Distributed Land-Cover Change Simulation
and Multidimensional Interpolation

M.W. Berry and K.S. Minser
Department of Computer Science
University of Tennessee
Knoxville, TN, 37996-1301
[berry,minser]@cs.utk.edu

Hoh Watershed

Statement of interest submitted to the Land Use Modeling Workshop at the USGS EROS Data Center, Sioux Falls, SD, June 5-6, 1997.


Statement of Interest:

LUCAS Computer simulations are used in landscape ecology to simulate the effects of human land-use decisions on the environment. Such decisions are influenced by both ecological and socioeconomic factors which can be represented by spatially explicit multidisciplinary data. With support from the U.S. Man and the Biosphere (MAB) program, we have developed (through a collaboration with several ecologists, economists, sociologists, and foresty personnel) the Land-Use Change Analysis System (or LUCAS) for the study of land-use effects on landscape structure in such areas as the Little Tennessee River Basin in western North Carolina and the Olympic Peninsula of Washington state. These effects include land-cover change and species habitat suitability. Using a geographic information system (GIS) to store, display and analyze map layers derived from remotely sensed images, census and ownership maps, topological maps, and output from econometric models, a parallel/distributed version of LUCAS (pLUCAS) was developed for simulations on a network of workstations. Targeting distributed computational environments reflects the resources available to most land-use planners, forestry personnel, and wildlife managers. We have recently conducted a formal performance evaluation of two pLUCAS distributed models on an ATM-based network of 12 SUN Ultra-2 workstations. Speed improvement factors as high as 8 (relative to serial runs on a single SUN Ultra-2 workstation) have been obtained using the PVM or MPI message-passing environments.

IMP As part of the Integrated Modeling Project (IMP) of Southern Global Change Program, we are responsible for the development of the (second) IMP module which facilitates the horizontal integration of forest responses to environmental stresses and disturbances through the use of micro-scale cellular automata. This module is being developed from the LUCAS modeling system prototype. Stochastic attributes used by LUCAS will incorporate the frequency distributions of output results generated by the IMP's Linked Dynamic Model. Overall focus of the Integrated Modeling Project (IMP) is to integrate forest health and productivity assessments of southern and southeastern forests by taking into account changing climate, air quality, and land use changes.

In order for LUCAS to aggregate site index and forest growth types (height vs age) attributes to forest stands across the southern region, response surfaces must be developed from selected outputs of the codes comprising the Linked Dynamic Model. Using 5-dimensional interpolation based on the Modified Shepard's Method developed by Robert Renka (Univ. of North Texas) we are building a portable object-oriented (C++) software package that will produce interpolated values for the

  1. mean,
  2. standard deviation,
  3. minimum value,
  4. maximum value, and
  5. distribution type (normal, lognormal, uniform, etc.)
of any model output (e.g, site index or the height of of dominant trees at age 25) based on the following independent variables (or conditions):
  1. atmospheric carbon dioxide,
  2. ozone exposure,
  3. nitrogen deposition,
  4. temperature, and
  5. precipitation.
The hypervolumes or response surface data produced from the proposed activity will be constructed using the netCDF (network Common Data Form) which is a popular machine-independent format for representing GIS and other scientific data. These hypervolumes will be used to integrate the influence of environmental factors on the forest conditions modeled by the codes in the IMP's Linked Dynamic Module. These hypervolumes may also be of great use by other researchers (outside of the IMP project) in the study of forest growth and production in the southeastern US.

In addition to the development of the 5-dimensional hypervolumes for the IMP driving variables, we will be porting the current MPI-based pLUCAS prototype to a newly acquired IMP SP/2 (total of 40 processors). Cycles on this machine will be available for future pLUCAS-based simulations associated with Module II of the IMP project.

Credits:


Our research in computational ecology has been supported by the Southeastern Appalachian Man and the Biosphere (SAMAB) Program under U.S. State Department Grant No. 1753-000574 and University of Washington Subcontract No. 392654, by the National Science Foundation under grants NSF-ASC-94-11394 and NSF-CDA-95-29459, and the USDA Forest Service under Contract Nos. 29-1286-96 and SRS-CA-96-067.



References:

M. W. Berry, R. O. Flamm, B. C. Hazen, and R. L. MacIntyre. Lucas: A System for Modeling Land-Use Change. IEEE Computational Science and Engineering, 3(1):24-35, June 1996.

B. C. Hazen. A Distributed Implementation of the Land-Use Change Analysis System (LUCAS) Using PVM. Master's thesis, University of Tennessee, Knoxville, August 1995.

B. C. Hazen and M. W. Berry. The Simulation of Land-Cover Change Using a Distributed Computing Environment. Simulation Practice and Theory, 1997. In Press.

R. J. Renka. Multivariate Interpolation of Large Sets of Scattered Data. ACM Trans. on Math. Soft. 14(2):139-148, June 1988.


Michael Berry
Tue May 13 20:58:03 EDT 1997