Get access

Validation and inter-comparison of three methodologies for interpolating daily precipitation and temperature across Canada

Authors

  • Nathaniel K. Newlands,

    Corresponding author
    1. Environmental Health Research Branch, Agriculture and Agri-Food Canada, 5403 - 1 Avenue South, Lethbridge, AB T1J 4B1, Canada
    • Environment, Research Branch, Agriculture and Agri-Food Canada, 5403 - 1 Avenue South, P.O. Box 3000, Lethbridge, Alberta T1J 4B1, Canada.
    Search for more papers by this author
  • Andrew Davidson,

    1. National Land and Water Information Service, 960 Carling Avenue, Agriculture and Agri-Food Canada, Ottawa, ON K1A OC6, Canada
    Search for more papers by this author
  • Allan Howard,

    1. National Agro-climate Information Service, Agriculture and Agri-Food Canada, 1800 Hamilton St., Regina, SK S4P 4L2, Canada
    Search for more papers by this author
  • Harvey Hill

    1. Climate Impacts and Adaptation, Agriculture and Agri-Food Canada, 1011, Innovation Blvd, Saskatoon, SK S7V 1B7, Canada
    Search for more papers by this author

Abstract

The use of daily climate data in agriculture has increased considerably over the past two decades due to the rapid development of information technology and the need to better assess impacts and risks from extreme weather and accelerating climate change. While daily station data is now regularly used as an input to biophysical and biogeochemical models for the study of climate, agriculture, and forestry, questions still remain on the level of uncertainty in using daily data, especially for predictions made by spatial interpolation models. We evaluate the precision of three models (i.e., spline, weighted-truncated Gaussian filter, and hybrid inverse-distance/natural-neighbor) for interpolating daily precipitation and temperature at 10 km across the Canadian landmass south of 60o latitude (encompassing Canada's agricultural region). We compute daily, weekly, and monthly-aggregated bias and root-mean-square (RMSE) validation statistics, examining how error varies with orography and topography, and proximity to large water. Our findings show the best approach for interpolating daily temperature and precipitation across Canada requires a mixed-model/Bayesian approach. Further application of interpolation methods that consider non-stationary spatial covariance, alongside measurement of spatial correlation range would aid considerably in reducing interpolation prediction uncertainty. Copyright © 2010 Crown in Right of Canada.

Ancillary