Ecological Risk Assessment at the Regional Scale1

R. L. Graham,2 C. T. Hunsaker,2 R. V. O'Neill,2 and B. L. Jackson3

2Environmental Sciences Division
Oak Ridge National Laboratory
Oak Ridge, TN 37831-6038

3Computing and Telecommunications Division
Oak Ridge National Laboratory
Oak Ridge, TN 37831-6038

November 16, 1990

Supplementary publication to Ecological Applications

_________________________

1Research was sponsored by the Office of Ecological Processes and Effects Research, U.S. Environmental Protection Agency, Washington, D.C., under EPA Interagency Agreement DW89932112-01-2 with the U.S. Department of Energy under contract DE-AC05-84OR21400 with Martin Marietta Energy Systems, Inc. This research has not been subjected to EPA review and therefore does nor necessarily reflect the views of EPA, and no official endorsement should be inferred.

The purpose of the computer programs for this paper is to simulate the effects of bark beetle infestation on forests in the Adirondacks area. There are of two different ozone scenarios considered, low and high. The programs calculate the resulting change in several landscape endpoints for 66 watersheds and ph for 66 lakes in the region. Landuse data for Adirondacks is available from USGS (scenes Glen Falls and Lake Champlain) in 200m x 200m resolution. Watershed boundaries for this area were obtained for the same resolution.

There are two executable programs.

The LAKEPATCH program is written in Fortran 77 for the IBM/MVS operating system. The functions of this program include the following:

1) Assign initial values to arrays which define the parameters of the two ozone scenarios (low and high). See the section in Graham [1] under 'Exposure.' and 'Effects.' for a description of the low and high ozone scenarios. The probability of an initial bark beetle attack on a coniferous or mixed forest pixel is .015 for the low ozone scenario and .04 for the high ozone scenario. See the main program listing for the parameter and array values. See Graham [1], figure 2, for histograms of the probability distributions of patch sizes.

2) Read in the landuse data and watershed data and overlay them. See the main program listing for the UTM boundaries of the geographic data. The possible coded landuse types of interest are the following:

LANDTYPE CODE
urban 11 - 17
agricultural 21 - 22
rangeland 31 - 33
forest
deciduous 41
conifer 42
mixed 43
water 51 - 54
wetland 61 - 62
barren land 71 - 77
tundra 81 - 85
snow or ice 91 - 92

3) Calculate endpoints for the landuse data before bark beetle infestation. The endpoints include the following:

a) proportions of different kinds of forest

b) dominance and contagion where
dominance = ln(N) + SUMk(Pk * ln(Pk))

contagion = 2*N*ln(N) + SUMijk(Pijk * ln(Pijk))
where N is the number of landuses that occur Pk is the proportion of each landuse, k=1,N Pijk is the proportion of 3-cell patterns of landuses c) proportion and number of pixels for each landtype d) edge counts of different kinds of forest (deciduous, conifer, and mixed) with non-forest landtypes e) number of interior forest pixels f) several indices formulated by Haralick g) percentages of each type of landuse for each watershed 4) Write out all the endpoints to flat files. 5) Simulate a bark beetle infestation by randomly selecting epicenters of infested patches. Find the extents of the patches according to the probabilities defined in the ozone scenario parameters. See Graham [1], figures 3 and 4, for representations of the infested patches. 6) Calculate all the endpoints described in 3) for the altered landscape. 7) Write out all the endpoints to flat files. 8) Do 5), 6), and 7) 100 times. Each iteration results in a 'Monte Carlo' simulation of a bark beetle attack. 9) Finish. The LAKESAS program is written in SAS version 5.18 for the IBM/MVS operating system. It reads in the flat files that were created by LAKEPATCH. The functions of this program include the following: 1) Read in all endpoints calculated by LAKEPATCH and store them in a permanent SAS dataset. Identify the records according to whether they are for the base case, or the Monte Carlo cases 2) Get watershed observations from an existing SAS database ('original' case). 3) Calculate ph of the lakes in each watershed according to a regression equation. See the SAS program listing for the regression coefficients and model definition. 4) Calculate average ph for Monte Carlo values. Compare these averages with ph for base case and original case. 5) Produce a table that shows lakes that have a shift in average Monte Carlo ph of greater than .2 from the base case ph. 6) Produce a table that identifies lakes that have a shift of ph from less than 5.5 to more than 5.5 from the base case ph to the average Monte Carlo ph. 7) Save all data of interest in permanent SAS data members. 8) Finish.

REFERENCE

R. L. Graham, C. T. Hunsaker, R. V. O'Neill, B. L. Jackson, "Ecological Risk Assessment at the Regional Scale", Ecological Applications 1:196-206.