Dataset for "Exploring the viability of a machine learning based multimodel for quantitative precipitation forecast post-processing"

Title: Dataset for "Exploring the viability of a machine learning based multimodel for quantitative precipitation forecast post-processing"

Authors: Luca Monaco, Francesco Laio, Roberto Cremonini, Giovanni Bindi, Secondo Barbero

Description:This dataset supports the study presented in the paper "Exploring the viability of a machine learning based multimodel for quantitative precipitation forecast post-processing". The work focuses on improving quantiative precipitation forecast over the Piedmont and Aosta Valley regions in Italy by blending outputs from four Numerical Weather Prediction (NWP) models using machine learning architectures including Multi-Layer Perceptrons (MLPs), U-Net and Residual U-Net as Convolutional Neural Networks (CNNs), and NWIOI as observational data (Turco et al., 2013).

Observational data from NWIOI serve as the ground truth for model training. The dataset contains 406 gridded precipitation events from 2018 to 2022.

Dataset contents:

obs.zip: NWIOI observed precipitation data (.csv format, one file per event)

subsets.zip: Events dates for 10 different training-validation-test sets, retrieved with 10-fold cross validation (.csv format, one file per set and per split)

domain_mask.csv: Binary mask (1 for grid points in the study area, 0 otherwise)

allevents_dates_zenodo.csv: Summary statistics and classification of all events by intensity and nature, used for subsets creation with 10-fold cross validation

Citations:

NWIOI: Turco, M., Zollo, A. L., Ronchi, C., De Luigi, C., & Mercogliano, P. (2013). Assessing gridded observations for daily precipitation extremes in the Alps with a focus on northwest Italy. Natural Hazards and Earth System Sciences, 13(6), 1457–1468.

Related Repository: https://github.com/MonacoL/QPFMultimodel