Modeling monthly temperature in mountainous ecoregions: Importance of spatial scale for ecological research

Christine R. Rollinson, Margot W. Kaye

Research output: Contribution to journalArticle

2 Citations (Scopus)

Abstract

Large gridded climate datasets facilitate correlation with landscape patterns (e.g. with species distribution), but point-based ecological data are often disconnected from the spatial extent and resolution of available climate data. Consequently, data users must understand the tradeoffs between spatial and temporal resolution, as well as potential biases arising from the modeling approach when selecting data. The Global Regression Ecoregional Analysis of Temperature (GREAT) model was developed to describe monthly temperature experienced by locations along elevation and latitudinal gradients in the central Appalachians. Model performance was assessed at 30 m and 1 km spatial resolutions, with subsequent analysis of the tradeoffs between extracting temperatures from an a priori spatial grid versus modeling temperature for specific locations. Results indicate that temperatures modeled at coarser resolution (1 km) had higher correlation with observed station temperatures because models developed at finer spatial scales (30 m) over-emphasize the influence of topographic variation. Monthly temperatures from the GREAT model were compared with temperatures extracted from the established, spatially weighted gridded products PRISM and Daymet. This comparison showed greater correlation between the spatially weighted modeling approaches than with the global regression-based method. However, monthly temperatures from the GREAT model had the highest correlation with temperatures observed at 14 independent climate stations that were not used in the development of any of the 3 models. The GREAT model allows temperature to be modeled specifically for each station and results in higher model performance than can be achieved with gridded datasets.

Original languageEnglish (US)
Pages (from-to)99-110
Number of pages12
JournalClimate Research
Volume64
Issue number2
DOIs
StatePublished - Jan 1 2015

Fingerprint

ecoregion
modeling
temperature
Temperature
Regression analysis
regression analysis
climate
latitudinal gradient
spatial resolution

All Science Journal Classification (ASJC) codes

  • Environmental Chemistry
  • Environmental Science(all)
  • Atmospheric Science

Cite this

@article{1db1e1d79a8e47f4a62f559fde5b6f64,
title = "Modeling monthly temperature in mountainous ecoregions: Importance of spatial scale for ecological research",
abstract = "Large gridded climate datasets facilitate correlation with landscape patterns (e.g. with species distribution), but point-based ecological data are often disconnected from the spatial extent and resolution of available climate data. Consequently, data users must understand the tradeoffs between spatial and temporal resolution, as well as potential biases arising from the modeling approach when selecting data. The Global Regression Ecoregional Analysis of Temperature (GREAT) model was developed to describe monthly temperature experienced by locations along elevation and latitudinal gradients in the central Appalachians. Model performance was assessed at 30 m and 1 km spatial resolutions, with subsequent analysis of the tradeoffs between extracting temperatures from an a priori spatial grid versus modeling temperature for specific locations. Results indicate that temperatures modeled at coarser resolution (1 km) had higher correlation with observed station temperatures because models developed at finer spatial scales (30 m) over-emphasize the influence of topographic variation. Monthly temperatures from the GREAT model were compared with temperatures extracted from the established, spatially weighted gridded products PRISM and Daymet. This comparison showed greater correlation between the spatially weighted modeling approaches than with the global regression-based method. However, monthly temperatures from the GREAT model had the highest correlation with temperatures observed at 14 independent climate stations that were not used in the development of any of the 3 models. The GREAT model allows temperature to be modeled specifically for each station and results in higher model performance than can be achieved with gridded datasets.",
author = "Rollinson, {Christine R.} and Kaye, {Margot W.}",
year = "2015",
month = "1",
day = "1",
doi = "10.3354/cr01306",
language = "English (US)",
volume = "64",
pages = "99--110",
journal = "Climate Research",
issn = "0936-577X",
publisher = "Inter-Research",
number = "2",

}

Modeling monthly temperature in mountainous ecoregions : Importance of spatial scale for ecological research. / Rollinson, Christine R.; Kaye, Margot W.

In: Climate Research, Vol. 64, No. 2, 01.01.2015, p. 99-110.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Modeling monthly temperature in mountainous ecoregions

T2 - Importance of spatial scale for ecological research

AU - Rollinson, Christine R.

AU - Kaye, Margot W.

PY - 2015/1/1

Y1 - 2015/1/1

N2 - Large gridded climate datasets facilitate correlation with landscape patterns (e.g. with species distribution), but point-based ecological data are often disconnected from the spatial extent and resolution of available climate data. Consequently, data users must understand the tradeoffs between spatial and temporal resolution, as well as potential biases arising from the modeling approach when selecting data. The Global Regression Ecoregional Analysis of Temperature (GREAT) model was developed to describe monthly temperature experienced by locations along elevation and latitudinal gradients in the central Appalachians. Model performance was assessed at 30 m and 1 km spatial resolutions, with subsequent analysis of the tradeoffs between extracting temperatures from an a priori spatial grid versus modeling temperature for specific locations. Results indicate that temperatures modeled at coarser resolution (1 km) had higher correlation with observed station temperatures because models developed at finer spatial scales (30 m) over-emphasize the influence of topographic variation. Monthly temperatures from the GREAT model were compared with temperatures extracted from the established, spatially weighted gridded products PRISM and Daymet. This comparison showed greater correlation between the spatially weighted modeling approaches than with the global regression-based method. However, monthly temperatures from the GREAT model had the highest correlation with temperatures observed at 14 independent climate stations that were not used in the development of any of the 3 models. The GREAT model allows temperature to be modeled specifically for each station and results in higher model performance than can be achieved with gridded datasets.

AB - Large gridded climate datasets facilitate correlation with landscape patterns (e.g. with species distribution), but point-based ecological data are often disconnected from the spatial extent and resolution of available climate data. Consequently, data users must understand the tradeoffs between spatial and temporal resolution, as well as potential biases arising from the modeling approach when selecting data. The Global Regression Ecoregional Analysis of Temperature (GREAT) model was developed to describe monthly temperature experienced by locations along elevation and latitudinal gradients in the central Appalachians. Model performance was assessed at 30 m and 1 km spatial resolutions, with subsequent analysis of the tradeoffs between extracting temperatures from an a priori spatial grid versus modeling temperature for specific locations. Results indicate that temperatures modeled at coarser resolution (1 km) had higher correlation with observed station temperatures because models developed at finer spatial scales (30 m) over-emphasize the influence of topographic variation. Monthly temperatures from the GREAT model were compared with temperatures extracted from the established, spatially weighted gridded products PRISM and Daymet. This comparison showed greater correlation between the spatially weighted modeling approaches than with the global regression-based method. However, monthly temperatures from the GREAT model had the highest correlation with temperatures observed at 14 independent climate stations that were not used in the development of any of the 3 models. The GREAT model allows temperature to be modeled specifically for each station and results in higher model performance than can be achieved with gridded datasets.

UR - http://www.scopus.com/inward/record.url?scp=84937034059&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84937034059&partnerID=8YFLogxK

U2 - 10.3354/cr01306

DO - 10.3354/cr01306

M3 - Article

AN - SCOPUS:84937034059

VL - 64

SP - 99

EP - 110

JO - Climate Research

JF - Climate Research

SN - 0936-577X

IS - 2

ER -