1 Introduction

According to common understanding, risk is an inherent component of the business activity associated with result expectations which can be disregarded due to unforeseeable events. Despite the diversity of approaches and methodologies of risk analysis and evaluation, some commonly accepted general assumptions on risk can be outlined (OECD 2009). Households and societies manage risks through multiple complementary strategies that can be separated into ex-ante (before a shock occurs) and ex-post (after a shock has occurred) measures (Heltberg et al. 2009). The most suitable strategies can be defined only after a risk analysis, starting from the study of the hazards, i.e. the unpredictable events changing the outcome expectations. The greater the entrepreneur’s knowledge of the factors that can influence the results, the lower the company’s exposure to risk is. Risk management, if successful, translates into resilience, which is the ability to avoid and reduce (or recover from) the negative impacts of the hazards (Bednar-Friedl et al. 2022). From a socio-economic point of view, resilience can refer to “how a system, community or individual can deal with disturbance, surprise and change” (Mitchell and Harris 2012).

As for weather-related hazards, many studies focused on risks in the context of climate change (IPCC 2012, European Commission 2021). Increasing climate extremes are expected to lead to an increase in “disasters” (Alexander 2016), defined as serious alterations in the normal functioning of human production systems in vulnerable conditions. The approach to risk assessment mainly adopted at an international level, also suggested by the Intergovernmental Panel on Climate Change (IPCC) and the Joint Research Centre of the European Commission (JRC), includes a robust analysis of the three components of the risk: hazard, exposure and vulnerability, where the hazard is the occurrence of extreme events potentially damaging; the exposure refers to the presence of systems subject to hazards (in this case, farms, crops and their values), while the vulnerability refers to the propensity of exposed systems to suffer adverse effects when impacted by hazard events (Handmer et al. 2013).

As known, the agricultural sector is crucially dependent on environmental, climate and weather conditions and the relationships between both food production and quality and climate factors are among the main objects and concerns in the studies on climate change (Bezner Kerr et al. 2022; Anderson et al. 2023; Bozzola et al. 2018; Ben Mhenni et al. 2021; Wang et al. 2017). At a global level, crops’ productions already show effects in yields, quality and distributions of pests and diseases, due to increasing temperatures, changing precipitation patterns and greater frequency of some extreme events, also leading to wide food safety issues and famine (Mbow et al. 2019; Charalampopoulos and Droulia 2021). The shift from the past risk (linked to a known probability distribution) to uncertainty of meteorological disasters directly influences farmers’ management, e.g. higher likelihood of floods in Europe (European Commission 2021; Handmer et al. 2013). In view of the direct relationship between the variability of production yields and uncertainty of meteorological conditions, there is a latent demand for vulnerability analysis and risk management tools in the primary sector (Monteleone et al 2023; Shah et al. 2021) that could be a key factor for the resilience and the very survival of farms (Glauber et al. 2021).

Many countries adopted risk policy schemes for agriculture in order to support farmers facing weather-related risks (Bielza Diaz-Caneja et al. 2009; Glauber et al. 2021; Clarke et al. 2012; Cafiero et al. 2007). The topic of risk management, as known, entered the European debate on the Common Agricultural Policy (CAP) only in recent decades (Barral 2023), receiving legislative attention in the Community with the approval of the Health Check (EU reg. 73/2009) which offered for the first time Member States the possibility of using a part of the financial resources intended for direct payments to support farmers in case of damages due to adverse weather conditions, plant diseases, epizootics and environmental emergencies. The EC Regulation 1305/2013 introduces the toolkit for risk management policy, supporting crop insurances (art. 37), mutual funds (art. 38) and income stabilization tools (art. 39). The new European policy tools had not the hoped-for success at European level: in the 2014–2020 period, approximately 2.67 billion euros were activated, less than 0.4% of the entire CAP budget (Bardají et al. 2016).

In Italy, the situation is substantially different: the Government planned an expenditure equal to approximately 1.64 billion euros, which alone represents over 60% of the total amount in Europe, and almost 8% of the available resources at a national level for the rural development policy (Bardají et al. 2016). Due to the long tradition of public intervention in this field, which started in 1970 with the establishment of the National Solidarity Fund for natural disasters in agriculture (NSF, law n. 364/70), in Italy, there are two distinct strategies to cover production risks: compensatory payments to farmers as ex-post strategy and contributions for the underwriting of insurance contracts as ex-ante strategy (Cafiero et al. 2007). Only in the period 2003–2018, the declared damages attributed to natural disasters exceeded 27 billion euros (Parisse et al. 2020). Across the last decades, a significant shift of public investments can be observed from compensatory measures to crops’ insurances. At the beginning, insurance contracts were concentrated on damages due to hail events, while during the last decades other kinds of contracts have been introduced and implemented, the pluri-risk insurances, covering both droughts and floods, for instance (ISMEA 2023b). More recently, an experimental mutual fund (called AgriCat Fund, ISMEA 2023b) has been launched in 2022 covering the catastrophic damages due only to drought, frost and floods. The evolution of the Italian risk policy scheme for agriculture shows the choice of insurances as the main and almost the only strategy (transferring the risk) to cope with weather-related risks, with very low effectiveness in relation to the public expenditure poured into the system. In fact, in Italy, farmers’ participation in crop insurance is very low, never exceeding 20% of agricultural gross domestic production (GDP) in the last 15 years. Only 11% of farms are involved, and there is a huge portfolio concentration in few areas of the country, with 80% of insured GDP in northern Italy (ISMEA 2023a). Moreover, the ex-post compensation paid by the NSF essentially managed less than 5% of the declared damages. A study conducted on the main strategies adopted by Italian farmers highlighted that the approach is still strongly oriented towards the technical means (agricultural practices, pesticides, fertilizers and irrigation), rather than available financial tools (Pontrandolfi et al. 2016). This is especially true for medium-small farms, the most represented in Italy, giving a sign of lack of information and of “mistrust” in the risk management policy scheme.

These aspects raise the crucial question whether scientific representation and related knowledge used in risk policy design in Italy have been incomplete or incorrect in its theoretical setting and implementation, firstly the choice to identify the risk management almost exclusively with crops’ insurances (Severini et al. 2019; Cordier and Santeramo 2020; Capitanio 2022). Among the main weaknesses, there is the lack of objective analyses on weather-related hazards behaviour that justify the preference of insurances over other risk management and/or adaptation strategies (e.g. structural and management measures). In fact, even with the uncertainties, the knowledge of trends and correlations between variables is needed to base the choices on more objective criteria (OECD 2009). In the agricultural sector, the potential impacts on crops that could be filled by an agrometeorological approach should be considered in the hazard analysis (Rijks and Baradas 2000; Das 2005; Pontrandolfi 2019). Although it is a very complex issue, the inclusion of socio-economics indices, e.g. through composite indices, could improve the comprehensiveness of the assessment (Navarro et al 2023), even if these aspects have been not yet fully dealt with in agriculture (Villani et al. 2022; Ronco et al. 2017; Meza et al. 2020; Hagenlocher et al. 2019).

This overview could also explain the unsatisfactory implementation of risk management policy for agriculture in the Campania administrative region (in Italy, the European and national policies are implemented at a regional level). At the end of 2019, the farms using insurance policies amounted to 2585, the 2.2% of the total number of farms in the region, corresponding to the 2.7% of the value added of the Campania agriculture.

Following all the above considerations, the Campania Region launched the programme “A tools risk management demand analysis for Campania Region” (regional law n. 218/2021). The overall objective is to design the best strategies to increase the resilience of Campania farms in relation to market crises and/or production/quality losses caused by weather extreme events. Among the specific objectives, there is to produce and promote the rapid communication of “certified” information on weather events, yields and prices, and their related trends, in order to better outline the farms’ risk profiles and to promote the dissemination of diversified protection tools. In particular, the definition of the prevailing weather-related hazards at a territorial level is foreseen (Specific Ob. n. 2), giving a reliable knowledge on events.

A growing interest on weather extremes analyses has emerged in the research community during the last two decades (Nicholls and Alexander 2007; Zwiers et al. 2013; WMO and GWP 2016; Seneviratne et al. 2021), leading also to the works of the Expert Team on Climate Change Detection and Indices — ETCCDI, that suggested 27 indices (Klein Tank et al. 2009). Weather-related hazard analyses focused on agriculture are quite spread in scientific literature (WMO and GWP 2016; Shah et al. 2021; Bacci 2017; Antle et al. 2015; Sivakumar et al. 2005; Groot et al. 2018), particularly on drought and floods and/or on specific crops, also in Italy (Moonen et al. 2002). However, other extremes need to be deeply investigated in relation to their potential impacts, such as extreme cold and hot temperatures, as well as heat waves, during sensitive crop stages (Monteleone et al. 2023), also considering future scenarios of potential alterations of bioclimate and trend acceleration towards warmer and dryer conditions, as for example reported for some Mediterranean agricultural areas (coastal countries of the Adriatic Sea) (Charalampopoulos et al. 2023).

In this research field, the trend analyses are gaining ever greater weight in scientific production (De Natale et al. 2023), providing invaluable guidance for researchers and policymakers (Pathak et al. 2018).

Despite the wide scientific literature on this field, the novelty of this work is the contribution to risk assessment building by emphasizing the role of hazard analyses to support policy choices at regional level.

2 Methods

2.1 Study area

The study area is the Campania region (Fig. 1), in Southern Italy, located between the Tyrrhenian Sea to the southwest and the southern Apennines to the northeast, which covers a total area of 13,671 km2 and is divided in five provinces (third level of the European Nomenclature of Territorial Units for Statistics — NUTS): Avellino, Benevento, Caserta, Naples and Salerno. Campania is the most populous and most densely populated administrative region of the South of Italy, third at national level in terms of number of inhabitants and second in terms of population density, with a very unbalanced distribution, since highly populated provinces (Naples, Caserta and Salerno), facing the sea, contrast with other sparsely inhabited or even depopulated ones (Avellino and Benevento), as reported by Minolfi et al. (2016).

Fig. 1
figure 1

Study areas and coverage of Era5-Land meteorological gridded data (see paragraph 2.2) in the five provinces: Caserta, Benevento, Napoli (Naples), Salerno and Avellino

Full size image

The regional orography is very heterogeneous: more than half of the total area is hilly (50.8%); 34.6% is mountainous (Southern Apennines); and only 14.6% is occupied by flat areas. The great part of the regional river basins are tributaries of the Volturno river, the longest one in Southern Italy, with a river basin of 5550 km2, located in the northern part of the region; the second main river is the Sele, with a basin area of 3200 km2 in the southern part of the region. The coastal zones are represented by alluvial plains and high rocky coasts, mainly located in the South (in Salerno province).

In general, the region is characterized by good water availability; withdrawals for the agricultural sector amount to 40%, and the area equipped for irrigation is equal to 24% of the utilized agricultural area (UAA, Eurostat 2019).

Climatic features slightly vary in the 5 provinces, from the coast to the mountain areas (Brunetti et al. 2004). In Fig. 2, a summary of the main climatic variables is reported, derived from data processing described in the following paragraph. Total precipitation and temperatures (minimum and maximum) are higher in the coastal provinces (Caserta, Salerno and Naples). As characteristic of Mediterranean climate, rainfalls are mainly concentrated in the autumn/winter seasons, while the growing period is generally drier. The highest values of reference evapotranspiration are reached in the summer quarter.

Fig. 2
figure 2

Source: data of Copernicus C3S elaborated by CREA (see paragraph 2.2)

Climatic characterization of the Campania provinces (1981–2010 monthly averages). Spatial aggregation was based on the mean value.

Full size image

If not differently specified, the data reported below are taken from a recent report on the agricultural sector in Campania (CREA 2023).

Agriculture is an important productive sector in the region, and according to the latest available data, the agricultural productivity, expressed in terms of agricultural value added at basic prices per unit of work, increased by 9% in the last years. The UAA covers 37.7% of the total regional area, slightly lower than the national value (41.5%). The ratio between agricultural and total employees is 4.2%, slightly higher than its national value (4.1%). In 2020, the farms in the region were 79,353, corresponding to the 7% of the Italian farms, and their size was mostly small, as around 71% had an area of less than 5 ha.

About half of the UAA is occupied by arable crops, of which the most frequent are cereals, mainly durum wheat and alternate forages. The remaining part is almost equally divided between permanent meadows/pastures and woody crops, of which the olive tree is the most widespread. The specialized livestock farms were 13,353 in 2020 (17% of the total number). Referring to woody crops, Campania holds the national record in the production of cherries, walnuts and hazelnuts, while the production of figs, apricots, peaches, plums and citrus fruits is also abundant.

More than 30 agricultural products are recognized by the European Union as Protected Designation of Origin (PDO) or Protected Geographical Indication (PGI, Regione Campania 2023). With reference to the cultivation of wine grapes, the region boasts 19 PDO and 10 PGI.

On a world scale, Campania is a net exporter, with over 3 Mt of exported products in 2020, and a total value of exports of 3896 million euros (growing over the years), especially towards other European countries (67%), America and Asia.

2.2 Dataset

Among several meteorological gridded datasets covering Italy, some products freely provided by the Copernicus Climate Change Service (C3S) (https://cds.climate.copernicus.eu/#!/home) are characterized by a set of variables and a spatial and temporal resolution suitable for agro-meteorological analyses. In particular, the ERA5-Land (E5L) gridded weather dataset (Muñoz-Sabater et al. 2021) was used to calculate the indicators because it is a high-resolution reanalysis product derived from the ERA5 dataset and optimized for the land surface and has a quite large temporal coverage suitable for trend analyses. Reanalysis combines past short-range weather forecasts with observations through data assimilation. The E5L dataset covers the period from 1950 to present on a regular grid with a spatial resolution of 0.1°, corresponding to a horizontal resolution of approximately 9 km for Italy. The raw hourly gridded meteorological data for the bounding box covering Italy (34.875–48.125 N, 4.875–20.125 E) was extracted from the E5L dataset through the Climate Data Store API client (a Python based library) for the following variables: near-surface air temperature, dew point temperature, precipitation, surface solar radiation downwards (shortwave radiation), both wind speed components and geopotential. The grid cells selected to cover the study area are in total 139.

E5L data was summarized to the daily series of the main agro-meteorological variables, according to the procedures explained in Parisse et al. (2023); the relative humidity was derived from hourly humidity time series derived in turn from dew point and air temperature to reduce bias issues in computing air humidity and in the derived estimate of reference evapotranspiration (Frei 2022), which was obtained according to the FAO Penman–Monteith method (Allen et al. 1998). The 1981–2010 monthly climate normals were derived from the daily series as the averages of climatic variables calculated over a 30-year period, according to the World Meteorological Organization — WMO (WMO 2018), and they were spatially aggregated for each province basing on their mean values at NUTS3 level (Table 1).

Table 1 SPEI classification system (modified from WMO-No. 1090, 2012)
Full size table

Data processing was performed using the open source software R (R Core Team 2018) also applying the libraries “climdex.pcic”, “SPEI” and “trend”.

2.3 Weather extreme indices

The analysis of weather extremes potentially affecting agriculture was carried out for the period 1981–2021, through the calculation of a set of indices, some of them linked to specific phenological stages more sensitive to weather stresses (European Environment Agency 2020; Jägermeyr et al. 2021; Maggiore et al. 2020). Moreover, percentile indices were also selected because they are less conditioned by outliers (Morak et al 2013) and more suitable for comparisons of changes across different regions (Alexander et al. 2006). The indices were calculated for each grid cell, and the percentile thresholds were derived from the 1981–2010 climatological period. The following indices were selected:

  • Precipitation due to very wet days: two yearly indices of heavy rain (Zhang et al. 2011) computed as the sum (R95pTOT in mm) of daily precipitation above the 95th percentile of precipitation at wet days (RR > 1 mm) and as the percentage fraction of R95pTOT in relation to the annual precipitation (R95pFRAC).

  • Extreme maximum temperatures: monthly percentage frequency of the days with maximum temperature exceeding the 90th percentile (TX90p) calculated on the distribution of the daily climate values, according to the definition of the ETCCDI (Klein Tank et al. 2009).

  • Drought indices: the first index is the Standardized Precipitation Evapotranspiration Index (SPEI) (Vicente‐Serrano et al. 2010), which is a dimensionless index, calculated for each month by comparing the Climate Water Balance (difference between precipitation and evapotranspiration) cumulated values of the previous n months with the corresponding values of a reference period (as long as possible). In this study, n was set to 3 and 6, which are the time scales more appropriate for agricultural analyses, and the reference period adopted was 1980–2021, including the 1980 data in order to compute the index at a 6-month scale from March to May 1981 The standard 7 classes of SPEI values, as defined in the WMO and GWP report (WMO and GWP 2016), are reported in Table 1. The SPEI is one of the two indices selected (together with the Palmer Drought Severity Index) by the IPCC for agricultural and ecological drought assessment in its Sixth assessment report (AR6) (Seneviratne et al. 2021), and it is receiving growing interest, particularly in large scale studies (De Natale et al. 2023; Ekundayo et al. 2022; Wang et al. 2022; Yildirim et al. 2022; Ahmed et al 2023). In addition, a second metric has been computed, the index of drought months, defined as the number of months when the SPEI value at a 6-month scale is lower than − 1.5, which corresponds to severe/extreme drought (to identify the most critical periods), assessed during the growing season (March–October). The same analysis was performed for SPEI at a 3-month scale.

  • Late frost: count of days with minimum temperature below 0 °C during April and March (LFd), when most tree crops are in the flowering stages (Zheng et al. 2015; Gobin 2018). Due to the progressive advance of plant phenological development, as a consequence of the temperature increasing (Mosedale et al 2015), the analysis was also extended to March, despite most studies in literature, related to the same geographic context, focused on the month of April (Espín Sánchez 2022). In particular, the most critical period for vegetation starts already in the third dekad of March (Zinoni et al. 2002).

  • Heat stress: number of days with maximum temperature above 35 °C (HSTd), which is generally considered as a critical temperature, beyond which the physiological processes can be interrupted (Jones and Goodrich 2008; Bois et al. 2014), calculated for the summer quarter (June–August).

Overall, a widespread use of these indices can be found in the scientific literature referred to risk assessment and adaptation policies support (European Environment Agency 2017; Donat et al. 2013; Russo et al. 2014; Zhang et al. 2011; Zolina et al. 2009). Concerning the agricultural sector, a general increase of this research field has been observed in the last decades (European Environment Agency 2017; Cogato et al. 2019), with most studies referring to drought indices (De Natale et al. 2023; Sisodia and Sharma 2022), also for Italy (Parisse et al. 2020; Vergni et al. 2021).

2.4 Trend analysis

With the aim to detect possible changes in the occurrence and in the intensity of the weather extremes, an analysis of trends and their statistical significance was performed of the above cited indices for each grid cell of the study area, across the period 1981–2021. The analysis was based on two widely used tests, the non-parametric Mann–Kendall (MK — Mann 1945; Kendall 1975) and Sen’s slope (Sen 1968) tests and on a modified version of the MK test suitable for data with periodicities, which is called seasonal Kendall (SK) test (Hirsch et al. 1982).

The rank-based MK test is widely used for climate data to identify trends. The MK test has the advantage of not requiring normally distributed data and of being low sensitive to abrupt breaks in the time series and not affected by outliers (Yadav et al. 2014), but it is not suited for data with periodicities (seasonality) in the time series. In case of seasonality, the SK test estimates the τ statistic separately for each month of the year. The MK and SK tau (τ) values range from − 1 to 1 corresponding to a perfectly downward or upward trend respectively, while a value of 0 indicates the absence of trend. In general, the higher the absolute value of τ, the higher its significance.

The MK and Sen’s slope tests were applied to the R95pTOT and drought months’ time series to determine trends on an annual scale. For TX90p, SPEI, LFd and HSTd indices, the SK test was applied, considering the different months of the year (trend at a monthly scale).

The MK and SK τ statistics were applied to estimate the presence of a monotonic trend in the series. The τ critical values for the different significance levels depend on the length of the time series, that in this study is 41 years for all indices but the SPEI6 and SPEI3, for which it is reduced to 40 years, because the index computation requires to consider 5 or 2 months (respectively) before the month analysed, for completing their aggregation period. For both the MK and SK tests, the maps of distribution of the significant trends were produced, for significance levels of 0.1, 0.05 and 0.01.

The Sen’s slope test was used to estimate the magnitude of the trend over time. This test is commonly used when trends are assumed to be linear. Its value indicates the steepness of the trend and represents the quantification of the change over time.

3 Results and discussion

3.1 Precipitation due to very wet days

During the 1981–2021 period, the Campania region was particularly affected by very heavy rains, i.e. over the 95th percentile of the 1981–2010 reference period (R95pTOT), which averagely amounted to 21% (R95pFRAC) of the annual total precipitation. As shown in Fig. 3, this fraction exceeds or approaches 30% in 4 years (1985, 2003, 2012 and 2021). In 2021, the Campania value was the maximum compared to the other Italian regions, reaching 36% (487 mm) and 33% (446 mm) in the Naples and Caserta areas, respectively (Alilla et al. 2023). The worst conditions were observed in Caserta where the R95pFRAC overcame the value of 30% for 9 times across the analysed period; 5 out of them happened since 2017. In absolute terms, the highest amount of very heavy rains (605 mm compared to a total precipitation of 2034 mm) was reached by Salerno in 2010 which was the rainiest year in the time series for the whole region.

Fig. 3
figure 3

Annual values of R95pTOT (green bars) and R95pFRAC (blue solid lines) over the period 1981–2021 at NUTS 3 level. The green and blue dashed lines indicate the mean values of R95pTOT and R95pFRAC, respectively (spatial aggregation was based on the median value)

Full size image

As shown in Fig. 4a, a positive trend has been estimated for precipitation due to very wet days in the whole region. The trend is significant at least at 0.10 significance level in the great part of the region and exceeds 0.05 in many areas, particularly in the inner ones. In some sparse locations, the significance reaches a level of 0.01. The most affected province is Benevento. In terms of magnitude, the highest slope characterizes Caserta and exceeds 3 mm/year in most of its territory (Fig. 4b). Significant high slope values (over 4 mm/year) are also present in the northern and southern coastal areas of Salerno, likely due to the close mountainous ridges.

Fig. 4
figure 4

Trend map based on Mann–Kendall tau (a) and Sen’s slope (b) values (mm/year) for R95pTOT across the period 1981–2021. Significant positive ( +) and negative ( −) trends in a are reported for significance levels of 0.10, 0.05 and 0.01. Black crosses in b indicate significant trends at a 0.05 significance level

Full size image

The results suggest that heavy rain is no longer exceptional in a trend perspective, confirming the common perception of this phenomenon. This fact makes these kinds of events less suitable to be addressed through risk management tools because of the inherent definition of risk as out of norm and unpredictable. These findings are consistent with other studies produced in Italy (Brunetti et al. 2004). As reported for the inner areas (Iannuccilli et al. 2021), there is a growing trend in the annual, spring and summer heavy rain amount, which will increase the potential damages. Moreover, most extreme rainfall events in winter and autumn are expected to increase in the Italian peninsula, mainly in the Tyrrhenian coastal areas (Faggian 2021), and future scenarios of increasing precipitation extremes were predicted for the Campania region, despite a general reduction of annual rainfalls (Giugliano et al. 2022).

3.2 Extreme maximum temperatures

Extreme maximum temperatures (TX90p), i.e. over the 90th percentile of the 1981–2010 reference period, were quite frequent and widespread in the region between 1981 and 2021. As shown in Fig. 5, the mean value is around 10% (of total days) everywhere in all the months of the year, even though the November values are generally 2% lower and those of September about 2% higher. It is remarkable that the percentages of extreme maximum temperatures exceeded or nearly reached 50% during the June–September period in almost all the provinces, in particular in August, when values above 55% were recorded in Naples, Caserta and Benevento.

Fig. 5
figure 5

Distribution of TX90p monthly percentage values in the 1981–2021 period at NUTS 3 level (spatial aggregation was based on the median value). Red points indicate the mean values

Full size image

Significant positive trends were detected for the TX90p index only in some months of the year while any negative trends were retrieved (Fig. 6). It is noticeable that the summer quarter is mostly affected by this phenomenon, with the highest significance in August. Significant positive trends were also highlighted in February, when some crops (like hazelnut) are already in reproductive stages and in March (in the coastal areas of Salerno) and April (mainly in Benevento) which is the typical period of flowering for many fruit trees in the study area (Chmielewski and Rötzer 2001). In addition, widespread positive trends with the highest significance are evident in November, which can be related to a possible deficit of chilling accumulation for fruit trees (Rodríguez et al. 2021; Luedeling et al. 2011; El Yaacoubi et al. 2020). Some geographical differences are evident in June and July, with part of the northern and the southern region, respectively, not interested by the trends. These findings agree with Andrade et al. (2012), who reported a significant positive trend for warm days throughout Europe. A general upward tendency in terms of frequency and intensity of the hot extremes was also observed in Spain by El Kenawy et al. (2011), particularly evident for the warm days (TX90p), equal to 2.7 days per decade.

Fig. 6
figure 6

Monthly trend maps based on Seasonal Kendall’s tau for TX90p across the period 1981–2021. Significant positive ( +) and negative ( −) trends are reported for significance levels of 0.10, 0.05 and 0.01

Full size image

3.3 Heat stress

As reported in Fig. 7, the number of heat stress days (maximum temperature over 35 °C) during the summer period shows a significant positive trend, which affect a great part of the inner areas, mainly in August and partially in June, whereas in July no significant trend was detected, except for few grid cells. This phenomenon is much stronger in the inner provinces of Avellino and Benevento, which are not affected by the mitigating effect of sea breeze on higher temperatures. In fact, for these provinces, a mean number of heat stress days equal to or greater than 1.2 was detected, compared to a value of 0.3 for Naples and Salerno and 0.6 for Caserta (Fig. 8). The maximum values were reached in Avellino and Benevento areas, where 10 HST days were counted during the summer quarter 2021, while 8 HST days resulted to be concentrated in August 2017. It is remarkable that in the same year a value of 5 or 6 days was reached in Caserta.

Fig. 7
figure 7

Monthly trend maps based on Seasonal Kendall’s tau for nHST across the period 1981–2021. Significant positive ( +) and negative (–) trends are reported for significance levels of 0.10, 0.05 and 0.01

Full size image
Fig. 8
figure 8

Average monthly and summer values of mean (in orange) and maximum (in red) HSTd over the 1981–2021 period at NUTS 3 level during the summer quarter (spatial aggregation was based on the median value)

Full size image

Overall, a general trend towards more extreme conditions is particularly evident for summer extreme temperatures (TX90p and heat stress days). The results confirm the general upward tendency of warm days (TX90) at a global level reported by Alexander et al. (2006) and Morak et al. (2013), and such a trend was also observed in neighbouring areas (Basilicata and Calabria regions), especially in summer days (Piccarreta et al 2015; Caloiero et al 2017). It is expected that this phenomenon will mainly affect crops whose growing cycle takes place during summer, such as tomato, grapevine and olive tree (Papadaskalopoulou et al. 2020).

3.4 Drought indices

The trends of drought conditions in the 1981–2021 period based on the SPEI index computed at 6-month and 3-month scales are shown in Fig. 9. No significant trends in the index value were identified for most of the growing season, except for March, April and October for SPEI6 (Fig. 9a), and March for SPEI3 (Fig. 9b) when an upward trend was detected, suggesting a progressive shift towards wetter conditions. For SPEI6, the trends are concentrated mainly in the northern part of the region in the months of March and April, reaching maximum significance in Caserta, while in October they are located close to the border between the inner provinces of Avellino and Benevento. The most widespread SPEI6 trend was detected in February. As for SPEI3, a less significant trend is shown which covers the whole areas of Caserta and Naples, and the coastal zones of Salerno, but it is remarkable that a local negative trend (towards drier conditions) appears in the coastal belt of Caserta and Naples in August.

Fig. 9
figure 9

Monthly trend maps based on Seasonal Kendall’s tau for SPEI6 (a) and SPEI3 (b) across the period 1981–2021. Significant positive ( +) and negative ( −) trends are reported for significance levels of 0.10, 0.05 and 0.01

Full size image

Another trend analysis has been focused on the index of drought months, i.e. the number of annual drought months during the growing season (from March to October) limited to severe and extreme drought (Fig. 10). This index shows a growing trend, despite the findings derived from the analysis of the value of SPEI indices, mainly located in the inner (Fig. 10a) and coastal areas (Fig. 10b) at a 6-month and a 3-month scale, respectively.

Fig. 10
figure 10

Mann–Kendall tau values for SPEI6 (a) and SPEI3 (b) annual drought months (severe to extreme drought) from March to October across the period 1982–2021. Tau values > 0.2 in b are significant at a 0.10 significance level

Full size image

These results highlight that the strongest droughts tend to be more frequent and more concentrated during the growing season, causing an increasing impact on crops.

SPEI is a quite complete index, useful to give a comprehensive view of the agricultural drought, as it takes into account the water balance between precipitation and evapotranspiration at a temporal scale suitable for assessing soil moisture deficit and consequent crop stress (Mishra and Singh 2010). Even if SPEI allows to identify the critical agricultural drought events, a wider approach to assess drought risk could be used in calculating more than one index, as suggested by Bachmair and Villani (Bachmair et al 2018; Villani et al. 2022), in order to consider other aspects, i.e. frequency, duration and extent, as for example performed in some recent trend studies about agricultural drought (Hayat and Tayfur 2023; Isfahani et al. 2022; Paulo et al. 2012).

3.5 Late frost

In the period of analysis, late frost events were on average 5 in Avellino and Benevento, 3 in Salerno and Caserta, and close to 0 in Naples (Fig. 11). As expected, the number of late frost days was larger in March than in April; its maximum value was reached in 1987 and was between 16 and 18 in all provinces except Naples, where it was equal to 9. Focusing on the most dangerous frost days in April, the worst year was 1997, when Benevento, Avellino and Salerno were affected by 7, 5 and 4 frost events, respectively. These kinds of events, coupled with the warming trend at the beginning of the growing season, leading to earlier flowering, could heavily impact the production of fruit tree orchards (Cannell and Smith 1986; Zavalloni et al. 2006; Pfleiderer et al. 2019), even though the frost exposure is expected to be reduced with the climate change (Parker et al. 2021), as well as grapevine (Mosedale et al. 2015).

Fig. 11
figure 11

Average March, April and early spring values of mean (in azure) and maximum (in purple) LFD over the 1981–2021 period at NUTS 3 level (spatial aggregation was based on the median value)

Full size image

The analysis of trends did not reveal a clear tendency for this phenomenon (Fig. 12): only in the month of March some scattered and significant negative trends were detected, especially near the coastal areas. Considering these results, the late frost days can be considered “unpredictable”, therefore they fully fall within the definition of “risk”.

Fig. 12
figure 12

Mann–Kendall tau values for LFd across the period 1981–2021

Full size image

4 Conclusions

This study focused on the weather-related hazards potentially affecting agriculture in a Southern Italy area (Campania region), giving several important suggestions for supporting the definition of risk policies aiming to enhance the farms’ resilience to weather-related risk. A special attention has been paid to the main hazards currently managed by risk financial tools: frost, drought and floods.

The analysis highlighted a general intensification of weather extremes in the last decades, with an increase of frequency and intensity of events potentially affecting crops, in line with the common perception of these phenomena. Specifically, the main indications refer to the growing amount of heavy precipitation potentially leading to flood, the increasing number of months with severe/extreme droughts during the vegetative season and the rising occurrence of extreme temperatures across the year. The phenomena interested a great part of the regional territory, with some areas particularly affected by both heavy rain and severe/extreme drought. On the other hand, no significant trends were detected for late frosts.

In terms of policy implications, the results seem to suggest that some extreme weather events can no longer be considered as exceptional at the present time and in a trend perspective, particularly for drought and floods. This fact makes these kinds of events less suitable to be addressed through the risk management tools (e.g. crop insurance) because of the inherent definition of risk as out of norm and “unexpected”. The risk management is a very complicated matter, also involving the vulnerability aspects, but this hazard analysis can still suggest that the current policy scheme, based almost exclusively on the strategy of transferring risks (insurances and more recent mutual funds) could be less and less effective, both for farmers and for the allocation of public resources. A more adequate policy could be designed in order to increase the resilience of farmers to drought, floods and extreme temperatures, with more diversified strategies with a medium-long perspective, such as measures for adaptation to climate change (structures and infrastructures, innovation in technologies, improvement of farms’ management etc.).

The information base provided through the choice of indices more related to agriculture seems more accurate to answer at the complexity of the sector compared to a purely meteorological approach. In addition, the monthly analysis allows for assessing different conditions throughout the agricultural season. The chosen indices seem to balance the descriptive capacity and the need of synthesis for policy purposes, but, of course, further indices could be explored. As a general result, the chosen indices based on percentiles showed to be more effective in detecting exceptional events referring to the local averages, while indices based on fixed thresholds allowed to highlight the most critical events that could be the object of exceptional public aids (ex-post intervention). Good potentialities of the E5L data emerged for the hazard assessment. Spatial aggregation at province level can be suitable for policies’ design purposes, albeit it often smooths the differences among areas.

Moreover, the analysis could be more representative of potential impacts using a crop-specific approach focused on the most sensitive phenological stages of each crop. Therefore, further developments of the study could involve a calibration of indices to make them as crop- and site-specific as possible, through the adoption of more diagnostic thresholds.

At last, the methodological approach and the data set adopted can be replicated in other regions, and at a national level, supporting agricultural and climate policy decisions. Future steps of this research will focus on linking hazards, exposure and vulnerability indices, following the scheme of climate risk assessment.

The study highlights the link between the innovation in risk analysis for agriculture and the reliability of policy choices, fostering the effectiveness of the strategies implemented to improve farmers’ awareness and resilience to weather-related shocks and damages.