Snow Model for Yellowstone National Park

Application of a Snow Model for Yellowstone National Park

4/23/2002

Gary Wockner, Francis Singer, Mike Coughenour, Phil Farnes

This document contains a description and instructions for the Yellowstone Snow Model.

Introduction

Yellowstone National Park (YNP) managers have increasing needs for a landscape level method to predict snow accumulations in YNP during winter. Snow pack may influence stream runoff, plant phenology and plant production during the subsequent growing season (Merrill et al. 1994, Coughenour and Singer 1996b). Snow packs may influence the habitats that ungulates feed in, their abilities to travel, their migration timing, their vulnerability to predators and their energetic expenditures during winter (Brunnel et al 1999, Huggard 1993, Hobbs 1989, Geese and Grothe 1995, Farnes et al. 1999, Cheville et al 1999.) Snowpacks may determine which areas elk or wolves may occupy during winter (Briggs 1995). Snow affects ungulate foraging, and this effect has been included in spatially explicit calculations of elk carrying capacity (Coughenour 1994,1996). Bison leaving the park create management problems, and snow cover in conjunction with population size could be used to predict when bison migrations from the park occur (Cheville et al. 1998). A 5-year multidisciplinary project on bison ecology and brucellosis in Yellowstone coordinated by Pete Gogan is nearing completion, and snow cover effects on bison distributions are being incorporated into a model of bison carrying capacity and emmigration (M. Coughenour, ms. in prep.) More recently, the reintroduction of wolves into YNP in 1995 has generated the scientific need to determine past and present snow packs in order to differentiate sort out any potential effects of wolves on ungulate distributions from effects due to variations in snow cover.

Snow hydrologists have been generating snowfall isopleth maps for many years by manually combining their knowledge of the impact of topography on precipitation, topographic maps, and with point-based weather station and snow depth data. Running et al. (1987) used a simple ratio of data on an isopleth map to the long term base station average to compute precipitation at any given site from base station values, but the outputs were not mapped. Methods for generating maps from point data for precipitation, snowfall, and snow depth have been developed during the last decade or so. In addition to SAVANNA (Coughenour 1991), more specialized models include MT-CLIM (Hungerford 1989), ANUSPLIN (Hutchinson 1989, 1995), and PRISM (Daley et al 1994).

Coughenour (1994,1996) combined GIS and snow depth data to generate bi-weekly snow depth maps for the purpose of estimating elk carrying capacity on Yellowstone’s northern range. The idea of developing a GIS‑based snow model for Yellowstone’s northern range has also been a goal of Phil Farnes (Snowcap Hydrology, Bozeman, MT.) for a number of years (Farnes and Romme 1993, Farnes 1996, Biggs 1995). Although snow features such as crusting density and depth hoar can influence ungulates, Farnes has argued that snow water equivalent (SWE) is the single best measure that incorporates most other snow pack features and can be used to predict ungulate responses (Farnes et al. 1999). Farnes et al. (2000) state, “It has been proposed that GIS be used to determine monthly SWE maps for the northern range for the previous 50 years using SWE values and algorithms generated for the report”.

The purpose of this effort was to respond to that recommendation and to develop the GIS‑based component of an operational snow model that could be used to predict SWE at any location on the northern range on any day. Thus SWE can be predicted for the visual or radiotelemetry generated locations of ungulates, or locations of wolves. Our study team makes this GIS model available for park staff and for the other researchers currently evaluating the factors that determine ungulate distributions.

Data Model Background

A data model consists of a highly integrated combination of a data set, and a model which uses the data to generate useful information. In this case, point data for snow water equivalents are used by a model to generate maps of snow water equivalents across a region. As such, the data and the model are both equally important.

The model is based on an algorithm to spatially interpolate point data, while correcting for effects of elevation. For each time period (eg. month), a regression is performed of precipitation against elevation. In some periods, the regression is not significant and is not used. Another key feature of the algorithm is that the values on the resultant map are guaranteed to match the observed values at each weather station.This algorithm was first developed by Michael Coughenour as part of a spatially explicit ecosystem model called SAVANNA (Coughenour 1992, 1993). The same algorithm was used in a Landscape Carrying Capacity Model for elk on Yellowstone’s northern elk winter range (Coughenour 1994, Coughenour and Singer 1996). The first application of the model to Yellowstone was at a research conference held in Yellowstone in 1991 (Coughenour 1994). In this application, GRASS GIS maps for elevation and vegetation were read into a model to calculate snow depth maps, available forage for elk, and elk carrying capacity on a biweekly basis throughout the winter. The model produced output files that were read into the GRASS GIS, to produce maps of snow depth and elk carrying capacity. These output maps were presented at the 1991 conference.

At about the same time, Phil Farnes was conducting studies of snow distributions on the Yellowstone northern elk winter range (Farnes and Romme 1993, Farnes 1994, 1996). He quantified the ways that slope, aspect, and tree cover affect snow pack, as compared to measurements made on a standard level, treeless sample site. He also developed ways to integrate data from numerous snow water sample sites into a unified data base, and ways to use snow water equivalent to calculate an index of winter severity that combines stress effects of cold temperature and heavy snow on elk.

The idea of combining the Coughenour model with the Farnes data into a stand-alone data model was the outcome of initial research on bison and elk carrying capacity by the two researchers in Grand Teton National Park (GTNP). The idea for that project was conceived by Robert Schiller and Francis Singer. Coughenour conducted preliminary SAVANNA modeling studies and Farnes collected snow data in GTNP. To create the stand-alone model, Coughenour combined his earlier elevation-based model with the slope/aspect/tree cover relationships of Farnes, in order to convert the snow data assembled by Farnes into maps of snow water equivalents in GTNP. The snow data model was delivered to GTNP by Coughenour and Farnes in 1999, at the same time Farnes delivered his unique data set (Farnes et al. 1999). Subsequently, a new phase of GTNP carrying capacity research was initiated by Tom Hobbs, F. Singer, G. Wockner, and L. Ziegenfuss.

In 2000, Gary Wockner, Tom Hobbs, and Francis Singer (CSU) obtained the model from Coughenour for this new phase of the GTNP project (Hobbs et al. 2001). Working with Farnes and Coughenour, Wockner obtained data to run it, worked through several software bugs, tested it, and then used it in a more complex carrying capacity model for the Jackson elk herd. A novel approach for handling the snow shadow effect east of the Teton Range was also developed. Results of this work can be seen at http://www.nrel.colostate.edu/projects/teton/. In addition, Linda Zeigenfuss used the model to predict snow water equivalents (SWE) for elk on the same days as overhead flights on the Jackson elk herd.

The model was applied and parameterized in the Fall of 2001 to Yellowstone National Park by Wockner, again using data provided by Farnes, and again for Linda Zeigenfuss for the same purpose in Yellowstone. This YNP snow modeling work was funded by Francis Singer. Coughenour had also been applying the core model to Yellowstone as part of his work on bison and elk carrying capacity with the SAVANNA model. Concurrent to the Wockner efforts, Coughenour applied the stand-alone data model to YNP, for use by other researchers on the YNP bison project (P. Gogan et. al.).

Other researchers working at YNP have expressed interest in obtaining the model. Therefore, this model is being made available on a website http://www.nrel.colostate.edu/projects/yellowstone/ through Colorado State University’s Natural Resource Ecology Lab. The model, which includes several files, is all zipped in a file named “Yellowstone_snow.zip”.

The model is being updated, and users are encouraged to submit their email contacts for notification of the most recent release.

Model Description

The model creates interpolated snow maps (SWE) of a one-day measurement for the entire Yellowstone study area. Several data sources drive the model:

digital elevation map – dem.asc
slope map – slope.asc
aspect map – aspect.asc
vegetation map – snow_veg.asc
snow site data – snowsites.dat, which contains the locations and elevations of the 28 snow measurement sites
snow measurement data -- yellsnow_data.dat, which contains the measured daily SWE dating from October 1, 1981 through September 30, 1999.

Using the data provided by Farnes measured at 28 YNP snow sites, this model creates a snapshot map of SWE on the study area. Specifically, using the DEM and the snow data, an initial grid is created using interpolation and regression. This grid is then readjusted for the effect of slope, aspect, and vegetation cover. Using slope and aspect, the more the cell tilts toward the sun, the more it is melted off; conversely, the more it is tilted away from the sun, the more snow accumulates. Using the vegetation data, less snow accumulates under conifers. The bigger the trees and the denser the stand, the less snow accumulation. The algorithmns expressing how these files are used are presented in Farnes et al, 2000.

All of these files are included in the zipped packet.

Instructions for Using the Model

Open the file “pptmap.dat”.

The model is in the folder “Snow Model”. The model reads an input file “pptmap.dat” which contains input parameters. It is best to open “.dat” files with ‘notepad’ rather than a word processing program. When the file is opened, it will contain the following parameter options:

1 /mapin GIS format 1-ARC ASCII

1 /mapout GIS format 1-ARC ASCII

1 /1-correct for slope-aspect and veg.cover with hard-coded routine sitefact.f, 0-do not

'dem.asc' /elevation map

'slope.asc' /slope map

'aspect.asc' /aspect map

'snow_veg.asc' /cover map

2 /base station file format 1-monthly, 2-daily

'snowsites.dat' /base station info file

'yellsnow_data.dat' /merged weather or snow data file

1 /method - 1-inverse distance weighting, 2-Farnes eqn, 3-equiv Farnes eqn

1 /power to use in inverse distance weighting 2-squared distance, 1-linear distance wt

1 /1-take average over a period, 2-take sum

0.20 /max fraction of missing values for a station to be included in regression on elev

3 /nregmin - minimum N for using a regression equation

0.1 /rsqmin - minimum R-square for using the regression equation

0. /slpmin - minimum slope for using the regression equation1

1997,1997 /first and last years (water) to use

0,0,0,0,0,0,0,0,0,0,1,0 /flagged months to use

15,15 /first and last days of month to use

0,0 /first and last julian dates to use

Change three input parameters.

The only options that need to be changed to make the model work on a specific day are the three lines -- “first and last years to use”, “flagged months to use”, and “first and last days of month to use”. Only one year, one month, and one day will be flagged at a time. In the above example, a map is created for November 15, 1997. Always use the same year and day before and after the comma in the options. Flag the month with a “1” as in the example above. As another example, if we were to run the model on January 6, 1996, the three lines would read:

1996,1996 /first and last years (water) to use

1,0,0,0,0,0,0,0,0,0,0,0 /flagged months to use

6,6 /first and last days of month to use

Ignore the option for julian dates.

Because the data in “yellsnow_data .dat” is organized by water year, the input in “pptmap.dat” must be in water year. The first day of the 1997 water year is October 1, 1996; the last day is September 30, 1997. To run the model on October 1, 1996, which is water-year 1997, the above file would read:

1997,1997 /first and last years (water) to use

0,0,0,0,0,0,0,0,0,1,0,0 /flagged months to use

1,1 /first and last days of month to use

Save the file

After changing theses three parameter lines, save the file.

Double-click on the application file “pptmap.exe”.

To run the model, simply ‘double-click’ on the application file “pptmap.exe”. The program will open a ‘DOS’ window and begin running through the rows of data. There are 1299 rows of data. On Wockner’s machine (750 Mhz, 400 MB RAM), it takes about 30 seconds to run. When it is done running, press “return” and the window will close.

Import the output into a GIS

The output map is the file “outmap.asc”. It is an Arc ASCII file with a header identical to the input files. We have been importing the file into Arcview, which converts it into an Arc grid. Then we change the legend and quickly scan the map to make sure the model has worked correctly. Since all input files have 100 meter cells, the outmap will also have 100 meter cells. The input/output is in UTM Zone 12, NAD 83, which is the same as used by YNP for the data that is downloadable from their website. Mike Coughenour recommends smoothing the outmap twice with a 5x5 filter before analyzing or using the results. If you have trouble importing the ASCII files into Arcview, make sure you don’t have spaces in the name given to the main folder in which the model is unzipped.

Measurable SWE is rarely present before November 1^st, and usually is gone by June 15^th at all sites. The site at Gardiner rarely registers any SWE. In the depth of winter, SWE predictions can be as high as 75 inches at the highest elevations in the park.

Several files are created along with the output map. They include:

Baseinfo.out = contains observed versus predicted values at the snow station locations.

Datesinc.out = contains the date of the output.

Debug.out = if there are problems they will appear here. Otherwise will say “doneread”.

Info.out = for this model, there will be warnings that base stations are outside the domain. These can be ignored.

Regress.out = contains the predicted values at base stations, the R², slope, and intercept of the regression equation.

Basesy.vec = the X/Y locations of the base stations.

Unless the output map looks incorrect, all of the above files can be ignored. We have run the model for various dates in the YNP system and the output always looks correct. Additionally, R² and slopes are always adequate. In the Teton system, we had trouble due to an anomalous snow shadow east of the Teton Range. In the Yellowstone system, no large anomalies exist, and the model runs smoothly with respect to its input.

Additional Information

Additional files (not used in the model) are in the folder “Images and Shapefiles”. The file Snowsites.jpg is an image of the entire study area including YNP boundaries, Northern Range boundaries, lakes, and snow station locations and names. We have also included shapefiles of these four GIS layers:

Yell_study.shp = Shapefile polygon of the Yellowstone study area

Yell_bound.shp = Shapefile polygon of YNP boundary

North.shp = Shapefile polygon of the Northern Range Boundary

Snowsites.shp = Shapefile points of the snow station locations

The file 1997_Ysnow.gif is an animated snow map of the 1997 water year. It begins in November 1996 and runs through June 1997. Images appear every week throughout the year. The snow-year 1997 was the heaviest on record in the 1981-1999 database. The files should play in a web browser.

Technical comments/questions on the model, contact:

Gary Wockner

Natural Resource Ecology Lab

Colorado State University

(970) 491-5724

gwockner@nrel.colostate.edu

Mike Coughenour

Natural Resource Ecology Lab

Colorado State University

(970) 491-5572

mikec@nrel.colostate.edu

Comments/questions about use, funding, applications, contact:

Francis Singer

Natural Resource Ecology Lab

Colorado State University

(970) 491-7056

francis@nrel.colostate.edu

Linda Zeigenfuss

USGS – BRD

(435) 789-0555

linnlarz@easilink.com

Literature Cited

Biggs, T. 1995. The development and use of a geographic information systems snow map for estimating equilibrium numbers of the gray wolf in Yellowstone National Park. B.S. Thesis, Princeton University, N.J. 30pp.

Brunnell, F. L., F. W. Hovey, R. S. McNay, and K. L. Parker. 1990. Forest cover, snow conditions, and black-tailed deer sinking depths. Canadian Journal of Zoology 68: 2403-2408

Cheville, N.F., D. R. McCullough, and L. R. Paulson. 1998. Brucelloses in the greater Yellowstone area. National Research Councel, Washington, D.C. 186pp.

Coughenour, M. B. 1992. Spatial modeling and landscape characterization of an African pastoral ecosystem: a prototype model and its potential use for monitoring drought. pp. 787-810 in: D.H. McKenzie , D.E. Hyatt and V.J. McDonald (eds.). Ecological Indicators, Vol. I. Elsevier Applied Science, London and New York.

Coughenour, M.B. 1994. Elk carrying capacity on Yellowstone’s northern elk winter range - Preliminary modeling to integrate climate, landscape, and elk nutritional requirements. pp. 97-112 in: Plants and Their Environments: Proceedings of the First Biennial Scientific Conference on the Greater Yellowstone Ecosystem, Mammoth Hot Springs, Sept. 1991. D. Despain (ed.). Technical Report NPS/NRYELL.NRTR-93/XX. USDI/NPS. Denver.

Coughenour, M.B. and F.J. Singer. 1996. Yellowstone elk population responses to fire - a comparison of landscape carrying capacity and spatial-dynamic ecosystem modeling approaches. pp. 169-180 in J. Greenlee (ed.), The Ecological Implications of Fire in Greater Yellowstone. International Association of Wildland Fire. Fairfield, WA.

Coughenour, M.B. and F. J. Singer. 1996b. Elk population processes in Yellowstone National Park under the policy of Natural Regulation. Ecol. Appl. 6:573-593.

Daly, C., R.P. Neilson, and D.L. Phillips. 1994. A statistical‑topographic model for mapping climatological precipitation over mountainous terrain. Journal of Applied Meteorology, 33, 140‑158.

Farnes, P., and W. Romme. 1993. Estimating localized snow water equivalents on the Yellowstone northern range. Pages 59-63 in Proceeds: 50^th Eastern Snow Conference, Quebec City, Quebec.

Farnes, P. 1996. An index of winter severity for elk. Pages 303-310, in F. Singer, ed. Effects of grazing by wild ungulates in Yellowstone National Park Technical Report. NPS-96-01, NPS, Denver, CO.

Farnes, P., C. Hayden, and K. Hansen. 1999. Snowpack distribution in Grand Teton National Park and Snake River Drainage above Jackson, Wyoming. Department of Earth Sciences, Montana State University.

Farnes, P., C. Heydon and K. Hansen. 2000. Snowpack in Yellowstone National Park. Department of Earth Sciences, Montana State University.

Geese, E. M., and S. Grothe. 1995. Analysis of coyote predation on deer and elk during winter in Yellowstone National Park, WY. American Midland Naturalist. 133: 36-43.

Hobbs, N.T. 1989. Linking energy balance to survival in mule deer: development and test of a simulation model. Wildlfie Monographs 101.

Hobbs, N. T., F. J. Singer and G. Wockner. 2001. Assessing Management Alternatives for Ungulates in the Greater Teton Ecosystem using Simulation Modeling, March 2001 Progress Report. Natural Resource Ecology lab, Colorado State University.

Hutchinson MF. 1989. A new objective method for spatial interpolation of meteorological variables from irregular networks applied to the estimation of monthly mean solar radiation, temperature, precipitation and windrun. CSIRO Division of Water Resources Tech. Memo. 89/5: 95‑104.

Hutchinson MF. 1995. Interpolating mean rainfall using thin plate smoothing splines. International Journal of GIS 9: 305‑403.

Huggard, D. J. 1993. Effect of snow depth on predation and scavenging by gray wolves. Journal of Wildlive Management. 57(2): 382-388.

Hungerford, R.D., R.R. Nemani, S.W. Running, J.C. Coughln. 1989. MTCLIM: A Mountain Microclimate Simulation Model. Res. Paper INT-414. Ogden, UT. U.S. Dept. of Agric. Forest Serc. Intermountain Res. Stn.

Merrill, E. H., N. L. Stanton, and J. C. Hak. 1994. Response of bluebunch wheatgrass, Idaho fescue and nematodes to ungulate grazing in Yellowstone National Park. Oikos 69:231-240.?

Running, S.W., R.R. Nemani, and R.D. Hungerford, 1987. Extrapolation of synoptic meteorological data in mountainous terrain and its use for simulating forest evaporation and photosynthesis. Canadian Journal of Forest Research, 17:472‑483.