Communication strategies to address geo-hydrological risks: 2 the POLARIS web initiative in Italy 3

. Floods and landslides are common phenomena that cause serious damage and pose a severe threat to


55
Although, in the last few years, the Italian scientific community has begun to communicate information on geo-

109
Risk perception is also important to determine the attitude towards risks and, when information campaigns and 110 risk communication strategies are designed, the public perception should be known (Plapp & Werner, 2006).

111
Risk perception is a subjective assessment of the hazard occurrence's probability and people's feelings of the

120
If it is globally accepted that risk perception has strong implication for the success of risk communication. It is 6/25 governmental authorities, national and local agencies, the majority resulted one-way efforts, focused solely on 133 improving hazard knowledge or raising risk awareness, mostly regarding flood hazard.

135
In this work, we adopt the terminology and definitions used in Google Analytics. We use the term session to

146
This role should be attained working in two directions: (i) providing mass media (e.g., journalists) with correct 147 and reliable information, which they can communicate (spread) further to the broader civil society, and (ii)

148
adopting less technical and more widely comprehensible language to better engage citizens. Figure

178
The "Events" section publishes information on specific meteorological events in Italy, using text, maps, videos, 179 photographs, and drawings. In this section, specific icons were designed to define the type of the geo- represent an important source of information and data for the mass media.

186
The "Alert zones" section provides information for 134 Alert Zones defined by the Italian National Civil

187
Protection system to forecast geo-hydrological hazards, including landslides and floods. The section provides 188 the possibility to query a number of information items, and a sidebar offers access to different thematic layers 189 and maps for each Alert Zone.

190
The "Are you prepared?" section offers information on suitable (and unsuitable) behaviours to adopt before,

217
Since POLARIS is published in Italian, it is not surprising that the sessions mainly originate from Italy (91%).    (Table 1).

248
To  Fig. 4). Again, the 265 analysis did not reveal a periodic trend.

266
To gain a better understanding of the temporal distribution of the user access, and to identify peak values, we 267 used the daily number of users and pageviews obtained from Google Analytics. We then related the peak values

270
(iv) the publication of press releases that used our data, and (v) the promotion of the website through media.
271 Figure 6 shows the daily user statistics (Fig. 6a), and a comparison between users and number of pageviews 272 (Fig. 6b), for the 21-month period of website publication, with icons located to identify possible relations. We   originates daily, we found that other peaks were the consequences of the daily activity of users from government 299 offices or agencies. In Fig. 6b, we plotted the users and the pageviews data together. The mean number of pages 300 per user, in the entire period, was 2.5; however, the inspection of Fig. 6b

312
Analysing the number of referrals from the social networks, corresponding to 14% of the total, we found that

320
To define the extreme rainfall conditions that occurred in Italy, we exploited an analysis based on hourly rainfall 321 measurements. The analysis was performed in the 84-day period between 1 August and 23 October 2015. We 322 exploited sub-hourly rainfall measurements by more than 2000 rain gauges distributed over the entire Italian 323 territory. According to the method described by Rossi et al. (2015), the empirical cumulative distribution 324 function (ECDF) of the cumulative rainfalls has been modelled for each rain gauge. The function allows the 325 calculation of the non-exceedance probability for any given cumulative rainfall and for a set of predefined 326 durations (3, 6, 12, 24 h), which estimates the non-exceedance probability of the cumulated rainfalls, for each 327 rain gauge. To obtain a continuous representation for the entire Italian territory, the rain gauge data have been 328 interpolated using an inverse distance weighted (IDW) algorithm. This process resulted in a set of four (one for 1 2 / 2 5 each duration) raster maps that show the non-exceedance probability of the cumulative rainfalls. The maps have 330 been analysed to identify the days when at least 10% of the Italian territory has been interested by a non-331 exceedance probability of 80%. This probability value corresponds to cumulative rainfall events that can be 332 defined as extreme events and that could have triggered geo-hydrological events.

333
The results of the analyses showed that, in the considered period, the extreme conditions occurred six times for 334 a duration of 3 h, 12 times for a duration of 6 h, 15 times for a duration of 12 h, and seven times for a duration 335 of 24 h. We plotted these extreme conditions in the daily distribution of Facebook users shown in Fig. 7. We