Technical note: Space–time analysis of rainfall extremes in Italy: clues from a reconciled dataset

2.1 Data sources

As a follow-up of the activities of the Italian National Group for the Prevention of the Hydrogeological Disasters (GNDCI) a comprehensive nationwide hydrological information system has been set up, within the “CUBIST project”, funded by the Italian Ministry of Education and Research within the funding PRIN 2005 (Italian Research Projects of National Relevance). The database includes about 6000 pluviographs and pluviometers, 700 temperature stations and about 400 river basins (Claps et al., 2008) and is available at http://www.cubist.polito.it; (last access: 26 October 2017). In detail, the database includes rainfall data from 1900 to 2001 (depending on the region) and constitutes the first important attempt to make the large Italian hydrological heritage freely available in a computer-readable format. The annual maxima data for different durations included in the CUBIST dataset are extracted with a sliding-windows process from manual tipping-bucket rain gauge data, equipped with a recording system that writes on diagram paper. Further information on the characteristics of the stations can be found in Acquaotta et al. (2016). The number of data per year is not constant across the analysed period, as it has increased in time as more stations have been installed in recent years. Data availability decreases in the period of the Second World War, as many records were missed in that period. After 1980, with the progressive dismissal of the SIMN and the development of the local hydrographic authorities, data availability decreases rapidly until 2001, when the rain gauges still under the SIMN were taken over by the local operational centres.

After the late 1980s, indeed, the local environmental agencies started to support the SIMN in its work. Gradually, the 21 regional hydrological services took over the networks and the tasks of the national one. In this period most of the old manual tipping-bucket rain gauges have been substituted with automatic stations, similar to the one described in Acquaotta et al. (2016) for the Piedmont region. Each hydrological service adopted its own rules for the publication of the collected data and, even if the Italian law adopted an open-source policy for the public data for non-commercial uses (under D.Lgs.82/2005, D.Lgs.36/2006, D.M.10/11/2011, L.221/2012, D.Lgs.179/2012 and L.114/2014), an updated database of the annual rainfall maxima for sub-daily duration at the national scale is still lacking. For the scope of this research, the different agencies have been contacted, and the regional annual maxima datasets for sub-daily durations were requested. The regions of Italy are shown in Fig. 1 together with the type of data provided, which will be described in the following section. Table 1 lists the names of the local authorities and the regional codes, aimed at identifying them in the database. The public availability of the original dataset is also reported.

Figure 1Names of the Italian regions and type of datasets provided by the regional authorities. The cases refer to the bullet list of Sect. 2.2.

Download

Table 1Regions of Italy with the assigned code and the related local Operational Center with references to the availability of digitized data.

¹ Manfreda, S., Sole, A. and De Costanzo, G.: Le precipitazioni estreme in Basilicata, Editrice Universo Sud, 2015.
² ARPACAL: Centro Funzionale Multirischi, http://www.cfd.calabria.it/, (last access: 1 August 2016).
³ ARPAL: Consultazione Dati Meteoclimatici, http://www.cartografiarl.regione.liguria.it/SiraQualMeteo/Fruizione.asp, (last access: 1 August 2016).
⁴ ARPA Lombardia: Progetto Strada, http://idro.arpalombardia.it/pmapper-4.0/map.phtml, (last access: 1 August 2016).
⁵ Protezione Civile Regione Marche: Annali Idrologici Regione Marche, http://console.protezionecivile.marche.it, (last access: 1 August 2016).
⁶ ARPA Piemonte: Banca dati meteorologica, http://www.regione.piemonte.it/ambiente/aria/rilev/ariaday/annali/meteorologici, (last access: 1 August 2016).
⁷ Protezione Civile Puglia: Annali Idrologici – Parte I, http://www.protezionecivile.puglia.it/centro-funzionale/analisielaborazione-dati, (last access: 1 August 2016).
⁸ SIR Toscana: Settore Idrologico Regionale, http://www.sir.toscana.it/, (last access: 1 August 2016).
⁹ Centro Funzionale di Protezione Civile Provincia Autonoma di Trento: Meteotrentino, http://www.meteotrentino.it/, (last access: 1 August 2016).
¹⁰ the autonomous provinces of Trento and Bolzano–Alto Adige, together, constitute the region Trentino-Alto Adige (CD: 22).

Download Print Version | Download XLSX

2.2 Cleaning and merging operations

Merging and harmonizing the different datasets is a quite long and difficult operation, that is still ongoing. The different operational centres provided different types of datasets, with different temporal coverages and spatial reference systems. Duplicate stations are often present in the databases of neighbouring regions.

The first steps of this work have been carried out at the regional scale. For each region all the data falling inside the regional boundaries have been considered. These data, according to the setting of the databases of the local operational centres, could belong to one of these three categories:

• data from the CUBIST database for the 1900–2001 period already available from the former national service

• data provided by the regional authority

• data provided by the regional authorities of the neighbouring regions that extend beyond their regional borders.

Observations dating from before 1916 have been discarded, as they are considered not significant and too unevenly distributed. Considering that most of the provided data have been validated from the related authorities, they are considered reliable and, at first, included directly in the I-RED. For information on the validation procedures, please refer to the Appendix Appendix A and to Barbero S. et al. (2017). In the presence of inconsistencies between the type (b) and type (c) data, preliminary manual merging was carried out. The sources of the inconsistencies could be various, according to the evolution of the monitoring systems of the different regions, and these inconsistencies are often due to the joint management of interregional basins. The different regional authorities often have adopted different codes and/or names for the same station, the first step has been thus to identify the presence of duplicate stations with the same or a similar name covering different time intervals. Sometimes, even for the same station, neighbouring regions can provide different data for the same years. This can be, for example, due to the fact that sometimes regions share rainfall data before their validation and official publication. If the same station was found in the database of more neighbouring regions, the first attempt of merging the series together was carried out by analysing the data recorded year by year. If the merging was not feasible, higher priority was given to the data provided by the authority of the considered region (that is usually also the owner of the network). This allowed the presence of duplicate series in the I-RED to be avoided.

Once the type (b) and type (c) datasets were merged for each region,, the resulting dataset has to be merged with the type (a) dataset. This operation has been quite complex, as the overlapping period between the different dataset was different for each region and because most of the authorities did not track the change in the name and code of the stations. The different procedures performed, according to the type of the dataset that the region has provided (as reported in Fig. 1) can be summarized as follows:

Figure 2(a) Data availability per year in the I-RED and CUBIST databases (the smallest value across the five considered durations is reported per year). (b) Number of series longer than fixed threshold values in the I-RED databases per duration (null values are ignored). (c) Number of null values per duration. (d) Length of the series in the I-RED database represented in space: the colour refers to the minimum length among the five available durations. If more stations overlap due to the resolution of the picture, the one with the longer series appears on the top.

Download

1. Regions that digitized the whole SIMN database for their areal domain and provide a complete merged database. The provided data were inserted in the I-RED without editing and without considering the CUBIST series. Only for the Abruzzo and Molise regions some preliminary refinement was needed, as the two regions were divided in 1963, and the databases of the two regions partially overlap. The stations were then divided according to the actual regional boundaries and the duplicate series removed.

2. Regions that provided datasets including data from their actual regional network partially merged with subsets of digitized data from the SIMN Hydrological Yearbooks. As not all the SIMN datasets were digitized from the local authorities, the dataset lacked part of the stations included in the CUBIST database. To maximize the available information, data from the regional databases and the CUBIST one were manually analysed and merged, in order to avoid duplicate values. Stations with the same name and similar coordinates were merged together in the presence of a 2-year consistent overlapping period. In the presence of inconsistencies between the values recorded by the two stations in the overlapping period, they were treated as different stations and renamed. If it was not possible to remove any doubt, the stations were considered as separate entities. For the Liguria region, the information in ARPAL (2013) was used to overcome the lack of information on the continuity of the series.

3. Regions that provided two different datasets: one containing the whole digitized data from the SIMN stations and another containing the digitized data from their actual networks. The data of the two databases were merged together, the overlapping period manually analysed to avoid overlapping, and the CUBIST database ignored. The operation was made possible by the collaboration of ARPA Piemonte, for Piedmont and Valle d'Aosta, and of the Università degli Studi di Firenze, for Tuscany.

4. Regions that provided only the data recorded from the network they actually manage. All the information concerning the SIMN stations was lacking. The provided dataset was therefore merged with the whole CUBIST database for the considered regions. Duplicate values were excluded when manually analysing the overlapping period, if present.

With the application of the above described rules, 20 complete regional datasets have been obtained. The regional datasets were finally merged together to generate the I-RED. After the merging phase some reliability check has been performed, in order to detect any problematic or incorrect information. These include the identification and removal of the duplicate data or stations, and reliability checks on the larger values of the dataset, comparing them to the absolute record-breaking events for all the durations (see Libertino, 2017), aimed at detecting inconsistencies in rainfall series. If any suspect value was found, its year of occurrence was compared, when referring to recent years, with the data from event reports or newspapers. If the data refers to a SIMN station, the Hydrological Yearbooks were consulted. If no evidence was found, the related authority was contacted. Most of the operations need human supervision, and thorough verification work. If it is not possible to remove any doubt the suspect value is discarded.

Due to the complexity of the check operations, further efforts and collaborations with the regional authorities are still ongoing to increase the consistency of the database. Nevertheless, to date (October 2017) the I-RED includes more than 4500 stations nationwide and constitutes the largest updated dataset of annual maxima for Italy.

Considering that most of the regional authorities supervise the use and widespread dissemination of their datasets in order to prevent improper use, a detailed description on how to access the I-RED is reported in the Data Availability section.

In the following, the spatio-temporal distribution of the assembled data will be described.