A Spatial Resampling Perspective on the Depiction of Global Air Temperature Anomalies

Abstract

Networks of near-surface climate stations often produce samples of air temperature that are spatially uneven and sparse. Spatial samples of air temperature, as a result, affect both the regional depiction of air temperature anomalies and large-scale spatial averages. A resampling procedure is used here to illustrate the general problem of how spatially variable station networks can influence estimates of climate change. As an example, the warm air temperature anomalies that occurred during 1988 are resampled using station networks from other years. During 1988, areas in North America and northern Asia experienced very warm conditions, producing a terrestrially averaged air temperature anomaly of over 0.4°C. If the 1988 air temperature anomalies were resampled using the 1981 station network, the warm event in northern Asia would not have been detected due to missing air temperatures over northern Russia during 1981. As a result, the 1981 station network would have estimated the 1988 terrestrial air temperature anomaly to be 0.25°C. Alternatively, the station network of 1888 was biased toward areas that had warm 1988 air temperatures and would estimate the 1988 air temperature anomaly to be 0.52°C. In addition to problems in estimating the terrestrially averaged air temperature anomaly, regional climatic variability is altered by resampling the 1988 anomalies with other networks. The sampling problems illustrated here also show the importance of having free and open exchanges of data worldwide. National networks of climatic stations are crucially important not only in the analysis of regional climatic variability, but also in the detection of global-scale climatic change.