshire View project. NHView is part of the AmericaView Consortium. In addition, a MS student worked on mapping the Coastal Watershed of NH and completed her MS Thesis as part of this project. Finally, a visiting PhD student from China conducted his PhD research as part of this project and helped investigate some of the theoretical aspects of the sampling issues. How have the results been disseminated to communities of interest?The results of this work have been published in two articles in peer-reviewed remote sensing journals. In addition, an MS Thesis was produced on the Coastal Watershed mapping in NH. A poster of this thesis work was presented at the UNH Graduate Research Conference. What do you plan to do during the next reporting period to accomplish the goals?A lot of work was accomplished on both major project goals this year, but there is much left to accomplish. Work will continue both locally to investigate sampling strategies for collecting reference data for object-based mapping projects and on issues and considerations for large area mapping. The next year will specifically concentrate on special considerations for distributing the reference data samples appropriately over large areas to obtain valid accuracy assessment results.

Impacts
What was accomplished under these goals? Land cover maps of the Coastal New Hampshire Watershed were created from Landsat 8 imagery using both a pixel-based and an object-based image analysis approach. The accuracy of the maps was evaluated using different sampling strategies for the object-based analysis vs. the pixel-based one. This work was published in a MS thesis completed in the September of 2015. Further work is continuing that will incorporate the use of low altitude imagery for collecting the reference data inside the objects as a possible substitute for the ground collected data. Work began on reference data collection strategies for large (continental) mapping areas with a review and uncertainty analysis of the current state of the art. A paper was published on this topic in the journal called Remote Sensing. Some theoretical analysis was also started looking into some of the issues related to how positional accuracy impacts thematic accuracy. This work was also published in the journal called Remote Sensing.

Publications

  • Type: Journal Articles Status: Published Year Published: 2014 Citation: Congalton, R. J. Gu, K. Yadav, P. Thenkabail, and M. Osdogan. 2014. Global land cover mapping: A review and uncertainty analysis. Remote Sensing, 6, pp. 12070-12093; doi:10.3390/rs61212070
  • Type: Journal Articles Status: Published Year Published: 2015 Citation: Gu, J. R. G. Congalton, and Y. Pan. 2015. The impact of positional errors on soft classification accuracy assessment: A simulation analysis. Remote Sensing. 7, pp. 579-599; doi:10.3390/rs70100579
  • Type: Theses/Dissertations Status: Published Year Published: 2015 Citation: Ledoux, Lindsay. 2015. Evaluating Landsat 8 Satellite Sensor Data for Improved Vegetation Mapping Accuracy of the New Hampshire Coastal Watershed Area. MS. Thesis. University of New Hampshire. 110p.