Influence of spatial information resolution on the relation between elevation and temperature
Abstract
The association between elevation and temperature is analysed by simple linear
correlations across several spatial scales. The minimum (tn) and maximum
(tx) temperatures (response variables), expressed at two time scales (monthly and
daily), are observed for 102 weather stations in east central France from 1980 to
2014 (12,784 days). Elevation (explanatory variable) is provided at 10 resolutions:
50, 100, 200, 500 m, 1, 2, 4, 8, 12, and 16 km. The coefficient of determination,
R2, is used to determine which resolution gives the best results. The slope given by
the regression is used to assess the drop in temperature per unit of elevation (temperature
lapse rate [TLR]). In most situations, monthly and daily temperatures are
optimally explained by the finest (50 m) resolution: the R2 is, respectively, 0.53
and 0.24 for tn and 0.78 and 0.39 for tx. The coarser resolutions produce results of
much lower quality. However, in one circumstance (monthly mean of tn), the highest
R2 value is obtained for the 4-km resolution, which is a meaningful result as
current regional climate models now achieve similar resolutions. Both monthly and
daily TLRs of tn and tx are, on average, slightly lower than −0.5 C/100 m at
50-m resolution. The TLR decreases with resolution: it is only −0.23 C/100 m for
tn and −0.13 C/100 m for tx at 16-km resolution. Other insightful results involve
the influence of the topographical context, which shows some additional effect
with that of elevation and which was quantified through partial correlations.