Importance and Availability of Climate Information

Forest pest risk maps are typically assembled by combining spatial data from three principal subject areas: host species distribution, pathways of pest movement, and key environmental factors (Bartell and Nair 2004). Climatic attributes such as temperature and moisture strongly shape pest behavior, affecting survival, reproductive rate, and in many cases, the ability to spread at a continental scale. Thus, climatic data provide an important coarse filter for forest pest risk analyses. Regularly gridded climate maps covering the entire geographic area of interest are typically required for analytical purposes. Such maps may be constructed by spatial interpolation of weather station data. These data are readily available for much of the United States, dating back several decades, from the National Oceanic and Atmospheric Administration (NOAA) National Climatic Data Center (NCDC).

Spatial Interpolation of Climatic Variables

A wide array of spatial interpolation algorithms, (e.g., geostatistical, regression, spline, inverse distance weighting) have been used to construct broad spatial-scale climatic data sets from weather station data (Daly 2006, Mardikis and others 2005, Nalder and Wein 1998, Price and others 2000, Xia and others 2000). Most currently accepted methods acknowledge that terrain is a significant factor governing climate at all but the broadest scales, and they use elevation measurements to represent terrain and adjust climatic variable values accordingly (Daly 2006). One well received interpolation approach is the Parameter-elevation Regressions on Independent Slopes Model (PRISM). Initially developed to generate precipitation maps for the Pacific Northwest (Daly and others 1994), the approach has since been applied to create maps of temperature, relative humidity, snowfall, growing-degree days, and many other variables (Daly and others 2000). In particular, the PRISM approach was applied to generate most of the maps in the recent version of the Climate Atlas of the United States (Plantico and others 2002), as well as similar products for Canada and China (Daly and others 2000). The PRISM approach is a knowledge-based system integrating a local climate-elevation regression with other algorithmic components: station weighting, topographic facets, coastal proximity, and a two-layer atmosphere (Daly and others 2002). When initially tested on precipitation in the Pacific Northwest, the PRISM approach outperformed other interpolation methods in comparative analyses (Daly and others 1994).

Limitations of Existing Interpolated Climatic Data Sets

There are several limitations of PRISM-derived or similar data sets with respect to their use for forest pest risk maps. First, most national-scale climatic data sets are calculated as normals, meaning an average of the variable of interest across a window of time, typically a 30-year period. For example, most data sets in the recent version of the Climate Atlas of the United States are based on inputs from 1961-1990 (Plantico and others 2002). Current weather data are not incorporated into the maps, so any pest risk map constructed from them will not include current events—and the accompanying variability—that may be relevant to an assessment of immediate risk.

Second, there are related issues of cost and data format. The Climate Atlas contains polygonal maps for a large number of potentially relevant climatic normals but does not include the regularly gridded data from which the maps are derived. These polygonal maps have limited attribute resolution, with the range of the original gridded data typically compressed into nine or fewer classes. Monthly gridded maps of a few variables—precipitation amount, mean minimum temperature, mean maximum temperature, and mean dewpoint—are available for public download from the PRISM group at Oregon State University (http://www.ocs.orst.edu/prism/). Notably, these maps are fairly current (finalized maps are available from 1997 through mid-2006), and the database is regularly updated, but it does not include many climatic variables that might be of interest for forest pest risk assessment, (e.g., relative humidity, number of days above freezing, or number of days with measurable precipitation). Regularly gridded data of these and other (30-year normal) variables, derived using the PRISM method, are available, but at substantial cost (from the Climate Source: http://www.climatesource.com/).

Third, most available climatic spatial data sets, whether derived using PRISM or other methods, are monthly or annual summaries depicting mean or extreme values over the time period. For some forest pests, the short-term, even daily status of multiple weather conditions may be relevant to the pest’s growth, persistence, or invasiveness. Fungal pathogens are particularly affected by the interaction of temperature and moisture availability. For example, the pathogen that causes late blight of potato (Phytophthora infestans) develops best at cool temperatures during extended periods of wet weather, as do many other Phytophthora species (Davidson and others 2002, Harvell and others 2002, Marshall-Farrar and others 1998). The interaction of climatic variables can also be important for some insect pests (Harrington and others 2001, Peacock and others 2006). Nevertheless, although there has been some effort to create maps of daily precipitation and temperature at a broad scale (Hunter and Meentemeyer 2005), there has been little attention paid to the co-occurrence of multiple weather conditions favorable to pest persistence and spread. Daily weather data allow the counting of how often, and for how long, variables meet certain threshold values. Creation of broad-scale maps from data derived in this manner may require a different spatial interpolation approach than that used for continuously distributed variables (van de Kassteele and others 2005).