Open AccessTHE USE OF AUTOMATED SYNOPTIC TYPING FOR CONDITIONAL VERIFICATION OF NUMERICAL WEATHER PREDICTION IN THE PERM REGION
Author(s) -
Sergey V. Kostarev,
AUTHOR_ID,
Igor N. Rusin,
AUTHOR_ID
Publication year - 2021
Publication title -
geografičeskij vestnik
Language(s) - English
Resource type - Journals
ISSN - 2079-7877
DOI - 10.17072/2079-7877-2021-1-68-80
Subject(s) - global forecast system , anticyclone , meteorology , advection , forecast verification , numerical weather prediction , synoptic scale meteorology , climatology , environmental science , principal component analysis , forecast skill , computer science , geography , geology , artificial intelligence , thermodynamics , physics
The article discusses the possibility of verification of short-term 2-meter air temperature forecasts with the Global Forecast System and Global Environment Multiscale numerical weather prediction models depending on the observed synoptic type (a case study of the Perm region for the period 2018–2019). As part of the study, we have developed a system of automated determination of synoptic type based on a two-stage procedure, including decomposition of mean sea level pressure fields via principal component analysis and the subsequent clustering of decomposition coefficients using K-means. It has been established that GFS forecasts are more dependent on synoptic type in summer than in winter. The decline of forecast quality, expressed in systematic underestimation of forecast temperature by 0.6°–1.2°, is noted for synoptic types associated with warm air advection.: In contrast, GEM forecasts tend to lack accuracy in winter. A sharp decrease in forecast quality has been discovered in the central area of anticyclone at night, when the forecast accuracy drops to 44%. The obtained results could be useful in operational forecasting and model postprocessing.