Project Title: Landscape Structure is an Integrated Measure of Earth System Response to Climate Change

NASA Reference Number: 1995-GlobalCh00404

Faculty Advisor: Dr. Susan L. Ustin

Institution: Department of Land, Air and Water Resources
University of California, Davis
Davis, CA 95616

Signatures:

Date: March 15, 1996
 

Academic Progress

Since completing my coursework in the fall, I have been preparing for my qualifying exams in ecology. The guidelines for prepartion of the exam stress that the candidate should critically review, reflect upon and evaluate his/her understanding of ecology and related disciplines. To prepare I reviewed my coursework and research experiences to this point, including two papers I co-authored which are currently in press or in submission (Grossman et al, 1996 and Zhang et al, 1996, respectively.) In addition I have prepared in depth in several examination areas relevant to my fellowship work (Table 2). My examiners are top level researchers and teachers in their respective examination areas and are listed in Table 2. My PhD qualifying exam is scheduled for March 13, 1996. I passed the written portion of the exam in April, 1995.

Table 2. Exam Committee and Subject Areas for Eric W. Sanderson’s Qualifying Examination for Ph.D. in Ecology, Univeristy of California at Davis

Examiner	Subject Area
Dr. Micheal Barbour	Landscape Ecology
Dr. Conrad Bahre	Environmental Remote Sensing
Dr. Eliska Rejmankova	Wetland Ecology
Dr. Wesley Wallender	Quanitative Methods using Geostatistics and GIS
Dr. Theodore Foin	Principles of Ecology (written exam)

In addition to finishing my coursework and preparing for exams, I have also been particpating in the Program in College Teaching, a unique UC Davis program which offers practical education in instruction for future professors. Graduate student participants are matched up with faculty mentors to teach a course, attend roundtable meetings, and complete a series of activity contracts revolving around teaching and professional development for instructors. Teaching is very important to me, and I believe it will be important to my success as a global change researcher. The ability to communicate concerns related to global change and to prepare and inform the next generation of researchers and a concerned citzenry seem to me vital to NASA’s global change effort.
 

Research Progress

This work is nearly complete. A draft manuscript (Sanderson et al, 1996) is appended to this report for your review. I intend to submit the manuscript by the end of March to Landscape Ecology.

In addition I have also completed some preliminary analyses of AVIRIS quick looks. Quick looks are one-band images, averaged every two lines and two samples, available for every site AVIRIS imaged between 1992-1994, which constitutes over 4000 scenes from over 100 different sites throughout North America. Quick looks have several advantages for the survey of ecosystem structure including that they are small (0.5 Mb) and easy to manipulate and have a pixel size (35 m) which approximates the spatial resolution of current land observation satellite systems like Landsat -TM, -MSS and SPOT. They have a spatial extent of approximately 8.6 km by 10.8 km, for a total area of 9626 ha. The single band is centered at 700 nm (nominally 10 nm wide), which avoids major atmospheric absorbtions (Lillesand and Kiefer, 1994), yet is in a region where basic terrestrial materials, like water, soil/rock, vegetation, and snow/ice can generally be distinguished (Lillesand and Kiefer, 1994). Initially I had planned to do simple level slicing to classify the scenes into spectrally distinct landscape components for analysis. However when the quicklooks are produced by the Jet Propulsion Laboratory, each scene is individually stretched, meaning that a simple automated classification procedure will not be possible. I am currently working other algorithms to solve this problem. For this preliminary work, each scene was classed into twenty equal sized spectral ranges.

For the preliminary work reported here, I chose six quicklook scenes from disparte parts of North America representing distinct ecological landscapes. The scenes chosen for analysis are shown in Figure 1 with approximate positions (by latitude and longitude) and the dates of acquisition. All scenes were chosen from 1994. Clearly these scenes show different landscape features to the eye. The scene from near Hubbard Brook, N.H., for example, shows an even coverage of the forest canopy, which is bisected by two roads. In this case the pattern is based on a nearly uniform biological canopy broken up by an anthropogenic feature, the roadways. Similarly the Petaluma River, CA scene shows the influence of human beings on the landscape, predominantly in regularly shaped fields sloping down to the wetland vegetation along the river. The scenes from Atenango, Mexico and Organ Pipes, AZ, show predominantly geomorphological features. The Mexican shows a landscape dissected by a meandering river network, while the Arizona scene is dominated by alluvial fans and outwashes. The Palisade Glacier, CA, and Saskatchewan, Canada, scenes show landscapes with mixtures of vegetation, rock and soil. One of the goals of this work is to determine whether these kinds of mixtures are distinctive for the climatic conditions or ecological communities within those landscapes. Another goal is to evaluate the processes which are structuring landscapes, because clearly (even from the preliminary analysis presented here) different landscapes will have different susceptibilities to change under global climate change scenarios, depending on the processes responsible for creating the observed patterns.

Image histograms for these six quick looks are presented in Figure 2. These histograms represent the frequency distributions of reflected light in the quick look near-infrared waveband. At least on a coarse basis, these histograms represent the proportions of distinct landscape spectrally distinct landscape elements in each scene. For example the Hubbard Brook, N.H., shows a clustering of reflectance magnitudes corresponding to the dominant forest canopy in that scene. The Palisade Glacier, CA, histogram shows a bimodal response, representing two distinct classes of landscape elements which are spectrally distinguishable.

For initial processing, I analysed each scene using the Fragstats software available from Oregon State University (McGarigal and Marks, 1994). The metrics calculated by this software are designed mostly for fragmentation studies, but several metrics overlap with in my work in landscape characterization. In addition to the metrics calculated by Fragstats, I am writing programs to evaluate landscape structure using various scale and texture metrics, including: spatial autocorrelation and blocking correlation for scale (Turner et al, 1991) and angular second moment (ASM) and inverse difference moment (IDM) for texture (Musick and Grover, 1991; Haralick, 1979).

The metrics calculated in this preliminary study were number of patches, mean patch size (ha), landscape shape index, double log fractal dimension, Shannon’s diversity and evenness indicies, and contagion and interspersion/juxtaposition. Each of these metrics is more extensively described in McGarigal and Marks (1994), but may be briefly summarized as falling into one of several classes. Number of patches and mean patch size are measures of patch density and size and are self-explanatory. Landscape shape index and fractal dimension are patch shape indicies based on perimeter to area ratios. Landscape shape index increases as patch shapes deviate from a regular geometric shape, in this case, a square. Fractal dimension is a measure of the complexity of the shapes (Milne, 1991). Shannon’s diversity and evenness indicies are analagous to indicies used in measuring species diversity and evenness, except in this case the indicies are calculated over number and distributions of patch types. Finally contagion and interspersion/juxtaposition are measures of the connectivity of areas of a similar type. Contagion is calculated on a pixel basis; interspersion/juxtaposition is calculated on a patch basis.

Table 3 show the values of these metrics for the six preliminary scenes. Prior to analysis, each scene was classed into twenty classes based on amount of reflectance at 700 nm. We have only just begun analysis of these data, so conclusions at this point are tentative at best. Clearly contagion varies between landscapes, reaching a maximum for Hubbard Brook, N.H., and a minimum for Palisade Glacier, CA. Other metrics appear to vary little

Table 3. Landscape Structure Metrics Calculated for Six AVIRIS Quicklook Scenes Classed into Twently Reflectance Categories (Preliminary Results)

	Quicklooks Scenes
Metrics	Saskatch.,Canada	Atenango, Mexico	Petaluma River, CA	Hubbard Brook, NH	Palisade Glacier,CA	Organ Pipes, AZ
Number of Patches	51311	48848	39229	22848	45288	49133
Mean Patch Size (ha)	0.188	0.197	0.245	0.421	0.213	0.196
Landscape Shape Index	122.483	119.736	109.409	88.564	114.339	120.891
Double Log Fractal Dimension	1.560	1.539	1.496	1.521	1.480	1.560
Shannon’s Diversity Index	2.834	2.821	2.714	2.105	3.010	2.846
Shannon’s Evenness Index	0.931	0.927	0.891	0.703	0.989	0.935
Contagion (%)	9.040	10.014	15.003	33.759	6.178	10.169
Interspersion/ Juxtaposition (%)	90.139	89.359	85.143	69.155	94.642	88.576

 

between scenes and it is not clear at this point whether they will be useful in describing distinct landscape structures. It is surprising how similar scenes from boreal Canada and semi-tropical Mexico are when viewed by these metrics. Such similarities indicate that the main difficulty in my task will not be calculating metrics, but in analysing their meaning in terms of the processes on the ground.

A number of papers related to my fellowship research have been published this last year, indicating that this is a timely and relevant research program. For example, Ritters et al (1995) published a factor analysis of 26 landscape structure metrics from the literature. They found that many of these metrics measure similar properities of the landscape and suggest that six metrics (average perimeter-area ratio, contagion, standardized patch shape, patch perimeter-area scaling, number of classes, and large-patch density-area scaling) may usefully summarize landscape structure. Benson and Mackenzie (1995) published a study of the effect of sensor spatial resolution on the landscape structure parameters for a lake district in Wisconsin. They found that several metrics, including fractal dimension, average patch size and contrast, varied with different sensor resolutions. They found that texture measures were largely scale invariant. My research will further this area of research by examining how spectral and spatial resolution interact when calculating landscape structure metrics using full AVIRIS data cubes. Finally Baker (1995) published a paper discussing changes in disturbance regime caused by climate change might affect landscape structure. His analysis indicates that global warming effects may hasten the response of landscapes to disturbance if the effect of global warming is to shorten the rotation time of the disturbance regime. As futher developments are made in the study of landscape structure and responses of landscape structure to globabl climate change, I will incorporate the latest findings and thinking into my own work.
 

References