Spatio–Temporal Dynamics of Groundwater Levels in the Piedmont Po Plain (NW Italy): Impacts of Climate Change and Land Use

Abstract

This study analyses spatio–temporal trends in groundwater level (GWL) across the Piedmont Po Plain (NW Italy), aiming to assess the impacts of climate change (CC) and human drivers on regional groundwater systems. Data were collected from a network of automated monitoring wells over the period 2010–2022, supported by meteorological records from regional weather stations. Results indicate a widespread decline in GWL, with an average of −4.32 cm/y and a maximum of −16.74 cm/y in the time period observed, particularly in recent years. These trends align with decreasing precipitation patterns observed in the region. However, deviations from this general behaviour are also identified. More specifically, local land use practices—specifically rice field and irrigation—may be artificially maintaining GWL. Moreover, hydrometric level data from the main rivers of the region were analysed to evaluate potential interactions with GWL variations. This comparison showed that, in general, no clear correspondence exists between river level fluctuations and groundwater dynamics, except in cases where monitoring wells are located within 50 m of a river channel. In addition, this study was compared with a previous study on the same area concerning temperature variations in groundwater, which allowed for an understanding of both the qualitative and quantitative impacts of CC on the groundwater in the area. The combined analysis highlights the interplay between CC and anthropogenic influences, emphasising the need for integrated groundwater management strategies that account for both climate variability and land use dynamics. Furthermore, the seasonal analysis of GWL revealed a clear hydrological cycle shaped by irrigation activity. In particular, the occurrence of GWL peaks during summer (irrigation season) confirms the role of irrigation in controlling aquifer behaviour in agricultural areas. The absence of a general correlation with river stage, combined with the occurrence of GWL peaks during summer (irrigation season), confirms that irrigation is the main driver of GWL fluctuations over the study period. This finding is critical for the accurate interpretation of groundwater responses and for developing effective and sustainable water management strategies in intensively cultivated regions.

1. Introduction

Climate change (CC) is a global phenomenon that exerts substantial influence on environmental components, including groundwater (GW) resources [1,2,3]. As a key component of the hydrological cycle, GW is highly sensitive to shifts in precipitation (P) patterns, temperature regimes, evapotranspiration rates, and land use changes [4,5,6]. These climate-induced variations affect recharge rates, storage capacity, and discharge processes, often leading to fluctuations in GW levels (GWLs) [7,8,9,10,11,12,13,14,15,16,17].

In particular, altered P patterns—characterised by changes in intensity, frequency, and spatial distribution—can either enhance or reduce GW recharge, while rising air temperatures typically increase evapotranspiration, exacerbating water stress [1,2,5]. Additionally, the anthropogenic footprint, especially through irrigation, urbanisation, and deforestation, further modifies GW recharge dynamics and amplifies vulnerability to climate extremes [1,18,19,20,21,22]. Understanding these complex interactions is essential for sustainable GW management, particularly in light of growing water demands and increasing climate variability. Developing integrated adaptation strategies that consider both natural and human factors is critical for ensuring long-term groundwater resilience and water security [23,24,25,26].

However, despite evidence of the direct impact of CC on GWLs, it is only rarely possible to have up-to-date studies at a regional level that provide an overview of the quantitative status of GW resources. This is particularly true for the Piedmont Po Plain, where a detailed understanding of GWLs responses to both climatic and anthropogenic pressures is crucial for effective water resource management.

This study aims to focus on the current quantitative state of GW in the Piedmont Po Plain (NW-Italy) to provide a comprehensive analysis of the impacts of CC and human drivers on GWLs in the period between 2010 and 2022. Such analysis is crucial given the known global critical role GW plays in sustaining agricultural, industrial, and domestic demands within this densely populated and economically significant region [27,28,29,30,31].

The Piedmont Po Plain, renowned for its diverse geological and hydrogeological characteristics, plays a pivotal role in sustaining agriculture, and GW represents the main water resource, for agricultural uses, industrial purposes and for drinking water supply [32,33]. For this reason, in the recent past, there has been a focus on studying the quantitative aspects of the resource by analysing GWL [16,32,34,35], often with a particular focus on recharge mechanisms [36,37].

The intensification of extreme weather events, such as more frequent and severe droughts in the Piedmont Region, underscores the vulnerability of the Po Plain to CC [38]. Furthermore, increased water demand for irrigation, coupled with altered precipitation regimes, has led to a significant increase in GW abstraction, further stressing these vital reserves [39,40,41,42]. This situation was particularly evident during the severe drought events documented in 2017 and 2022, which strongly affected northern Italy [38,40,42]. In this context, the present study also contributes to understanding the quantitative response of shallow aquifers to such extreme periods, providing insights into the resilience and vulnerability of GW resources under prolonged climatic stress.

Consequently, the objective of this paper is to observe how the quantitative status of GW in the Piedmont Po Plain has changed from 2010 to 2022 in relation to extreme events caused by CC [43]. More specifically, this study aims to analyse the decadal variations in GWL across 15 monitoring wells uniformly distributed throughout the Piedmont Po Plain. Secondarily, an analysis of GW variation due to other human drivers (irrigation and land use) was also conducted in order to complete the overall picture and obtain greater clarity regarding groundwater dynamics. This comprehensive spatiotemporal assessment will integrate high-resolution climate data with detailed land use information to identify the primary drivers of GWL fluctuations in this critical region. This integrated approach will allow for an understanding of how climatic variability, rivers and direct human interventions collectively shape the region’s hydrogeological response, providing critical insights for future water resource planning and climate adaptation strategies. Such an exhaustive analysis is paramount for developing robust policies that ensure the long-term sustainability of GW resources in the face of ongoing environmental transformations and escalating anthropogenic pressures.

2. Study Area

The study area is the Piedmont Po Plain in northwest Italy, which hosts a critical and extensive GW resource for the Piedmont Region. The Piedmont Po plain represents the westernmost part of the larger Po Plain, one of Europe’s most significant lowland areas. Covering approximately 27% of Piedmont’s territory [32], this vast plain is crossed by rivers and characterised by fertile alluvial soils deposited by the Po River and its tributaries over millennia. The plain is a crucial agricultural and economic hub, supporting the cultivation of crops such as rice, corn, and fruit [34].

2.1. Climatic Setting

The climatic setting of the Piedmont Po Plain is shaped by the interaction of continental, Mediterranean, and Atlantic influences, resulting in a temperate climate with well-defined seasonal variability. Historical meteorological data (1958–2009) indicate a mean annual air temperature (AT) of approximately 12.5 °C in lowland areas, with average summer and winter temperatures of 22 °C and 3.3 °C, respectively [44].

AT varies seasonally and spatially, typically ranging from around 0 °C (in winter) to 25 °C (in summer). Since the mid-1980s, a more pronounced rise in average temperature has been observed, particularly during the winter, spring, and summer months [37] (Figure 1). P patterns also reflect this climatic complexity [38,45]. Annual P between 2010 and 2022 ranged from approximately 400 to 1400 mm, with higher values generally recorded in the western and northern sectors of the plain due to orographic effects [44]. On average, annual P from 2002 to 2017 exceeded 900 mm, exhibiting a bimodal distribution typical of a continental (prealpine or subalpine) climate, with two maxima in spring (March–April) and autumn (September–October), and two minima in winter (January) and summer (June) [44].

To better illustrate recent trends in AT and P within the study area, Figure 2 reports the annual average values recorded at the Torino–Giardini Reali meteorological station [46], located in the provincial capital of the study region. The graph shows a clear increasing trend in AT and a decreasing trend in P between 2010 and 2022. Notably, 2017 and 2022 stand out for remarkably low P (in 2017 around 500 mm/y and in 2022 around 300 mm/y), representing a local extreme that may have significantly affected surface and GW dynamics [44]. These deviations confirm the increasing frequency of hydro–climatic extremes.

More broadly, recent decades have shown clear evidence of CC, which is altering regional climatic regimes [44]. These shifts include increasing AT, a higher frequency of extreme events, such as heatwaves, and significant changes in P patterns [37,38,47]. These factors, in turn, are impacting the hydrological cycle, particularly groundwater recharge and availability [8,16,47,48].

The effects of CC are particularly evident in recent years. Notably, the year 2022 was recorded as the hottest and driest in the past 65 years, with the highest temperature anomaly and the most pronounced P deficit in the entire historical series [38,49].

2.2. Hydrogeological Setting

From a hydrogeological point of view, from top to bottom, the Piedmont Po plain consists of the following (Figure 3) [32,50]:

-

Quaternary alluvial deposits (lower Pleistocene–Holocene), that host a shallow unconfined aquifer. Formed by fluvial deposits due to the erosion and sedimentation processes of rivers, including the Po and its tributaries, and by fluvioglacial deposits, this complex consists mainly of gravel and sand. Alluvial deposits are prevalent in the low-lying areas of the plain and contribute to its fertile soils;

-

Villafranchiano transitional deposits (late Pliocen–early Pleistocene), which host a multi-layered aquifer. They are predominantly lacustrine, swamp and fluvial sediments and consist of both permeable deposits (pebble, gravel, sand) and low-permeability deposits (silt and clay);

-

Pliocenic marine deposits (Pliocene), hosting a confined aquifer. Predominantly found in the southern and western margins of the plain, marine deposits originate from ancient seas that once covered the area. These deposits comprise layers of sand, silt, and clay, reflecting past marine environments. Marine deposits may contain valuable fossil records.

For its features and its size, the Piedmont Po plain is considered the most important GW reservoir of the Piedmont region [32,52].

The analyses in this study were conducted through the observation of the shallow unconfined aquifer, hosted in the Quaternary alluvial deposits. This aquifer consists of coarse gravels and sands of fluvial or fluvioglacial origin, with minor intercalations of silty clay deposits, and a thickness between 20 and 50 m [53]. At the base of the shallow aquifer, locally thick and continuous layers of silt or clay-rich deposits are present, which represent the separation between the shallow aquifer and the deep aquifers [54].

The decision to study the shallow aquifer in detail is twofold. The first is linked to the susceptibility of this aquifer to CC since, being the most superficial, it is the most exposed to the climatic variations. The second motivation is linked to the quantity of data relating to this aquifer from the main regional agencies (Arpa Piemonte and Regione Piemonte) and the ease with which it is possible to carry out measurements.

3. Materials and Methods

Fifteen monitoring wells (all capturing the shallow aquifer) were selected across the Piedmont Po Plain to better investigate variations in the phreatic GWL throughout the study area (Figure 4).

The selection of these wells was based on the availability of accurate and continuous GWL data over the entire period of analysis, but also because these same monitoring points were previously used to assess groundwater temperature (GWT) variations for the same time interval, as reported in [51]. All selected wells are part of the regional monitoring network and were chosen to ensure a spatially uniform coverage of the study area. GWL data are freely available and consist of two automatic measurements per day, recorded at 00:00 and 12:00 [55]. For all analyses, these raw data were first averaged daily and subsequently aggregated to monthly means; the resulting monthly average values were then used for statistical analyses.

Each well in the study area is instrumented with an OTT Orpheus Mini (OTT Hydromet GmbH, Loveland, CO, USA), an integrated pressure sensor and datalogger for level measurement in GW.

The Mann–Kendall test and the Thei–Sen estimator were selected and applied as the primary methods for trend analysis in line with the most commonly adopted approaches in the scientific literature for this type of study [56,57,58].

Before doing a statistical analysis, a completeness index (CI) analysis [59] was carried out for each monitor point, according to the following equation:

$$CI={\displaystyle \frac{number of available GWL data}{maximum number of data in the considered interval}}\times 100$$

(1)

GWL variations were analysed using the Mann–Kendall trend test method [60,61] with the PAST programme [62]. With this programme, a trend non-parametric test is employed, requiring a single column of data. Missing values are excluded, and the number of observations in the dataset (‘n’) is adjusted accordingly, following Gilbert’s procedure [63]. The data considered, denoted as x₁, x₂,…, x_n, represent, in this specific case, the sequence of GWL measurements ordered in time.

An indicator function is defined as follows:

$$sgn \left(X\right)= \left\{\begin{array}{c}1 if X>0\\ 0 if X=0\\ -1 if X<1\end{array}\right.$$

(2)

where X corresponds to the difference between two data values (x_j − x_i).

The Mann–Kendall S statistic is computed by summing over all pairs of values as follows:

$$S= {\sum }_{i=1}^{n-1} {\sum }_{j=i+1}^n sgn \left(X_j- X_i\right)$$

(3)

A negative S indicates a decreasing trend, S = 0 indicates no trend, and a positive S indicates an increasing trend.

For small samples (n ≤ 0), the p-value is derived from an exact values table [63]; for larger samples (n > 10), a normal approximation is applied.

When identical values occur in the dataset, they form tied groups. Each tied group j contains t_j identical values, and the total number of these groups is denoted as g. These tied values must be accounted for in the variance of S. Then, the standard deviation (SD) of S is estimated using the following formula:

$$SD=\sqrt{{\displaystyle \frac{1}{18}}}[n\left(n-1\right)\left(2n+5\right)-\sum _{j=i}^g{t}_j(t_j-1)({2t}_j+5)]$$

(4)

Next, the Z statistic is computed using the following formula:

$$Z={\displaystyle \frac{\left|S\right|-1}{SD}}sgn S$$

(5)

This Z statistic is utilised to determine the p-value from the cumulative normal distribution with the customary continuity correction of subtracting 1.

Once the Mann–Kendall test confirmed the presence or absence of a statistically significant monotonic trend, the Theil–Sen estimator [64,65] was applied to quantify its magnitude. The Theil–Sen method estimates the median slope of the trend, which is less sensitive to outliers than parametric approaches. This method makes it possible to quantify the total change in the parameter analysed (in this study GWL and AT), providing a reliable estimate of the trend.

The slope is calculated as the median of the slopes determined by all possible pairs of points (X_i, Y_i) and (X_j, Y_j) in the time series, where j > i. The slope for each pair is defined by the following formula:

$${Slope}_{ij}={\displaystyle \frac{Y_j-Y_i}{X_j-X_i}}, for j>i$$

(6)

Once the slopes for all pairs have been calculated, the Theil–Sen estimator is given by the median of these slopes as follows:

$$Slope Theil-Sen=Median\left({\displaystyle \frac{Y_j-Y_i}{X_j-X_i}}\right)$$

(7)

This median provides a robust estimate of the overall rate of change of the variable over time, reducing the influence of outliers that could distort results in parametric methods.

The final result of the estimator can be interpreted directly as the average change per unit time of the variable under consideration. Positive values of the slope indicate an increase over time, while negative values indicate a decrease.

To better understand the dynamics behind the observed variations in GWL, P and river water levels data were also analysed across the study area. Three representative meteorological stations and three representative river stations, chosen from the regional monitoring network [46], were selected based on their geographic coverage and data availability, as shown in Figure 4. Daily P and water level data from these stations were aggregated into monthly averages, with P data also aggregated into annual averages. As with GWL, the monthly mean values were then used to perform a visual assessment of temporal trends.

4. Results

With regard to the completeness index CI, all points had a good level of completeness (≥89%), as shown in

Table 1. Completeness index CI of the analysed monitoring wells.

Monitoring Well	DST	P8	P14/1	P21	P23	P43	PII06	PII11	PII19	PII31	PII32	PII45	PII51	SI5	T2
CI (%)	100	100	100	96	99	92	100	97	100	94	95	97	97	89	100

. The CI confirmed that all 15 monitoring points had a sufficiently robust dataset for trend analysis, which allowed the analysis to continue on all points using the methods described above.

The results of the Mann–Kendall trend test identified a statistically significant negative trend in GWL in 13 out of the 15 wells examined (

Table 2. Mann–Kendall trend test results. S indicates the Mann–Kendall statistic, Z is the standardised test statistic, and p-value represents the significance of the trend. Out of 15 monitored points, 14 show a statistically significant trend (13 negative and one positive), while only 1 point does not show a statistically significant trend.

	S	Z	p Value	Statistical Trend
DST	−835	−1.278	0.20123	No
P8	−1878	−2.8764	0.0040229	Negative
P14/1	−3919	−6.004	1.92 × 10−9	Negative
P21	−4580	−7.5156	5.67 × 10−14	Negative
P23	−3011	−4.7025	2.57 × 10−6	Negative
P43	−5537	−9.6621	4.37 × 10−22	Negative
PII06	−1283	−1.9646	0.049464	Negative
PII11	−3613	−5.7543	8.70 × 10−9	Negative
PII19	−1658	−2.5392	0.01111	Negative
PII31	−2003	−3.353	0.00079949	Negative
PII32	−3110	−5.1545	2.54 × 10−7	Negative
PII45	−2406	−3.8314	0.0001274	Negative
PII51	2819	4.5339	5.79 × 10−6	Positive
SI5	−2613	−4.7563	1.97 × 10−6	Negative
T2	−3084	−4.7245	2.31 × 10−6	Negative

), indicating a consistent and widespread decline across the study area over the period 2010–2022.

Only two points deviate from the general statistically significant negative trend:

-

DST does not show a statistically significant trend;

-

PII51 shows a positive statistical trend.

Subsequently, for the 14 points showing a statistically significant trend (positive or negative), it was possible to calculate the Theil–Sen estimator, the results of which are shown in

Table 3. Results of the Theil–Sen estimator. N° of data indicates the number of observations used for each point, Theil–Sen trend line slope represents the estimated slope of the groundwater level trend, GWL variation (m/13 years) shows the total change over the 13-year period, GWL variation (m/y) is the average annual change in meters per year, and GWL variation (cm/y) is the same expressed in centimeters per year.

Monitoring Well	N° of Data	Theil–Sen Trend Line Slope	GWL Variation (m/13 Years)	GWL Variation (m/y)	GWL Variation (cm/y)
P8	156	−0.002	−0.312	−0.02	−2.40
P14/1	156	−0.0034	−0.5304	−0.04	−4.08
P21	149	−0.0075	−1.1175	−0.09	−8.60
P23	154	−0.0094	−1.4476	−0.11	−11.14
P43	143	−0.0041	−0.5863	−0.05	−4.51
PII06	156	−0.0006	−0.0936	−0.01	−0.72
PII11	152	−0.0038	−0.5776	−0.04	−4.44
PII19	156	−0.0018	−0.2808	−0.02	−2.16
PII31	147	−0.0026	−0.3822	−0.03	−2.94
PII32	148	−0.0147	−2.1756	−0.17	−16.74
PII45	152	−0.0017	−0.2584	−0.02	−1.99
PII51	151	0.0075	1.1325	0.09	8.71
SI5	139	−0.0037	−0.5143	−0.04	−3.96
T2	156	−0.0046	−0.7176	−0.06	−5.52

. The results show an average lowering of GWL of −4.32 cm/year, with a maximum decline of −16.74 cm/year recorded at well PII32.

As regards the average monthly GWL trends, the results are reported in full for all points in Figure 5 and graphically illustrate what has already been confirmed by the static test results, showing a general decrease in GWL throughout the entire area.

Figure 6 shows both clear seasonal oscillations and long-term variations. Most wells show a regular intra-annual cycle, with higher GWL typically occurring in summer and lower levels during winter. Regarding what concerns interannual dynamics, it is possible to observe that throughout the study period, a general tendency toward declining GWL can be seen in the majority of wells, particularly from around 2015 onward. The downward trend becomes especially evident during the drought years of 2017 and 2022, which correspond to the lowest GWL values recorded in many wells, such as PII32, P14/1, and P21. Moreover, as already stated by the statistical analysis, well PII51 stands out from the regional pattern by displaying a steadily increasing GWL throughout the entire period.

Figure 7 presents a direct comparison between monthly P and GWL for wells PII45 (with Torino weather station), P23 (with Morozzo weather station), and PII32 (with Novara weather station).

In all three cases, despite seasonal oscillations typical of temperate climates, a long-term decrease is evident in both P and GWL. The linear trendlines confirm a general reduction in monthly P totals, which aligns with the observed negative GWL trends detected in the area.

All three weather stations show a current peak of P is typically observed in the spring and autumn months, consistent with the bimodal precipitation regime typical of the Piedmont region [48,66]. Despite the persistence of this seasonal cycle, a general decrease is evident at all locations and appears gradual but consistent, particularly after 2015.

Regarding the analysis of the mean monthly river water level trends at the three selected monitoring stations (Figure 8), the results show a generally regular seasonal pattern across the Sesia, Po, and Tanaro rivers. Over the long-term series (2010–2022), the fluctuations appear consistent with natural hydrological cycles, characterised by higher river water levels in spring and autumn and lower values during summer. This seasonal behaviour, with a clear minimum in July–August, coincides with the peak of the irrigation period.

5. Discussion

The analysis of GWL time series from 15 monitoring wells across the Piedmont Po Plain during the period 2010–2022 reveals important insights into both seasonal and long-term GW dynamics. To better contextualise and interpret these results, land use data of the region were also considered, as reported in Figure 9.

The analysis of the monthly GWL trends (Figure 5 and Figure 6) confirms a coherent and repeated seasonal hydrological cycle across most of the study area. GWL typically reach their annual maxima in summer and their minima in winter. This behaviour clearly indicates the strong influence of irrigation practices on groundwater dynamics. In fact, as shown in Figure 4, the land use of the Piedmont Po plain is dominated by arable lands (shown in yellow in the figure), rice paddies (in orange/red, mostly in the northeastern part of the region), and a smaller portion of fruit trees in the western area (light orange). The dominant portion of arable lands, mainly cultivated with wheat and corn, is characterised by an irrigation period during the summer months, starting from May/June and lasting until July/August, whereas for rice cultivation, irrigation typically starts in May and finishes in August.

Under natural, non-irrigated conditions, GWLs in the Piedmont region would generally peak in spring (caused by snowmelt and seasonal rainfall) and have a minimum in summer/fall (dry season) [16,34]. However, in this study, the observed summer peak, often occurring in August, coincides with the agricultural irrigation season. This highlights how, in lowland agricultural areas, the hydrogeological cycle is heavily shaped by irrigation schedules rather than purely climatic or natural hydrological drivers. This is a crucial factor to consider for sustainable GW resource management, as it underscores the need to account for land use and irrigation practices when assessing aquifer behaviour. This interpretation is further supported by the seasonal analysis comparing P and GWL (see Figure 7), which clearly shows a temporal shift: the P peak occurs in spring, while the GWL peak is delayed to summer (usually August), indicating that the aquifer responds more strongly to irrigation inputs than to natural ones. To better understand the relationship between GWL and P, correlation plots were created using mean monthly data over the entire period (2010–2022) for the points already analysed in Figure 7 (Figure 10). These scatter plots did not show any evident correlation, with very low correlation values (R² < 0.25). This confirms that GWL dynamics cannot be explained by precipitation alone, but are also influenced by other factors, such as irrigation practices.

Moreover, the analysis of river water levels in the study area (see Figure 8) highlights that the summer GWL maximum cannot be attributed to river–aquifer interactions, since river water levels show their annual minimum during the same period.

Beyond the seasonal variability, longer-term trends indicate a widespread and statistically significant decline in GWL across most of the study area (see Figure 5). The decreasing trend becomes especially evident from 2015 onward, with 2017 and 2022 standing out as years of particularly sharp decline, coinciding with documented drought events in northern Italy [38,49,69].

More specifically, for the study are, the statistical tests used showed that 13 of the 15 wells analysed show a statistically significant downward trend. The only two wells that deviate from this trend are the DST well and the PII51 well.

The DST well is located in Masio (in the province of Alessandria), approximately 40 metres from the Tanaro River, and is the only monitoring well that does not have a statistical trend. It also represents, in the study area, the only monitoring point located at a distance of less than 100 m from a river. Given its proximity to the river, to better understand the relationship between the DST well and the Tanaro River, a river monitoring station located in close proximity to the DST well was considered [70]. Monthly average hydrometric level data were used and compared with the monthly average piezometric level data from the DST well for the same time period (January 2010 to December 2022). The resulting scatter plot showed in Figure 11 highlights a R² value of 0.74, indicating a good correlation between piezometric and hydrometric levels, and thus confirming the strong connection between the two parameters. The variability of DST levels is therefore explained by the dynamics of the Tanaro River, which masks or counterbalances possible long-term declining trends of the GWL in this monitoring well.

Well PII51, on the other hand, is the only monitored well showing a statistically significant upwards trend. This is a piezometer located in the province of Novara, which is known for rice fields (shown in Figure 9). This type of agriculture uses special irrigation techniques that are recognised in the literature [34] as being able to influence the hydrogeological regime even on a small scale. Although other monitoring wells in the same area display a statistically significant decreasing trend, this behaviour can be justified by different degrees of subsoil permeability. Although the northern sector is characterised by finer soils that typically require smaller irrigation volumes, the upwards trend at PII51 can be attributed to local water retention in fine-grained soils, which slows drainage and promotes accumulation, as well as possible contributions from nearby canals or small-scale irrigation practices [71]. According to studies conducted in the region, even such a relatively small area is characterised by different degrees of permeability, which strongly affect GW recharge. Moreover, rice cultivation practices can vary, ranging from dry seeding to continuous flooding and these different irrigation methods may influence GWLs even at such a local scale [72,73,74].

For what concern the correlation between GWL and P in time series, this is clearly evident revealing an almost parallel downward trends over the 2010–2022 period, especially during drought years (in particular 2022). The temporal coincidence between low GWL values and prolonged dry periods reinforces the hypothesis that reduced recharge due to altered P regimes is a major driver of groundwater decline in the region over the long term.

This is further supported by a comparison with GWT data from previous research, which used the same set of monitoring wells and time period [51]. While this study demonstrates a widespread reduction in GWL, the previous one on GWT revealed a concurrent GWT warming trend across the Piedmont Po Plain. These findings confirm that CC is exerting a dual impact on regional GW systems: quantitatively through declining GWL and qualitatively through rising GWT. Such changes have implications not only for water availability but also for aquifer thermal regimes, with potential effects on GW-dependent ecosystems. In fact, even small GWT increases of about 1–2 °C, if sustained over time and combined with reduced discharge, may also impact sensitive habitats, aquatic species, and biological processes [75,76,77].

Moreover, PII51 shows a rising trend in GWT stripes [51], consistent with the general warming observed at other monitoring points. This suggests that the local increase in GWL at PII51 is not climate-driven, but more likely influenced by land use practices (paddy fields) which highlights the distinct hydrological behavior of rice irrigation systems compared to other agricultural areas.

These results are consistent with what has been observed at the global scale, where GW systems are undergoing widespread declines, often accelerated in agricultural floodplains. Large-scale analyses involving more than 1600 aquifer systems about one-third of global aquifers show accelerating depletion trends, with declines often exceeding 0.5 m/year [56]. The strongest links are observed in intensively cultivated drylands, where irrigation accounts for around 70% of global GW withdrawals [78]. This highlights that the dynamics observed in the Piedmont Po Plain are not isolated, but form part of a broader global pattern linking irrigation practices, climatic stress, and GW depletion.

When placed in a global context, the Piedmont case mirrors conditions in other agricultural floodplains, such as the Indo-Gangetic Plain [79,80] and the North China Plain [81,82,83].

Conversely, there are also cases of partial recovery. For example, in the Bangkok Basin, groundwater levels rose after the introduction of pumping fees and well licensing [84]; in Iran’s Abbas-e Sharghi Basin, declines were reversed through surface water diversion from the Kharkeh Dam [85]; and in Arizona (USA), managed aquifer recharge (MAR) using Colorado River water helped stabilize GW storage [86]. These success stories in areas also similar to Piedmont Po plain underline that recovery is possible with a correct and sustainable management of water resources, though usually at slower rates than depletion.

6. Conclusions

This study provides a detailed analysis of GWL trends in the Piedmont Po Plain over the period 2010–2022, clarifying how shallow aquifers respond to the combined pressures of climate variability and human activities. The findings reveal a widespread and statistically significant decline in GWL across most monitoring points, with an average decrease of −4.32 cm/year, and peak declines reaching up to −16.74 cm/year. The coincidence between sharp declines and drought years, such as 2017 and 2022, highlights the system’s vulnerability to prolonged dry periods. A key contribution of this work is the integration of climatic variables, land use patterns, and river–aquifer interactions to interpret the observed trends.

The seasonal analysis reveals that irrigation practices, particularly in corn and rice cultivation areas, have reshaped the natural hydrological cycle, producing unusual summer groundwater maxima. This demonstrates how agricultural practices can locally override climatic drivers, creating heterogeneous responses even within short distances, as exemplified by the contrasting behaviour of wells PII51 and PII32.

Moreover, the correlation between drought years (such as 2017 and 2022) and accelerated GWL decline highlights the vulnerability of the aquifer system to drought periods. An exception to this general trend is well PII51, located within a rice cultivation area, which displayed a positive GWL trend likely driven by artificial recharge through traditional irrigation practices. This highlights the importance of considering land use and irrigation methods when interpreting aquifer dynamics.

By linking GWL dynamics with previous research on GWT in the same monitoring network [51], this study also reinforces the evidence of CC impacts on both the quantitative and thermal regimes of GW systems of the shallow aquifer of the Piedmont Po plain. This dual perspective underlines the importance of integrated approaches when assessing GW vulnerability.

Overall, the observed trends (seasonal, interannual, and spatial) underscore the complex interplay between climate variability and human activity in shaping GW dynamics in the Piedmont Po Plain. This paper underscores the urgent need for integrated GW management strategies that account for climatic variability, land use patterns, and water use efficiency. Future research should focus on refining the analysis of rainfall and temperature trends, as well as incorporating land use data and seasonal GW abstraction rates. In addition, future work should aim to expand monitoring networks to include deeper aquifers, such as the Villafranchiano multi-layered aquifer, for which new monitoring wells are currently being constructed. Furthermore, developing comprehensive and publicly accessible datasets on groundwater abstraction in the study area would greatly enhance the understanding of aquifer dynamics and support more effective management strategies. Such efforts are crucial to developing adaptive management policies capable of safeguarding GW resources under ongoing and future CC scenarios.