Floods and droughts are the events that most threaten crop production; however, the impact of floods on crops is still not fully understood and often under-reported. Nowadays, multiple sources of information and approaches support the estimation of agricultural losses due to floods. This study aims to understand the differences in agricultural loss estimates provided by two conceptually different approaches (crop models and expert-based models), evaluating their sensitivity to flood hazard inputs. We investigated the challenges in flood agricultural loss assessments referring to a case study for which, in addition to model simulations, information from surveys and on-site inspections were available. Two crop models (APSIM and WOFOST) and the expert-based model AGRIDE-c were applied to evaluate agricultural yield losses after the flood event of the Panaro River (Emilia-Romagna, Northern Italy) that took place on the 6 December 2020. Two modelling tools were used to reproduce the event: the hydraulic model HEC-RAS and the image-based tool FwDET. Additionally, surveys among local farmers were conducted in the aftermath of the event to evaluate the flood features (water depth, extent and duration) and crop losses. The main findings of the study are that APSIM and WOFOST provide similar estimates of yield losses, while AGRIDE-c tends to underestimate yield losses when the losses over the entire study area are evaluated. The choice of the flood simulation technique does not influence the loss estimation since the difference between the yield loss estimates retrieved from the same model initialized with HEC-RAS or FwDET was always lower than 2%. Information retrieved from the surveys was not sufficient to validate the damage estimates provided by the models but could be used to derive a qualitative picture of the event. Therefore, further research is needed to understand how to effectively incorporate this kind of information in agricultural loss estimation.

Common main limitations affect standard approaches to regional frequency analysis (RFA) of rainfall extremes. Our study focuses on three of them that are rather frequent: regional models address (a) a single duration, or (b) a single exceedance probability at a time, and/or (c) hold a small-to-medium homogeneous region only. We use unsupervised ensembles of artificial neural networks (ANNs) to set up four alternative RFA models of sub-daily rainfall extremes. These are fed with annual maximum series of rainfall depth of any length collected at 2238 raingauges in a large and climatically and morphologically heterogeneous region. Our models can predict parameters of a Gumbel distribution for any location within the study area and any duration in the 1–24 h range. Prediction is based on mean annual precipitation (MAP), or on twenty morphoclimatic covariates. Validation is performed over an independent set of 100 gauges, where locally fitted Gumbel distributions are used as reference. A common literature approach where Gumbel parameters are functions of MAP is used as benchmark. Our results show that multivariate ANNs remarkably improve the estimation of percentiles relative to the benchmark approach. Finally, we show that the very nature of the proposed ANN models makes them suitable for interpolating predicted sub-daily rainfall quantiles across time-aggregation intervals and space and can be adapted for considering more flexible target frequency distributions (e.g. 3-parameter models).

This study provides an in-depth analysis of the frequency of extreme streamflow in Italy, adopting the innovative perspective of the theory of records, and focusing on record-breaking floods. (i.e., annual maximum series, AMS) observed in Italy between 1911 and 2020. Our research employs an extensive dataset of 522 annual maximum series (AMS) of streamflow observed across Italy between 1911 and 2020. We consider three time intervals (1911–2020, 1911–1970, and 1971–2020), and we define pooling-groups of AMSs based on (a) hydrological (e.g., catchment size, mean annual precipitation, etc.) and (b) spatial proximities of the gauged sites. First, within each group and for each time period, we compute the regional average number of record-breaking events (NRbins). Second, with a series of resampling experiments that preserve the spatial correlation among the AMSs, we test the hypothesis that NRbins result from a group of stationary sequences. Our results show spatially coherent patterns of an increasing number of record-breaking floods in central and in northeastern Italy over the last 50 years. In the same time interval, significant deviations in the regional number of record-breaking events from what would be expected for stationary flood sequences seem to be more common in drier climates or at higher altitudes, while the catchment size does not seem to be a meaningful descriptor.

Recent literature shows several examples of simplified approaches that perform flood hazard (FH) assessment and mapping across large geographical areas on the basis of fast-computing geomorphic descriptors. These approaches may consider a single index (univariate) or use a set of indices simultaneously (multivariate). What is the potential and accuracy of multivariate approaches relative to univariate ones? Can we effectively use these methods for extrapolation purposes, i.e., FH assessment outside the region used for setting up the model? Our study addresses these open problems by considering two separate issues: (1) mapping flood-prone areas and (2) predicting the expected water depth for a given inundation scenario. We blend seven geomorphic descriptors through decision tree models trained on target FH maps, referring to a large study area (∼ 10⁵ km²). We discuss the potential of multivariate approaches relative to the performance of a selected univariate model and on the basis of multiple extrapolation experiments, where models are tested outside their training region. Our results show that multivariate approaches may (a) significantly enhance flood-prone area delineation (accuracy: 92 %) relative to univariate ones (accuracy: 84 %), (b) provide accurate predictions of expected inundation depths (determination coefficient ∼ 0.7), and (c) produce encouraging results in extrapolation.

We propose an effective and reproducible framework to assess the informative content of teleconnections (or climate indices) for representing and modeling the frequency regime of rainfall extremes at regional scale. Our dataset consists of 680 annual maximum series of rainfall depth, with 1 and 24 h durations, located in Northern Italy. We compute at-site time series of L-moments (i.e., the mean and the L-coefficient of variation) through sliding time windows; then we discretize the study region into tiles, where L-moments time series are averaged. We compute the 30-years sliding mean for six teleconnections: North Atlantic Oscillation, Pacific Decadal Oscillation, East Atlantic – West Russia pattern (EA-WR), El Niño Southern Oscillation, Mediterranean Oscillation Index, and Western Mediterranean Oscillation Index (WeMOI). Then, we calculate L-moments-teleconnection Spearman correlations for single sites and for tiles with several resolutions, and retain correlations with p-values ≤ 0.05. We observe spatial patterns of strong correlation between several teleconnections and gridded L-moments. These spatial patterns are clearly visible at various tiles' resolutions, and may be used for setting up regional prediction models. The strongest influence is detected for the sliding mean on the WeMOI and EA-WR. Finally, we show a preliminary application of climate-informed regional frequency analysis, through a hierarchical framework, where the L-moments are modelled as functions of teleconnections. We observe high variability of teleconnection-driven predictions of rainfall percentiles, and an increase in overall goodness-of-fit of the climate-informed regional models relative to stationary models. Overall, our research suggests promising pathways for climate-informed local and regional frequency analysis of rainfall extremes, and describes a general method, that can be adapted to different geographical and climatic contexts, as well as environmental variables.