High-resolution monthly precipitation climatologies over Norway: assessment of spatial interpolation methods

2018 
Monthly precipitation climatologies at 1 km resolution have been produced over the Norwegian mainland for 1981-2010. The observed station normals are interpolated over a regular grid by applying a multi-linear local regression kriging (MLRK). The statistical method aims at modeling the influence of the main geographical features, such as: latitude, longitude, elevation and sea nearness on the precipitation field at a local scale. The MLRK is composed of two steps, (i) a background precipitation field is computed through a multi-linear local regression based on the geographical information, then (ii) a kriging interpolation is applied to adjust the field so to better fit the station residuals (i.e., the difference between the observed normals and the background field). The interpolation accuracy is evaluated by reconstructing the station normals with a leave-one-out approach and by comparing the model performances with those of other interpolation methods. The poor observation coverage over remote and mountainous regions in Norway has motivated us to consider precipitation fields produced by numerical models. In fact, numerical model output provides a reference field for the evaluation of MLRK that is not dependent on the station density, though it is not as accurate as the observations. Specifically, a regional climate simulation with a resolution of 2.5 km, based on the dynamical downscaling of the global reanalysis ERAInterim and available for the time period 2003-2016, has been used. In addition to the MLRK validation, the combination of numerical model fields and observed climatologies has been investigated and the results show that this integrated approach provides more accurate high-resolution climatologies over the Country. The numerical background introduces valuable information over remote and mountainous regions, whereas in-situ data correct for model biases.
    • Correction
    • Source
    • Cite
    • Save
    • Machine Reading By IdeaReader
    9
    References
    2
    Citations
    NaN
    KQI
    []