Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
8866531 | Remote Sensing of Environment | 2018 | 11 Pages |
Abstract
We developed a methodology to evaluate quantitative uncertainty characteristics of satellite data retrievals including the contribution of systematic error and statistical error. This is introduced by assessing the total product error of optical, satellite data-based, Fractional Snow Cover (FSC) estimates. Here the FSC estimation is based on an algorithm allowing the consideration of the effect of different error sources; a semi-empirical reflectance model describing the relationship of the observed reflectance and FSC through several variables and parameters. We assume that after the statistical error analysis, the remaining portion of total product error arises due to systematic factors. Hence, first we define a statistical error component through the theory of error propagation, and then estimate the total product error (PE) by using in situ observations on FSC, and finally derive the systematic error from these two error components. The experimental approach for estimating PE is conducted through an analysis of the observed estimation errors (i.e. residuals) in the GlobSnow Snow Extent (SE) v2.1 products on FSC. In practice, independent in situ snow course observations from Finland on FSC are compared to corresponding satellite FSC estimates to quantify the residuals. The approach is then illustrated for an extended region of corresponding European boreal forest. Our results show that the total PE in the GlobSnow FSC product is significantly higher than the originally provided statistical error. This is due to deficiencies in the parameterization of the applied forward modelling approach, in particular in the consideration of the forest canopy effects.
Related Topics
Physical Sciences and Engineering
Earth and Planetary Sciences
Computers in Earth Sciences
Authors
Miia Salminen, Jouni Pulliainen, Sari Metsämäki, Jaakko Ikonen, Kirsikka Heinilä, Kari Luojus,