Will Patterson

Integrating ArcView and the Spatial Analyst Extension with the PRISM Climate Expert System

Geographic Information Systems (GIS) are increasingly being associated with climatological research as flexible utilities in which to develop and explore modeled data. PRISM (Parameter-elevation Regressions on Independent Slopes Model) is a climate-mapping expert system developed by Dr. Christopher Daly and staff from the Oregon Climate Service (OCS) at Oregon State University (OSU). PRISM utilizes a digital elevation model (DEM) and a weighted regression technique to distribute point climate measurements to a geographic grid. PRISM exhibited superior results over various kriging methods in a comparative study of regional precipitation mapping. While the PRISM software is not available to the public, many products and datasets produced by PRISM are available over the Internet. This paper will provide a brief overview of climate mapping with PRISM and describe how ArcView and the Spatial Analyst extension have been customized with Avenue to assist with formatting PRISM input files and assessing PRISM output data. The availability of GIS-compatible PRISM datasets will also be discussed.


Introduction

GIS has not only enhanced data visualization within disciplines, but perhaps more importantly, greatly improved data relations across discipline boundaries. For example, with the power of GIS overlays, a soil scientist can now directly compare categorized soil-type polygons with precipitation spatial data developed by a meteorologist. This basic yet essential functionality has helped fuel the rapid demand for GIS data. In particular, spatial distributions of climate parameters are proving invaluable as base data for overlay analysis and as inputs for other agricultural, natural resource, and hydrological models.

Recent advances in the provision of spatial climate data have been accomplished through development of the PRISM expert system. PRISM is a group of seven stand-alone programs that work together to produce gridded estimates of climate parameters. PRISM has been termed an expert system because it attempts to duplicate the procedures an expert would use to develop climate maps. The system is based on the philosophy that topographic orientation and elevation are primary driving factors of climate distributions. PRISM was developed originally to address the problems of precipitation mapping in complex terrain. However, as the system evolved, it was discovered that other climatic properties such as snow and temperature could also be successfully mapped using similar modeling assumptions.

The author of this paper has been using PRISM in a precipitation modeling project for the National Weather Service Office in Eureka, California. ArcView has been used as an assistive tool for formatting necessary files to input into PRISM and also for viewing PRISM output data. Both PRISM input and output data are ASCII files with specific formats. It was discovered that Avenue, ArcView's object-oriented programming language, provides several features for reading and writing ASCII data. By bundling together related Avenue routines for developing and importing PRISM-format files, a PRISM Tools extension for ArcView was developed to assist users of the expert system.


The PRISM Method

PRISM uses a DEM to estimate the elevation of each input weather station. Next, each DEM cell is assessed for slope orientation and assigned to a topographic "facet" (a group of cells with a common topographic orientation). Localized climate-parameter/elevation regression functions are then formed from nearby weather stations and a parameter estimate is produced. In the regression equations, stronger weights are assigned to weather stations with locations, elevations, and topographic positionings similar to that of the DEM cell. Finally, PRISM produces estimate prediction intervals to approximate the levels of uncertainty involved.


PRISM Versus Kriging

Kriging has become a popular method for interpolating many types of data, including precipitation and other climate themes. Kriging utilizes a semivariogram to determine the optimal station weights for interpolation purposes. PRISM was compared to kriging, detrended kriging, and cokriging in a regional precipitation mapping exercise for the Willamette River Basin, Oregon (Daly, et al., 1994). Of these methods, PRISM exhibited the lowest overall bias and mean absolute error. The latter two methods also require area-wide relationships between the climate parameter and elevation (i.e. between precipitation and elevation). This relationship does not often exist when modeling large domains such as states or nations. PRISM avoids this problem by continually adjusting its cell neighborhood to adapt to changes in orography at multiple scales.


Acceptance of PRISM

The PRISM Evaluation Group (PEG) was formed to critique and review PRISM methods and products. PEG consists of State and Regional Climatologists, representatives of national agencies such as the Natural Resources Conservation Service (NRCS), and other potential users from state and local governments. After considerable evaluation and scrutiny, PEG determined that PRISM-generated precipitation maps were of equal or greater quality than the best manually prepared maps available. Under an agreement with NRCS, PRISM is currently being used to produce climate maps for all fifty states.


The Seven PRISM Programs

The PRISM expert system consists of the following seven programs (Daly, et. al., 1996). The programs can be used separately for specific tasks but are more commonly used sequentially in a complete modeling session.

1. LATTICE utilizes a Gaussian filter to smooth and/or convert grids to different spatial resolutions. It is typically used to convert a DEM to a resolution that is appropriate for the climate parameter being modeled. LATTICE will also convert GRASS or Esri-format ASCII Grids to PRISM-format ASCII grids used within the expert system.

2. FACET uses the output DEM from LATTICE to develop grids of slope orientation at up to six separate spatial scales. Adjacent cells with similar slope orientations are grouped together to form contiguous topographic zones called facets. Nine different orientations are possible, including eight compass directions and a flat orientation.

3. ASSAY uses an iterative jacknife cross-validation technique to determine optimal modeling settings.

4. PRISM is the core interpolation module that produces gridded estimates of climate parameters based on input station data, FACET grids, and a DEM. Gridded predictions are produced from a weighted parameter-elevation regression function calculated for each DEM cell.

5. POLISH is a postprocessing routine for evaluating parameter-elevation regression slopes produced by PRISM. Some smoothing of an output grid may be performed in cases where POLISH detects abrupt transitions between adjacent topographic facets with opposing climate characteristics.

6. MIRROR produces point value estimates from POLISH output grids and various performance summary and diagnostic statistics.

7. GISLINK performs unit conversions if necessary, and also converts native PRISM-format output grids to one of two GIS compatible formats, GRASS or Esri ASCII Grid (also known as ArcInfo ASCII Grid).


PRISM Flowchart

The following flowchart illustrates a typical PRISM modeling session.

Click on the image to see the full size version.

PRISM Flowchart

(Image courtesy of Wayne Gibson, Oregon State University)


The Challenge

The LATTICE and GISLINK programs within the expert system provide the necessary transitions between internal PRISM-specific file formats and GIS-compatible data. However, it was hypothesized that some customized utilities within a GIS environment could supplement these programs with additional routines for creating or reformatting PRISM data. The ability to view various intermediate PRISM files produced during a modeling session without having to put each one through several GIS conversion and importing steps could prove to be helpful.

Some examples of PRISM input and output data include:

Input Data:

Output Data: ArcView was chosen to try to fulfill the challenge of providing a unified platform for managing PRISM input files and evaluating PRISM output data. When used with the optional Spatial Analyst extension, ArcView provides a robust set of raster tools that includes the capability of importing ASCII Grids in an Esri-format. However, it wasn't clear whether ArcView would provide adequate customization features for the task at hand.


Avenue Provides a Solution

It was soon discovered that Avenue provides the exact utilities necessary to develop shortcuts for dealing with PRISM data. Avenue is a programming language that utilizes objects and requests to complete user-desired tasks. An object is basically something that you are working with, and requests are used to initiate a particular function or property of the object. For example, a linefile is an object that allows Avenue to read through ASCII files line by line. Linefile objects are controlled by specific requests that will read, store, and write new ASCII files. By using these objects and requests, conversion between different ASCII file formats is made possible.

The following is an example algorithm to import a PRISM-format ASCII grid directly into ArcView, thereby creating a new Esri Grid:

1. Convert the input PRISM-format ASCII grid to a linefile object.
2. Read the header of the PRISM grid, calculate cell size information,
   and store the coordinate data.
3. Create a temporary linefile object to act as an Esri-format ASCII grid.
4. Using the stored coordinate and cell size data, write out an Esri-format
   ASCII grid header into the temporary linefile.
5. Read the cell values from the PRISM grid linefile, line by line.
6. Write the cell values to the temporary linefile, line by line.
7. Import the temporary linefile as an Esri-format ASCII grid directly into
   a view; this will create a regular Esri Grid.
8. Delete the temporary linefile.

Essentially, this algorithm converts a PRISM-format ASCII grid with this header:
GEO
north:      42.686631944580
south:      40.979166666666
east:     -130.230034722646
west:     -125.02083333334
rows:    145
cols:    217

To a temporary Esri-format ASCII grid with this header:
ncols 217
nrows 145
xllcorner -125.02083333334
yllcorner  40.979166666666
cellsize  0.04166666667

And then imports the temporary Esri-format ASCII grid directly into ArcView, thereby creating a regular Esri Grid. Before finishing, the routine also deletes the temporary Esri-format ASCII Grid.

This entire procedure can be represented by one Avenue program, or script. When executed, the grid conversion and importing process is instantaneous and transparent to the user. By using similar approaches, a variety of assistive conversion routines were developed for use with PRISM, ranging from importing weather station files as shapefiles to exporting Esri Grids directly to PRISM-format ASCII grids.


The PRISM Tools Extension

To simplify the installation and accessibility of the PRISM shortcut scripts, an ArcView "PRISM Tools" extension was developed. An extension is essentially a group of Avenue scripts and customized interface features that can be turned on or off at the discretion of the user. Eleven separate scripts for working with PRISM data as tables (databases), Grids, and shapefiles were included in the extension. Because PRISM Tools also uses the Spatial Analyst extension, many other powerful features are available. For example, several handy climate-style color ramps are available for viewing Grids, including some specifically for use with precipitation and temperature.

The following images illustrate the look and feel of the PRISM Tools extension, and include some background examples of various diagnostic and final products produced by the expert system.

Click on the images to see the full size versions.

ArcView Project-Extensions Menu with PRISM Tools extension available:

ArcView Project-Extensions Menu

ArcView Table-PRISM Menu with FACET grid and weather station shapefile in background:

ArcView Table-PRISM Menu

ArcView View-PRISM Menu with final precipitation grid in background:

ArcView View-PRISM Menu


Conclusions

The PRISM expert system has provided unique and powerful techniques to produce climate GIS data. In a similar way, ArcView, Avenue, and the Spatial Analyst extension are providing outstanding environments for viewing and evaluating climate data. The lessons learned from the integration of these technologies not only apply to climate research, but also to any undertaking where ASCII format data needs to be visualized.


Availability of PRISM Data

A wide variety of final polished PRISM products and metadata are in the public domain and are available for download from the Oregon Climate Service web site. The climate grids are available in Esri ASCII Grid formats, and can be used directly with the ArcView Spatial Analyst extension or the ArcInfo GRID module. Various polygon interpolations of the climate grids are also available in ArcInfo coverage format.

Oregon Climate Service web site:
http://www.ocs.orst.edu

Specific URL for PRISM Products:
http://www.ocs.orst.edu/prism/prism_products.html


Acknowledgments

Dr. Christopher Daly of Oregon State University and staff from the Oregon Climate Service for assistance with this project and allowing me to test the PRISM expert system.

Mr. Mel K. Nordquist for meteorological advice, data classification, and academic consultation.

Dr. Lawrence Fox III and Dr. Steven A. Carlson, both of Humboldt State University, for research and advisory guidance.

The California Department of Fish and Game, Technical Services Branch, GIS Unit for GIS assistance and programming advice.


References and Additional Readings

Daly, C., Neilson, R.P. and Phillips, D.L., 1994: A statistical-topographic model for mapping climatological precipitation over mountainous terrain. Journal of Applied Meteorology 33, 140-58.

Daly, C., et. al. 1996: Overview of the PRISM model - an Internet publication.

Daly, C., G.H. Taylor, and W.P. Gibson., 1997: The PRISM approach to mapping precipitation and temperature. In: Proceedings of the 10th AMS Conference in Applied Climatology, American Meteorological Society, Reno, NV, Oct. 20-23, 10-12.

Esri Educational Services, 1997: Programming with Avenue, three-day course notebook with exercises and training data.

McDonnell, R.A., 1996: Including the spatial dimension: using geographical information systems in hydrology. Progress in Physical Geography 20, 159-177.


Author Information

ArcView PRISM Tools Extension Developer and Author of this Paper:

Will Patterson, GIS Analyst
California Department of Fish and Game
Technical Services Branch
1730 I Street, Suite 100
Sacramento, California 95814
Telephone: (916) 323-1484
Fax: (916) 323-1431
E-mail: wpatters@hq.dfg.ca.gov

Also representing:

Spatial Analysis Laboratory
Department of Natural Resources, Planning, and Interpretation
Natural Resources Graduate Program
Humboldt State University
Arcata, California 95521


Other Contacts

PRISM Author and Primary Contact:

Dr. Christopher Daly
Oregon State University
USFS Forestry Sciences Lab
3200 SW Jefferson Way
Corvallis, OR 97331
E-mail: daly@fsl.orst.edu

PRISM GIS Products Contact:

Wayne Gibson
Oregon Climate Service
326 Strand Ag. Hall
Corvallis, OR 97331-2209
E-mail: oregon@ats.orst.edu