Who Wins a Battle Between Efficiency and Accuracy? A Case of Spatial Interpolation for Environmental Applications

• Activity Number: 516
• Type: Contributed
• Date/Time: Wednesday, August 6, 2014 : 10:30 AM to 11:15 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Ilya Rozonoyer*+ and Elena Naumova and Alexander Liss
• Companies: Tufts University and Tufts University and Tufts University
• Keywords: interpolation ; kriging ; spatial ; climate ; time series ; temperature

Abstract

Environmental models often include directly or indirectly weather parameters with a high spatial and temporal resolution. Economic or operational constraints usually limit availability of such data. We aimed to compare deterministic Inverse Distance Weighting (IDW) and probabilistic Ordinary Kriging (KO) spatial interpolations of weather parameters with application to environmental research. We abstracted daily temperature, precipitation, and snow cover measurements from weather gauges in two distinctly different localities: Alaska and the Continental USA. We interpolated the data feed to 0.1 degree spatial grid with IDW and KO, and compared accuracy and precision of both methods using cross-validation. We found that for homogenous terrain both methods were comparable. For heterogeneous terrain IDW offered slightly better accuracy and stability than KO. We demonstrated that for modern applications IDW method can provide an attractive alternative to computationally expensive KO without substantial loss of accuracy or precision. These results promise substantial savings in computational costs and faster, more affordable solutions. Extension to time-space autoregressive is suggested.