The effects of point or polygon based training data on randomForest classification accuracy of wetlands

Jennifer Corcoran, Joe Knight, Keith C Pelletier, Lian P Rampi, Yan Wang

Research output: Contribution to journalArticle

13 Citations (Scopus)

Abstract

Wetlands are dynamic in space and time, providing varying ecosystem services. Field reference data for both training and assessment of wetland inventories in the State of Minnesota are typically collected as GPS points over wide geographical areas and at infrequent intervals. This status-quo makes it difficult to keep updated maps of wetlands with adequate accuracy, efficiency, and consistency to monitor change. Furthermore, point reference data may not be representative of the prevailing land cover type for an area, due to point location or heterogeneity within the ecosystem of interest. In this research, we present techniques for training a land cover classification for two study sites in different ecoregions by implementing the RandomForest classifier in three ways: (1) field and photo interpreted points; (2) fixed window surrounding the points; and (3) image objects that intersect the points. Additional assessments are made to identify the key input variables. We conclude that the image object area training method is the most accurate and the most important variables include: compound topographic index, summer season green and blue bands, and grid statistics from LiDAR point cloud data, especially those that relate to the height of the return.

Original languageEnglish (US)
Pages (from-to)4002-4025
Number of pages24
JournalRemote Sensing
Volume7
Issue number4
DOIs
StatePublished - Jan 1 2015

Fingerprint

polygon
wetland
land cover
ecoregion
ecosystem service
GPS
ecosystem
summer
effect

Keywords

  • LiDAR
  • Object based image analysis
  • Optical and infrared sensors
  • Segmentation
  • Topographic
  • Wetlands

Cite this

The effects of point or polygon based training data on randomForest classification accuracy of wetlands. / Corcoran, Jennifer; Knight, Joe; Pelletier, Keith C; Rampi, Lian P; Wang, Yan.

In: Remote Sensing, Vol. 7, No. 4, 01.01.2015, p. 4002-4025.

Research output: Contribution to journalArticle

@article{5707f9ab21e947698cdc4b64b8e34a35,
title = "The effects of point or polygon based training data on randomForest classification accuracy of wetlands",
abstract = "Wetlands are dynamic in space and time, providing varying ecosystem services. Field reference data for both training and assessment of wetland inventories in the State of Minnesota are typically collected as GPS points over wide geographical areas and at infrequent intervals. This status-quo makes it difficult to keep updated maps of wetlands with adequate accuracy, efficiency, and consistency to monitor change. Furthermore, point reference data may not be representative of the prevailing land cover type for an area, due to point location or heterogeneity within the ecosystem of interest. In this research, we present techniques for training a land cover classification for two study sites in different ecoregions by implementing the RandomForest classifier in three ways: (1) field and photo interpreted points; (2) fixed window surrounding the points; and (3) image objects that intersect the points. Additional assessments are made to identify the key input variables. We conclude that the image object area training method is the most accurate and the most important variables include: compound topographic index, summer season green and blue bands, and grid statistics from LiDAR point cloud data, especially those that relate to the height of the return.",
keywords = "LiDAR, Object based image analysis, Optical and infrared sensors, Segmentation, Topographic, Wetlands",
author = "Jennifer Corcoran and Joe Knight and Pelletier, {Keith C} and Rampi, {Lian P} and Yan Wang",
year = "2015",
month = "1",
day = "1",
doi = "10.3390/rs70404002",
language = "English (US)",
volume = "7",
pages = "4002--4025",
journal = "Remote Sensing",
issn = "2072-4292",
publisher = "Multidisciplinary Digital Publishing Institute (MDPI)",
number = "4",

}

TY - JOUR

T1 - The effects of point or polygon based training data on randomForest classification accuracy of wetlands

AU - Corcoran, Jennifer

AU - Knight, Joe

AU - Pelletier, Keith C

AU - Rampi, Lian P

AU - Wang, Yan

PY - 2015/1/1

Y1 - 2015/1/1

N2 - Wetlands are dynamic in space and time, providing varying ecosystem services. Field reference data for both training and assessment of wetland inventories in the State of Minnesota are typically collected as GPS points over wide geographical areas and at infrequent intervals. This status-quo makes it difficult to keep updated maps of wetlands with adequate accuracy, efficiency, and consistency to monitor change. Furthermore, point reference data may not be representative of the prevailing land cover type for an area, due to point location or heterogeneity within the ecosystem of interest. In this research, we present techniques for training a land cover classification for two study sites in different ecoregions by implementing the RandomForest classifier in three ways: (1) field and photo interpreted points; (2) fixed window surrounding the points; and (3) image objects that intersect the points. Additional assessments are made to identify the key input variables. We conclude that the image object area training method is the most accurate and the most important variables include: compound topographic index, summer season green and blue bands, and grid statistics from LiDAR point cloud data, especially those that relate to the height of the return.

AB - Wetlands are dynamic in space and time, providing varying ecosystem services. Field reference data for both training and assessment of wetland inventories in the State of Minnesota are typically collected as GPS points over wide geographical areas and at infrequent intervals. This status-quo makes it difficult to keep updated maps of wetlands with adequate accuracy, efficiency, and consistency to monitor change. Furthermore, point reference data may not be representative of the prevailing land cover type for an area, due to point location or heterogeneity within the ecosystem of interest. In this research, we present techniques for training a land cover classification for two study sites in different ecoregions by implementing the RandomForest classifier in three ways: (1) field and photo interpreted points; (2) fixed window surrounding the points; and (3) image objects that intersect the points. Additional assessments are made to identify the key input variables. We conclude that the image object area training method is the most accurate and the most important variables include: compound topographic index, summer season green and blue bands, and grid statistics from LiDAR point cloud data, especially those that relate to the height of the return.

KW - LiDAR

KW - Object based image analysis

KW - Optical and infrared sensors

KW - Segmentation

KW - Topographic

KW - Wetlands

UR - http://www.scopus.com/inward/record.url?scp=84937941472&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84937941472&partnerID=8YFLogxK

U2 - 10.3390/rs70404002

DO - 10.3390/rs70404002

M3 - Article

VL - 7

SP - 4002

EP - 4025

JO - Remote Sensing

JF - Remote Sensing

SN - 2072-4292

IS - 4

ER -