Corrigendum: An Application of Image Processing Techniques in the Calibration of Catastrophe Models

Titus Kipkoech Rotich; Artur Hambardzumyan; Eliud Koech; Olga Poghosyan; Joseph Mungatu

doi:10.66573/001c.165343

Rotich, Titus Kipkoech, Artur Hambardzumyan, Eliud Koech, Olga Poghosyan, and Joseph Mungatu. 2026. “Corrigendum: An Application of Image Processing Techniques in the Calibration of Catastrophe Models.” Variance 19 (August). https://doi.org/10.66573/001c.165343.

View more stats

In Table 1, the F₁ Score for the GLM is incorrect. It should be 0.6675. The corrected table is below.

Table 1.A summary of the fitted GLM and CNN (U-Net) model performance measures

Model	MAE	Precision	Accuracy	FAR	Recall	F₁ Score
U-Net	1.7175	0.7010	0.7104	0.2989	0.8497	0.7604
GLM	1.9797	0.6578	0.6701	0.3401	0.6775	0.6675

In the paragraph that introduces Table 1, the text should read:

“We used these models to calculate various measures, as summarized in Table 1. The U-Net model outperformed the GLM on all measures. We then used the fitted U-Net model to predict the frequency of hail and compared it with the test data. Figure 12 shows the results.”