Subject: Acceptance test of precipitation re-distribution algorithms in QC2 (if precip. missing for >= 1 days)
Participants: Per-Ove, Solfrid, Ole Einar, Matthias (author of this page)
Not present: Paul
Issues discussed:
Test with different kinds of precipitation data: Convective, terrain (complex/flat), proximity to Swedish border
Precipitation measurements have already 10% measurement inaccuracy; good result if we can achieve a certain (yet unknown) percentage accuracy in estimated values
Convective case especially challenging: Half of stations could have precipitation, other half not.
Automatic QC2 should give better result than HQC –> Could this be tested? (Hypothetical HQC test dataset could be used and compared with automatic QC2)
Tests of QC2 should be carried out using historic data. Results should be compared with measurements.
We have enough data for extensive evaluation and reliable statistical error estimations.
Should HQC still check results of automatic QC2 (probably not feasible)
Is QC2 giving better results than values of numerical models that are stored in KDVH. –> Test this