Using feature-based verification methods to explore the spatial and temporal characteristics of the 2019 chlorophyll-<i>a</i> bloom season in a model of the European Northwest Shelf

Abstract

Abstract. Two feature-based verification methods, thus far only used for the diagnostic evaluation of atmospheric models, have been applied to compare ∼7 km resolution pre-operational analyses of chlorophyll-a (Chl-a) concentrations to a 1 km gridded satellite-derived Chl-a concentration product. The aim of this study was to assess the value of applying such methods to ocean models. Chl-a bloom objects were identified in both data sets for the 2019 bloom season (1 March to 31 July). These bloom objects were analysed as discrete (2-D) spatial features, but also as space–time (3-D) features, providing the means of defining the onset, duration and demise of distinct bloom episodes and the season as a whole. The new feature-based verification methods help reveal that the model analyses are not able to represent small coastal bloom objects, given the coarser definition of the coastline, also wrongly producing more bloom objects in deeper Atlantic waters. Model analyses' concentrations are somewhat higher overall. The bias manifests itself in the size of the model analysis bloom objects, which tend to be larger than the satellite-derived bloom objects. The onset of the bloom season is delayed by 26 d in the model analyses, but the season also persists for another month beyond the diagnosed end. The season was diagnosed to be 119 d long in the model analyses, compared to 117 d from the satellite product. Geographically, the model analyses and satellite-derived bloom objects do not necessarily exist in a specific location at the same time and only overlap occasionally.