Premium
Comparison of regression techniques for mapping fog frequency: application to the Aragón region (northeast Spain)
Author(s) -
VicenteSerrano S. M.,
LópezMoreno J. I.,
VegaRodríguez M. I.,
Beguería S.,
Cuadrat José M.
Publication year - 2010
Publication title -
international journal of climatology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.58
H-Index - 166
eISSN - 1097-0088
pISSN - 0899-8418
DOI - 10.1002/joc.1935
Subject(s) - fog , terrain , precipitation , multivariate interpolation , variable (mathematics) , climatology , range (aeronautics) , environmental science , meteorology , linear regression , linear model , generalized additive model , geography , physical geography , cartography , statistics , mathematics , geology , mathematical analysis , materials science , composite material , bilinear interpolation
Abstract We compare different spatial interpolation techniques in mapping the monthly frequency of fogs in the Aragón region (northeast Spain). The local and spatially complex nature of the fogs makes them more difficult to map than other climatic variables such as precipitation and temperature. We found clear seasonal differences in the quality of the obtained maps. The localized nature of spring and summer fogs, mainly restricted to valley bottoms in mountainous areas, gives rise to several limitations. The modelling of fog frequency is more complex than that of other climate variables; to improve the model predictions, it is necessary to consider topographic variables that simulate the terrain structure. Moreover, the highly complex nature of the relationship between fog frequency and terrain means that simple linear models perform poorly in accounting for the role of geographic and topographic variables in determining the spatial distribution of fog frequency. The inclusion of non‐linear relationships between fog frequency and terrain variables in the models following a general additive model (GAM) procedure leads to an improvement in model performance because the flexibility of GAMs enables the inclusion of non‐linear relationships and the generation of response‐curve shapes that detail the exact relationship between the dependent variable and predictors throughout the entire range of the variable. Copyright © 2009 Royal Meteorological Society