The growing concern about the possible effects of climate change on flood frequency regime is leading Authorities to review previously proposed reference procedures for design-flood estimation, such as national flood frequency models. Our study focuses on Triveneto, a broad geographical region in North-eastern Italy. A reference procedure for design flood estimation in Triveneto is available from the Italian NCR research project “VA.PI.”, which considered Triveneto as a single homogeneous region and developed a regional model using annual maximum series (AMS) of peak discharges that were collected up to the 1980s by the former Italian Hydrometeorological Service. We consider a very detailed AMS database that we recently compiled for 76 catchments located in Triveneto. All 76 study catchments are characterized in terms of several geomorphologic and climatic descriptors. The objective of our study is threefold: (1) to inspect climatic and scale controls on flood frequency regime; (2) to verify the possible presence of changes in flood frequency regime by looking at changes in time of regional L-moments of annual maximum floods; (3) to develop an updated reference procedure for design flood estimation in Triveneto by using a focused-pooling approach (i.e. Region of Influence, RoI). Our study leads to the following conclusions: (1) climatic and scale controls on flood frequency regime in Triveneto are similar to the controls that were recently found in Europe; (2) a single year characterized by extreme floods can have a remarkable influence on regional flood frequency models and analyses for detecting possible changes in flood frequency regime; (3) no significant change was detected in the flood frequency regime, yet an update of the existing reference procedure for design flood estimation is highly recommended and we propose the RoI approach for properly representing climate and scale controls on flood frequency in Triveneto, which cannot be regarded as a single homogeneous region.

One of the most common tasks in hydrology is to produce an
accurate estimation of the design flood at ungauged or sparsely gauged river
cross-sections

The objective of our study is threefold: (1) to check whether climatic and
scale controls on flood frequency regime in Triveneto are similar to the
controls that were recently found in Europe; (2) to verify the possible
presence of changes in flood frequency regime by looking at changes in time
of regional L-moments of annual maximum floods

The selection of the most suitable regional parent distribution is a key
aspect in any regional flood frequency analysis. The scientific literature
recommends using the L-moment ratio diagrams for addressing this task (see
e.g.

Location of the 76 Triveneto gauging stations (black points) considered in the present study. Grey lines represent catchments boundaries. Pieve di Cadore, an example target basin, is highlighted in red.

Catchment characteristics of the Triveneto data set (black points).
Grey points represent the Austrian, Slovakian and Italian data sets
considered in

L-moment ratio diagrams for the subsets defined by intermediate
area

The study region consists of 76 catchments (see Fig.

Figure

The routinely hydrological and hydraulic risk assessments of dams located in
Triveneto are related to the growing concern about the possible effects of
climate change on the region flood-frequency regime. A second main objective
of this study is therefore to verify the possible presence of changes in
flood frequency regime by looking at the variability of regional L-moments of
annual maximum floods in time. In fact, according to

Left: L-moment ratio diagram for three-parameter distributions.
Blue, red and green points represent the record length weighted average of
L-Cv and L-Cs in Triveneto for PRE1965, POST1965 and POST1965 without 1966
sub-samples, in this order. Right: dimensionless growth-curves associated
with the above mentioned three sub-samples. The choice of the corresponding
three-parameter distribution was made by applying Hosking and Wallis test
(see

In this context, we decided to check the presence of changes in the frequency regime of AMS in Triveneto with reference to 1965, by dividing the total AMS sample into two sub-samples: PRE1965 (all years before 1965, for a total of 1217 station-years of data) and POST1965 (all years from 1965, included; 1148 station-years of data). It is important to highlight that POST1965 sub-sample includes 1966, which was characterized by exceptionally heavy floods in Triveneto, and Italy in general. Year 1966 is well-known for the 4–5 November floods in Florence, but the entire autumn of 1966 was particularly difficult for North-East Italian regions: Triveneto counted 87 deaths and about 42 000 displaced people (data from the Italian NCR). Hence, we decided to test the importance of this year in the flood frequency regime, by analysing a further sub-sample, obtained by removing AMF of 1966 from the POST1965 sub-sample (1117 station-years of data).

For each of the considered three sub-samples we plotted the record length
weighted average of L-Cv and L-Cs in the corresponding diagram and estimated
the associated regional three-parameter distribution (see Fig.

Left: L-moment ratio diagram for three-parameter distributions.
Blue, red, green and grey points represent the record length weighted average
of L-Cv and L-Cs for PRE1965, POST1965 and POST1965 without 1966 RoI
sub-samples and total RoI sample, in this order. Right: dimensionless
growth-curves associated with the above mentioned four sub-samples. The
choice of the corresponding three-parameter distribution was made by applying
Hosking and Wallis test (see

It is important to highlight that the dimensionless growth-curves in
Fig.

The third and last main objective of our study is to update the Triveneto
reference procedure for flood frequency estimation. Therefore we tested and
applied the so called Region of Influence (RoI) approach (see

The RoI approach arranges catchments according to their dissimilarity with
the target site, pooling together only the most similar ones. The number of
catchments to include in the RoI is determined by considering the return
period of the design flood

We report herein an application example for the Pieve di Cadore dam, for
which a 49 AMS (from 1951 to 1999) has been compiled from inflow series to
the artificial reservoir. The RoI sample consists of 16 catchments, for a
total of 555 observations, which enable one to get a reliable prediction of
the 100-year quantile according to the “

Our study confirmed the value of including climatic and
physiographic information in statistical regionalization. Catchment
descriptors as mean annual precipitation (MAP), basin area, mean and minimum
elevation and catchment location can be used as surrogates of climate and
scale controls. In particular, our study shows that climatic and scale
controls on flood frequency regime in Triveneto are similar to the controls
that were recently found in Europe by

By splitting the Triveneto sample into two distinct sub-samples (i.e. before and after 1965) and then removing the 1966 year from the second one, we observe very similar conditions in terms of flood frequency regime, highlighting the remarkable influence that a single exceptional year (1966) can have in the representation of flood frequency regime.

Finally, our update of Triveneto AMS data set falsified the main assumption
of the current reference procedure for design-flood estimation in Triveneto

The study is part of the research activities carried out by the working group: Anthropogenic and Climatic Controls on WateR AvailabilitY (ACCuRAcY) of Panta Rhei – Everything Flows Change in Hydrology and Society (IAHS Scientific Decade 2013–2022).