Please use this identifier to cite or link to this item:
Image Evaluation of Landsat Data for Monitoring Change in Forest Vigor: The Ohia Rain Forest Decline on the Island of Hawaiʻi
|uhm_phd_8111345_r.pdf||Version for non-UH users. Copying/Printing is not permitted||6.65 MB||Adobe PDF||View/Open|
|uhm_phd_8111345_uh.pdf||Version for UH users||6.59 MB||Adobe PDF||View/Open|
|Title:||Image Evaluation of Landsat Data for Monitoring Change in Forest Vigor: The Ohia Rain Forest Decline on the Island of Hawaiʻi|
|Authors:||Rockie, John D.|
|Advisor:||Wingert, Everett A.|
show 4 moreHawaii
forests and forestry
earth resources technology satellites
|Issue Date:||Dec 1980|
|Publisher:||[Honolulu] : [University of Hawaii at Manoa], [December 1980]|
|Abstract:||A concern among biologists and foresters in Hawaii in recent years has been the decline (dieback) of ohia (Metrosideros) rain forests, particularly on the island of Hawaii where the decline has reached epidemic proportions. Various explanations, theories and hypotheses have been advanced for the cause of the decline. Regardless of the cause(s), to effectively monitor this phenomenon necessitates periodic mapping of the areal extent of decline.|
This study evaluates LANDSAT images for mapping variations in rain forest vigor/density in order to monitor ohia decline on the island of Hawaii. Two LANDSAT scenes from February 1973 and July 1975 provided the basic data used in considering a 5,000 hectare study area. Several types of image enhancement were evaluated, including density slicing, color enhancement, multispectral comparisons, densitometry and change detection. Density slicing received primary emphasis because of objectivity and repeatability.
Density slices were made of the four MSS bands of each image, showing variations in radiance as recorded by MSS sensors for the various combinations of ohia canopy and understory. These were compared with classifications based on underflight aerial photographs, supplemented by ground and helicopter observations. Four classes of ohia were stratified, based on combinations of stand vigor and density. The ohia classes ranged from class 1 (dense, healthy ohia) to class 4 (dense ohia with very severe decline; closed ohia with severe decline; or open ohia with moderate or greater decline). Grid overlay comparisons were made between density slices for each LANDSAT image and the classification from aerial photographs for tile respective year, using the classification as the 'true' condition of ohia vigor and stand density. Two statistics were used, a coefficient of areal correspondence and a coefficient of contingency, each ranging from 1.00 for complete areal correspondence (complete association) to zero for no correspondence (complete independence). The 1973 data ranged from 0.40 to zero; for 1975, from 0.44 to zero. These results were then considered by ohia class and LANDSAT band. Class 2 discrimination was slightly better than class 1, followed by class 3, with class 4 much lower. Band 6 correlation was highest, with bands 5 and 7 only slightly lower; band 4 results were insignificant because of low density contrast in the image. To help explain the results several methods were used, including digital analysis, subarea comparisons and slight shifting of grid overlays.
The methods of enhancement and interpretation used in addition to density slicing gave negligible results except for color enhancement. A color composite image was interpreted to discriminate between areas of varying ohia vigor. The method was considered nearly comparable in reliability and timeliness to density slicing.
Conclusions: 1. Principal reasons for low correspondence levels were (a) heterogeneity of both canopy and understory of the rain forest; (b) inaccuracies in generalized vegetation maps used to evaluate density slice maps; and (c) planimetric errors in LANDSAT images due to both geometric errors and misleading pixel portrayal. 2. The change in ohia decline patterns between February 1973 and July 1975 was too small to be measured by either enhancement of LANDSAT images or interpretation of vegetation maps compiled from aerial photographs. 3. The density slicing method employed in this study was in itself sound--the problem was the variable nature of the ohia rain forest.
The problem studied proved to be on the margins of LANDSAT capability, due both to the limitations of LANDSAT sensitivity and the strong constraints of the unique phenomenon being studied. Despite the heterogeneous character of the ohia rain forest and the resulting difficulties in analysis, valid analytical procedures were developed that could be applied to vegetation stress studies involving more homogeneous species.
|Description:||PhD University of Hawaii at Manoa 1980|
Includes bibliographical references (leaves 193–201).
|Pages/Duration:||xii, 201 leaves, bound : illustrations (some color), maps ; 28 cm|
|Rights:||All UHM dissertations and theses are protected by copyright. They may be viewed from this source for any purpose, but reproduction or distribution in any format is prohibited without written permission from the copyright owner.|
|Appears in Collections:||Ph.D. - Geography|
Please contact email@example.com if you need this content in an ADA compliant alternative format.
Items in ScholarSpace are protected by copyright, with all rights reserved, unless otherwise indicated.