Data regression on sphere for luminance map creation from sky scanner measurements

• Sky scanner measurements are used for computation of sky luminance distribution.
• Three different regression methods are tested and compared.
• Hybrid adaptive splines seem to be the best solution for luminance map creation.
• Diffuse horizontal illuminances are calculated and compared for each regression.

Sky luminance distribution acquired by a whole-sky scanner has well-known constraints in resolution because of the restricted field of view and limited number of measured sky elements. The interpolating methods on measured data can significantly reduce the uncertainties of the scanners to create more accurate sky luminance maps depending on the mathematical approach. The general lack of many regressions is their low adaptability to significant sky luminance changes, e.g. on the cloud edges or near the sun disk. Three regression methods like inverse distance to a power, smoothing splines on the sphere and hybrid adaptive splines were tested on real skies measured with a self-made portable spectral sky scanner. Data regressions are compared and verified by the luxmeter in terms of calculated diffuse horizontal illuminance.