IFMAP Remote Sensing Report - DNR

Integrated Forest Monitoring Assessment and Prescription
IFMAP
Review of Remote Sensing Technologies for the IFMAP Project 2nd Revision
Prepared for: State of Michigan
Michigan Department of Natural Resources
Revised September 21 2001
| |Prepared by: |
| |Space Imaging Solutions |
| |Pacific Meridian Resources |
| |455 East Eisenhower Parkway, |
| |Suite 70, |
| |Ann Arbor, MI 48108 | |[pic] |
CONFIDENTIAL NOTICE The information contained in this report is proprietary and confidential.
This report and its contents may not be used, duplicated, communicated, or
disclosed, in whole or in part without the express written permission of
the Michigan Department of Natural Resources.
Michigan Department of Natural Resources IFMAP
Remote Sensing Report |Table of Contents |
Executive Summary 6
1.0 Introduction 7 1.1 Background 7
1.2 Purpose of this Document 8
1.3 Definition of Terms 8
1.4 Content of the Report 9
1.5 Classification System 9
1.5.1 Development of classification system 9
1.5.2 Development of rules 11
1.5.3 Classes 12
1.5.4 Timetable for rest of State 12 2.0 Coarse Resolution Map 13 2.1 Goals and Objectives 13
2.1.1 Goals 13
2.1.2 Functional Objectives 13
2.2 Imagery Source 14
2.2.1 Sensors 14
2.2.2. Timing/Seasonality of imagery 14
2.2.3 Imagery acquired for Southern Michigan 15
2.2.4 Imagery acquisition approach for remainder of state 16
2.3 Methodology 16
2.3.1 Stratification of training sites 16
2.3.2 Identification of training areas 17
2.3.3 Field training 17
2.3.4 Collection of training sites 18
2.3.5 Digitizing of training sites 18
2.3.6 Classification approach 19
2.3.7 Editing of raster 23
2.3.8 Draft Map Review 23
2.3.9 Delivery of final map 23
2.4 Accuracy Assessment Approach 24
2.4.1 Minimum mapping unit for verification 24
2.4.2 Potential for use of FIA 24
2.4.3 Error analysis, point selection and statistical approach 25
2.5 Other Issues investigated 28
2.5.1 Use of radar 28
2.6 Coarse-Resolution Classification of the Pilot Area 29
2.6.1 Goals 29
2.6.2 Imagery and Training Site Selection 29
2.6.3 Accuracy Assessment 29 3.0 High Resolution Map 31 3.1 Goals and objectives 31
3.1.1 Overall Goals 31
3.1.2 Specific Functions 31
3.1.3 Changes in methodology during the project 31
3.2 Sensor Evaluation Overview 32
3.3 Methodology 32
3.3.1 Species Modeling 32
3.3.2 Size Class 36
3.3.3 Canopy Closure 36
3.3.4 Stand delineation 37
3.4 Space Imaging - Digital Airborne Imaging System 39
3.4.1 Approach 39
3.4.2 Classification 39
3.4.3 Segmentation 42
3.4.4 Costs 43
3.4.5 Delivery 43
3.4.6 Operational considerations 43
3.4.7 Conclusions 44
3.5 DCS430 - Emerge 46
3.5.1 Approach 46
3.5.2 Classification 46
3.5.3 Segmentation 48
3.5.4 Costs 49
3.5.5 Delivery 49
3.5.6 Operational considerations 49
3.5.7 Conclusions 49
3.6 IKONOS - Space Imaging 51
3.6.1 Approach 51
3.6.2 Segmentation 51
3.6.3 Costs 53
3.6.4 Delivery 53
3.6.5 Operational considerations 53
3.7 AISA - 3DI 54
3.7.1 Approach 54
3.7.2 Classification 55
3.7.3 Bispectral Plots 56
3.7.4 Segmentation 59
3.7.5 Costs 59
3.7.6 Delivery 59
3.7.7 Operational considerations 59
3.8 Overall Conclusions 60
3.8.1 Approach 60
3.8.2 Classification 60
3.8.3 Segmentation 61
3.8.4 Costs, operational considerations and recommendations 63 4.0 Summary and conclusions 65 4.1 Classification scheme 65
4.2 Coarse resolution map 66
4.3 High resolution map 66 Bibliography 68
Appendix A 69
Appendix B 70
Appendix C 71
Appendix D 72
Appendix E 73
Appendix F 74
Appendix E 75
Appendix F 76
Appendix G 77
Appendix H 78
Appendix I 79
Executive Summary
This report is an evaluation and summary of work conducted by Pacific
Meridian Resources/ Space Imaging Solutions with respect to the remote
sensing component of the IFMAP project. It also lays out the future
direction of the remote sensing component of the project. Remote sensing is
s central component of this project and aims to produce both produce both a
coarse resolution Landsat TM derived land cover map of the whole of the
State of Michigan and a high resolution land cover map for lands managed by
the Michigan Department of Natural Resources. To achieve these objectives
different sensors and techniques have been evaluated. To develop a land cover map a classification scheme needs to be developed.
The project has developed a complex hierarchical classification scheme that
is based on canopy characteristics. This classification scheme was arrived
at after extensive consultation with field staff in all ecoregions. The
classification scheme starts with the major land cover types and subdivides
these types into more detailed classes. The natural land classes have the
greatest detail with forest type associations at Level 3 detailed land
cover classes at Level 4. The Level 4 class label also has a stand size
variable and a canopy closure variable. The final landscape stratification
that will be used for the calculation of the stratified inventory will fall
somewhere between Level 3 and Level 4. The classification system is
consistent across spatial scales. The statewide map will be built from the analysis of triple date Landsat
imagery, this imagery is taken in spring, summer and fall, which allows the
use of seasonal differences between cover types to aid interpretation. A
combination of supervised and unsupervised classification techniques have
been used to classify Southern Michigan and will be applied to the rest of
the State. Four types of high resolution imagery were captured over the pilot area in
Grand Traverse and Wexford Counties. Each of these types of imagery has
been evaluated for its utility for the IFMAP Project. A number of methods
have been investigated that classify and segment the imagery. Although the
results have proved useful, they do not capture the level of detail desired
by the field, likewise the delineation of stands is sufficiently different
from what foresters are used to looking at that it was decided that
automated processing of the high resolution imagery was not going to
generate the information field forester require. The imagery and
classification methods used are detailed in the report. Consequently at a meeting of the IFMAP advisory group on September 18th
2000 it was recommended and decided to move forward on a plan that would
use manual delineation methods over the digital imagery, and field labeling
of these polygons. This method would be facilitated with the GDSE software
and would be within a quality-controlled system. As a result of this investigation Emerge Imagery was acquired over 767,000
acres over the State of Michigan this summer, that will be used for mapping
in the coming year.
1.0 Introduction 1.1 Background
The use of remotely sensed data is an integral part of the Michigan
Department of Natural Resources IFMAP Project, and will be incorporated
into the project wherever it will provide advantages in cost or efficiency
while maintaining accuracy required for decision making. Remote sensing
will provide information at two scales of analysis: an ecosystem scale
where data will aid management of DNR lands within the context of their
ecosystem, and an operational scale: where data will aid the day to day
management of DNR by foresters, wildlife biologists and other DNR
personnel. Remote sensing at an ecosystem scale will help produce a map of land cover
over all ownerships within the State of Michigan (Figure 1). This map will
be used in combination with stratified inventory data over all ownerships
measured by the United States Department of Agriculture, Forest Service,
(USDA, FS). These data are collected as part of the Forest Inventory and
Analysis Program (FIA). The combination of the statewide maps and the FIA
inventory data will allow the production of a variety of maps and reporting
products that can help answer landscape scale, and statewide planning
questions. The detail of this map will not support specific operational
decisions. More details on the use of FIA and other inventory data are
presented in the companion document "IFMAP; Inventory Report, January
2001". Since the resolution of the statewide map will not support operational
decision making it is necessary to develop information at a higher
resolution, in terms of the maps and the information stored in them. It is
intended to use higher resolution imagery (1 m vs. 30 m) to support the
development of a high-resolution land cover map. This map will support the
standardized field survey that will gather location specific information
that can be used for management. Details of the high-resolution inventory
will not be covered in this report but are given in the companion document
"IFMAP; Inventory Report, January 2001". This report summarizes the remote sensing