Extraordinary 21st Century Drought in the Po River Basin (Italy)

Abstract

Recent research identified 2022 as being the year of lowest seasonal April–May–June–July (AMJJ) observed streamflow for the Po River Basin (PRB) in the past two centuries. Expanding upon this research, we applied filters (2-year to 30-year filters) to the AMJJ observed streamflow and identified the late 20th and 21st century as displaying extreme drought. In this study, we introduce PALEO-RECON, an automated reconstruction tool developed to leverage tree ring-based proxies and streamline regression processes. Using PALEO-RECON, we reconstructed the AMJJ streamflow, applying traditional regression techniques and using a nested approach in which 30-, 40-, and 50-year windows within the ~200-year observed streamflow record (1807 to 2022) were evaluated to capture uncertainty. Focusing on the 21st century (2000 to 2022), while several droughts in the ~2000-year paleo record may have exceeded the 2000 to 2022 drought, the recent PRB drought ending in 2022 was the lowest 23-year period in approximately 500 years, acknowledging that uncertainty increases as we move further back in time. When examining paleo and observed AMJJ streamflow records, deficit and surplus periods were evaluated, focusing on the potential “whiplash” between drought and pluvial events. We observed an increase in the frequency of whiplash events, which may be associated with a changing climate.

1. Introduction

The Po River Basin (PRB) (Figure 1) in northern Italy is the largest watershed in Italy, covering ~75,000 km² (~71,000 km² in Italy). The PRB is a complicated watershed that includes: (1) mountainous regions along the western and northern boundaries, which contribute snowmelt-driven runoff in the late spring and early summer months; (2) internal rainfall-dominated river systems that exhibit minimal seasonal flows during the summer months; and (3) the PRB delta, a critical biosystem that covers ~700 km² [1,2,3]. Average precipitation across the watershed is ~1200 mm per year, while the average annual discharge rate from 1807 to 2021 at the Pontelagoscuro gauge (Figure 1) was ~1500 m³/s [4,5]. The importance of the PRB to a wide range of sectors cannot be understated. The PRB hosts ~20 million residents, which represent approximately one-third of Italy’s total population. The PRB is a major supplier of energy, producing over 30% of the total Italian power production from ~900 hydro and ~400 thermal power plants [2]. The PRB accounts for ~35% of the total agricultural produce generated in Italy [6]. Thus, the impact of drought in the PRB would have a wide and severely negative impact on both the Italian economy and populace.

In the spring–summer of 2022, the severe drought in the PRB generated attention, both in Europe and worldwide. Headlines included: (1) “We are in extreme crisis. Italian Parmesan producers fear for future amid drought”; (2) “World War 2 mystery solved as German vehicle rises from river after severe drought”; (3) “Drought in Italy Reveals Sunken World War II Barge”; (4) “In the Po Valley, a historic drought is threatening Italy’s breadbasket”; and (5) “Italy’s Lake Garda shrinks to near-historic low amid drought” [7,8,9,10,11]. To quantify the 2022 drought, Ref. [12] investigated the ~200-year Pontelagoscuro gauge streamflow record [5], focusing on the spring–summer season including the month of July (J) and the periods of June–July (JJ), May–June–July (MJJ), and April–May–June–July (AMJJ). For each of these (the month or periods), Ref. [12] observed that 2022 was the lowest recorded year of streamflow in the ~200-year period of the records.

While [12] identified 2022 as the single worst drought year in over 200 years, we investigated longer-duration drought periods ranging from 2 to 30 years in the ~200-year Pontelagoscuro gauge streamflow record. For example, how does the most recent 10, 20, or 30 years of streamflow in the PRB rank in terms of drought when compared with the ~200-year observed period of record? Was the 2022 drought an anomalous year, or are we witnessing increased drought that may be associated with a changing climate?

The hypotheses of the current research include: (1) the 2022 spring–summer drought represented the end year of a multi-decadal drought in the PRB that was the most severe in the ~200-year observed record; (2) skillful streamflow reconstructions (using tree ring-based proxies) of PRB spring–summer season streamflow can be achieved through the development of PALEO-RECON, a reconstruction tool that modernizes regression-based reconstructions and significantly improves efficiency as an open-source model; (3) the streamflow reconstructions provide a paleo perspective of the recent 21st century (2000–2022) PRB drought, showing it to be the most extreme drought in ~500 years; and (4) whiplash events (sudden transitions between wet (pluvial) and dry (drought) periods) are increasing from the paleo to the observed record and may be associated with a changing climate.

While quantifying the recent 2022 drought using the observed record provides context for water managers and planners, comparing this drought with pre-observed (paleo) records would be beneficial. Initial efforts by [13] were later expanded by [14] in which a regional reconstruction of PRB streamflow was developed. Applying the methodologies of [15,16,17] to PRB regional streamflow using tree ring-based proxies [18], Ref. [14] developed a PRB “basin-wide” regional estimate of paleo streamflow variability. While helpful in identifying past (paleo) drought and pluvial periods, developing a “gauge-specific” (e.g., Pontelagoscuro) reconstruction of streamflow would allow for the quantification of the recent 2022 drought over two centuries. In this study, we introduce PALEO-RECON, version 1.0.1, a novel open-source tool designed to automate the reconstruction process and support reproducible research across different watersheds. PALEO-RECON was developed at The University of Alabama, Tuscaloosa, AL, USA, and is openly accessible as detailed in [19]. Using PALEO-RECON, we developed a skillful AMJJ seasonal reconstruction of PRB (Pontelagoscuro gauge shown in Figure 1) streamflow. This paper is organized in four sections, in addition to this introduction. In Section 2, we present the data, methodology, and the PALEO-RECON software. Section 3 discusses the results, and Section 4 provides an analysis of these findings. Finally, Section 5 draws conclusions based on the study.

2. Materials and Methods

2.1. Observed Streamflow and Self-Calibrated Palmer Drought Severity Index (scPDSI)

Monthly streamflow data was obtained for flows at Pontelagoscuro, which is located near the outlet of the Po River [5,12]. The monthly flow rates (m³/s) were converted to monthly volume (million cubic meters (MCM)), and the data begin in January 1807 and end in August 2022. Previous research efforts [12] examined spring–summer flows by selecting four windows of interest: July (J), June–July (JJ), May–June–July (MJJ), and April–May–June–July (AMJJ). For each of these four windows, Ref. [12] showed that the 2022 drought was unprecedented in the previous two centuries. A further examination of the spring–summer flows was performed in which filters (2-year to 30-year) were applied (Table 1). In addition to confirming the observations of [12] that 2022 was indeed the driest year in ~200 years, the late 20th and early 21st century displayed, for multiple filters, 2022 as representing the end year of many multi-decadal droughts. We selected the four-month (longest) season (AMJJ) and the 23-year filter (2000 to 2022 as this represents flows in the 21st century) for further study.

The Old World Drought Atlas (OWDA) provides the annual June–July–August (JJA) self-calibrated Palmer Drought Severity Index (scPDSI) for 5414 grid points across Europe from 0 to 2012 AD [18]. Per [15,16], the current research utilizes the OWDA scPDSI as a proxy for Pontelagoscuro AMJJ streamflow reconstructions and will include 196 scPDSI cells within a 450 km search radius (Figure 1).

2.2. Streamflow Reconstruction

The AMJJ streamflow will serve as the dependent variable for reconstruction in the forwards–backwards stepwise linear regression (SLR) model [17], while the scPDSI will serve as the independent variable [15,16]. In lieu of developing a single reconstruction model, a nested approach was conducted in which 30-, 40-, and 50-year reconstruction windows were evaluated within the 216-year (1807 to 2022) period of records to capture uncertainty. The retained SLR models were subject to a rigorous evaluation. Prior to input into the SLR model, prescreening (correlation and stability) was performed. Initially, we inspected the correlation between AMJJ streamflow and scPDSI cells to identify positively significant (p ≤ 0.01 or 99% significance) scPDSI cells. Next, we investigated temporal stability analysis by performing a moving correlation window (using 33% of the reconstruction window) between AMJJ streamflow and scPDSI cells, and scPDSI cells containing negative correlation values were considered unstable and were not considered in the SLR model [20]. SLR models for AMJJ streamflow were generated and evaluated for model skill, multi-collinearity, and over-fitting. A forward and backward (standard) SLR model was developed. The model adds and removes predictors until all variables that are not in the model have p-values that are greater than the specified alpha-to-enter value (0.05) and proceeds until all variables in the model have p-values that are less than or equal to the specified alpha-to-remove value (0.10) [14,17]. To evaluate skill, traditional statistical measures were used, including $R^2$ (amount of variance in each model) and $R_{\mathrm{Predicted}}^2$ (using leave-one-out cross-validation, also referred to as drop-one cross-validation). The drop-one cross-validation method is recognized as a more rigorous evaluation when compared with the split sample approach. The variation inflation factor (VIF) indicates the extent to which multi-collinearity is present, and a VIF value close to 1.0 (maximum of 10) is preferred [21]. The Durbin–Watson (DW) statistic was used to determine if autocorrelation was present [22], and the sign test (ST) counts the number of agreements and disagreements between instrumental and reconstructed flow. Finally, the reconstructed annual AMJJ flows were bias-corrected [23].

2.2.1. PALEO-RECON: An Automated Streamflow Reconstruction Tool

To streamline and expand the process detailed in Section 2.2, we developed PALEO-RECON, an automated tool that replicates the reconstruction workflow using scPDSI proxies. This tool enables researchers to apply the same method of selecting, evaluating, and modeling scPDSI cells to reconstruct streamflow not only for the Po River Basin (PRB) but for any point in Europe or North America. PALEO-RECON, openly accessible to the scientific community via GitHub https://github.com/SAIL-UA/PALEO-RECON, accessed on 11 November 2024, allows users to access 5414 cells from the OWDA dataset for Europe, spanning from 0 to 2012 AD [18], and 11,396 cells from the Living Blended Drought Atlas (LBDA) for North America, spanning from 0 to 2005 AD [24]. This automation enables each step of the reconstruction process—previously performed manually—to be completed with greater speed and consistency.

PALEO-RECON automates the entire paleoclimatic reconstruction process via SLR, from selecting scPDSI cells within a user-defined radius to preselecting based on correlation and stability, running stepwise linear regression, and evaluating model quality with metrics such as $R^2$ and $R_{\mathrm{Predicted}}^2$ via leave-one-out cross-validation; DW; and VIF. The tool simplifies the input requirements: as shown in Figure 2, users specify observed data (e.g., precipitation or streamflow), geographic coordinates of the observation site, desired search radius, and reconstruction window length, and PALEO-RECON performs the rest. The tool also provides transparency by mapping the selected cells, and it highlights the river basin if the observation site is within one of the world’s major basins, using data from the Global Runoff Data Centre [25]. Additionally, PALEO-RECON includes a bias correction module based on quantile mapping (RQUANT) [23], which reduces systematic errors and aligns reconstructed values more closely with observed data. With PALEO-RECON, this rigorous reconstruction process becomes readily accessible and reproducible for researchers, advancing paleoclimatic research in any region covered by scPDSI datasets [18,24].

Machine learning (ML) and deep learning (DL) models for paleoclimate reconstructions have the potential to enhance accuracy in certain contexts. PALEO-RECON builds on this potential by providing a versatile foundation for the future integration of advanced ML and DL techniques, extending beyond traditional regression methods. While these methods can outperform traditional approaches, their effectiveness is highly dependent on the quality and quantity of observed data, which can limit reliability when data are sparse or biased. By enabling comparative analysis between traditional SLR and ML/DL models, PALEO-RECON supports deeper insights into hydrological responses under changing climate conditions. This scalable tool positions PALEO-RECON as a valuable resource that can evolve with ongoing advancements in computational methods and expand the options available for paleoclimatic research.

2.3. Drought and Surplus Definitions

The definition of drought (or deficit) and surplus periods was adopted from [26]. These periods are identified when two or more consecutive years of deficit flow (below average flow) or surplus (above average flow) occur. The historical time period of the record (1807 to 2022) is used to assign and define the average to ensure that all analyses are evaluating the impact in relation to current historical conditions. This approach recognizes the varied nature of drought (and surplus) definitions and uses a standard statistical measure. For both deficit and surplus periods, the duration and magnitude (defined as the cumulative of the departure from the average) are determined.

2.4. Whiplash Analysis

Weather (or water) whiplash is defined when the hydrologic state shifts suddenly to overcome an accumulated deficit (or surplus). The method applied here is based on [27], where wet and dry periods are identified in the records based on yearly deviations from the long-term average of the historical record. A wet period (or dry period) is calculated by summing up consecutive dry or wet years to obtain an accumulated surplus (or deficit). Whiplash occurs when the accumulated surplus (or deficit) for the next state is greater than the amount needed from the opposite state in the previous accumulated period.

3. Results

3.1. Streamflow Reconstruction

Ten SLR models were retained (Table 2), and each model achieved a minimum $R^2$ of 0.50; an $R_{\mathrm{Predicted}}^2$ was within 0.1 or less of $R^2$ and passed VIF, DW, and ST statistics. Additionally, no two models could have the same regression equation as in the same combination of retained self-calibrated Palmer Drought Severity Index (scPDSI) cells within the model.

We combined the ten AMJJ stepwise linear regression (SLR) bias-corrected reconstruction models to capture uncertainty (gray lines and 5th and 95th percentiles) and applied a 23-year filter (the black line is the average of the ten models) (Figure 3). The 2000 to 2022 (23-year) drought (14,600 million cubic meters (MCM)) is shown (red line), and a visual inspection was conducted on the 2000-year reconstruction. The 2000 to 2022 23-year AMJJ drought appears to be the most extreme drought since the early 16th century (~500 years, 1522 AD). However, the recent 2000 to 2022 23-year drought was exceeded several times in the paleo record. The mid-5th century (450 AD) was the most extreme 23-year drought (12,506 MCM), followed by the mid-1st century (55 AD, 13,204 MCM), then the early-15th century (1422 AD, 13,572 MCM), and finally the previously identified 1522 AD drought. We next evaluated drought/surplus periods and whiplash in annual bias-corrected reconstructed AMJJ streamflow (average of ten models, 0 to 1806) and annual observed AMJJ streamflow (1807 to 2022).

3.2. Drought and Surplus Periods

The results of deficit and surplus periods (presented as volume and durations) are presented for the entire paleo and historical record (see Figure 4). In addition, the results of the whiplash analysis are presented in Figure 4d. The paleo reconstruction and historic flows show that the largest deficit periods (both in terms of duration and magnitude) occurred in the period 0–500. The largest surplus periods occurred in the 900s and 1600s. The historic period from the 1800s to present show approximately equal presence of surplus and dry periods and magnitudes that are “normal” in the entire record. It is important to note that the uncertainty of paleo records increases considerably as we analyze earlier periods, which may impact the interpretation of extreme events in ancient times.

3.3. Whiplash

The whiplash analysis (Figure 4d) shows a trend toward an enhanced number of whiplash years (wet to dry and dry to wet) during the historical record compared with the paleo record. Overall, the number of whiplash years in a century increased by 2.4 episodes (deficit to surplus) and 2.2 episodes (surplus to deficit) in the entire record. The historical record shows the largest number of years, potentially indicating enhanced climate extremes in the modern period. However, the further back we go in the paleo record, the greater the uncertainty becomes, making it essential to interpret ancient whiplash trends with caution.

4. Discussion

The recent (late 20th and early 21st century) multi-decadal drought in the PRB was extreme. The current research focused exclusively on the 23-year 2000 to 2022 period (21st century) and the April–May–June–July (AMJJ) season. We suspect reconstructions of additional spring–summer months and seasons would yield similar findings of extreme drought. We also believe additional filters of varying lengths (5-, 10-, 15-, and 20-year filters) would show that the end year 2022 was indeed a historical, decadal, and multi-decadal drought period. The impacts in the Po River Basin (PRB), from agricultural loss to discoveries of World War II relics, confirm the severity of the 2022 drought. Future research could examine precipitation across the PRB in an attempt to confirm a similar drought pattern. Additionally, novel artificial intelligence (AI) methods including machine learning (ML) and deep learning (DL) could be applied and compared with the traditional regression approaches in the current research.

Previous research efforts [28] have developed skillful reconstructions of streamflow for the Sava River (Slovenia). The streamflow reconstructions were generated “manually”, in which observed streamflow records were obtained and, based on the location of the gauge, self-calibrated Palmer Drought Severity Index (scPDSI) cells were identified within a 450 km radius. For the Catez gauge (Latitude 45.893333, Longitude 15.61), three streamflow reconstructions were generated, and skill statistics were provided [28]. The same analysis was conducted using PALEO-RECON, yielding results identical to those obtained with the “manual” method (see Supplementary Materials). This independent validation demonstrates that PALEO-RECON is a reliable and effective tool for generating tree ring-based proxy (scPDSI) reconstructions in Europe.

While streamflow reconstructions are limited in Europe, recent research [29] has evaluated 21 gauges in various watersheds (including the Danube, Inn, Elbe, Loire, Main, Rhine, and Wesser) across Europe, which resulted in 14 skillful reconstructions of annual streamflow from 1501 to 2000. It should be noted that no gauges from [29] were in Italy or Slovenia. The streamflow reconstructions consistently identified extreme drought years across Europe in the mid-16th century and mid-17th century [29]. Referring to Figure 3, the most recent extreme drought periods identified in the PRB since 1500 AD occurred in the early–mid-16th century and mid–late-17th century. The 14 reconstructions were evaluated, and the Inn River gauge was identified as the spatially closest watershed to the PRB. Because the Inn River is a tributary of the Danube River, the Danube River gauge was also included in the analysis. Yearly (1501 to 2000) reconstructed calendar year streamflow data for both the Inn River and Danube River were extracted. Additionally, yearly (1501 to 2000) reconstructed AMJJ streamflow data for the PRB were obtained. Each of the three vectors was standardized (mean of zero and standard deviation of one) for the 1501 to 2000 period of record, and a 23-year end-year filter was applied to align with the current research (Figure 5). Although it is acknowledged that the comparison involved annual streamflow for the Inn and Danube Rivers and AMJJ streamflow for the PRB, this represented the best available data for analysis. A visual inspection of Figure 5 indicated that the most severe drought periods (early–mid-16th century and late–19th century) were similar across all three watersheds. The primary difference in the temporal patterns of the three watersheds involved pluvials (wet periods). The PRB displayed extreme wet periods at the end of the 16th/beginning of the 17th century and the early–mid-19th century that were not observed in either the Inn or Danube reconstructions. It was observed that all three watersheds behaved very similarly from the mid-19th to the mid-20th century.

While previous research efforts from the deficit and surplus analysis indicate that the historical record (from the 1800s to present) has “normal” occurrences of deficit and surplus compared with the entire paleo record, future analysis should evaluate the potential of changes in the occurrence of deficit and surplus periods, similar to the study of [26], who found significant changes in both drought and surplus periods depending on climate change scenarios (i.e., model and Representative concentration pathways (RCPs) 4.5, 8.5). The water whiplash analysis did indicate a change in the tendency of shifts from surplus to deficit or from deficit to surplus in short time periods. This is a trend that is seen in other studies for Mediterranean regions around the world [27] and that is likely to continue under climate change scenarios.

The use of paleo data provides a broader temporal context for analyzing long-term drought patterns and extreme hydrological events, extending beyond the instrumental record. These data contribute to understanding climate-driven variability and can inform future research on drought resilience. However, the inherent uncertainty, particularly in older periods, requires careful consideration when interpreting findings and drawing conclusions.

5. Conclusions

This study extends recent findings on the extreme 21st century drought in the Po River Basin (PRB) by reconstructing streamflow over the past two millennia. Our analysis confirms that the 2000–2022 drought is among the most severe multi-decadal droughts since the early 16th century, with similar events occurring only a few times in the past 2000 years. The observed increase in hydrological “whiplash” events, marked by abrupt transitions between drought and pluvial conditions, highlights a potential shift in climate dynamics that could have significant implications for water resource management in the PRB and other Mediterranean regions.

Our analysis also confirmed that 2022 represented the end year of several multi-decadal spring–summer (April–May–June–July (AMJJ)) streamflow droughts (Table 1). Focusing on the 21st century (2000 to 2022), this 23-year period was the most extreme dry (drought) period in the ~200-year observed record. Multiple skillful PRB spring–summer (AMJJ) reconstructions were developed to capture uncertainty. These reconstructions provided a paleo perspective of spring–summer (AMJJ) streamflow, revealing that the 23-year period, 2000 to 2022, drought was the most extreme drought in ~500 years. The combined observed and newly developed paleo streamflow records provide a useful dataset. In analyzing PRB drought (duration, magnitude, severity, intensity), water managers and planners now have ~2000 years of paleo-observed record in lieu of ~200 years of observed record. The paleo-observed record revealed increasing whiplash events, which may be associated with a changing climate.

In this study, we introduce PALEO-RECON, an automated tool that standardizes and streamlines the reconstruction process, facilitating reproducible paleoclimatic studies across Europe and North America. The development of PALEO-RECON, tested using Sava River streamflow [28], provided an efficient and time-saving open-source tool for generating reconstructions using tree ring-based proxies (self-calibrated Palmer Drought Severity Index (scPDSI)). By automating traditional regression methods, PALEO-RECON offers a scalable foundation that could integrate advanced machine learning and deep learning models in future research, providing a flexible platform for investigating hydrological responses to changing climate conditions.

Author Contributions

Conceptualization, A.A.R.M., G.T., and T.P.; methodology, A.A.R.M., G.T., and T.P.; software, A.A.R.M.; validation, A.A.R.M. and G.T.; formal analysis, A.A.R.M., G.T., G.F., and T.P.; investigation, A.A.R.M., G.T., G.F., and T.P.; resources, G.T. and G.F.; data curation, A.A.R.M., G.T., G.F., and T.P.; writing—original draft preparation, A.A.R.M. and G.T.; writing—review and editing, A.A.R.M., G.T., G.F., T.P., and J.G.; visualization, A.A.R.M., G.T., G.F., and T.P.; supervision, G.T. and J.G.; project administration, A.A.R.M. and G.T.; funding acquisition, G.T. All authors have read and agreed to the published version of the manuscript.

Data Availability Statement

The data supporting the results of this study are available and archived at https://alabama.box.com/s/0bwpiyakqq6aihge7frx36qrn9129fku, accessed on 9 November 2024. The PALEO-RECON GitHub repository is openly accessible at https://github.com/SAIL-UA/PALEO-RECON, accessed on 11 November 2024. Additionally, the specific version of PALEO-RECON used in this study is archived on Zenodo and can be accessed via DOI: 10.5281/zenodo.14089242.

Acknowledgments

The authors wish to thank the University of Alabama, Alabama Water Institute (AWI), and Cooperative Institute for Research to Operations in Hydrology (CIROH) for their support. Tootle wishes to thank the members of the J. William Fulbright Scholarship Board; and Paola Sartorio, Barbara Pizzella, and Chiara Petrilli of The U.S.—Italy Fulbright Commission; and Virna Eccli of the Università degli Studi di Trento for their support.

Conflicts of Interest

The authors declare no conflicts of interest.

Abbreviations

The following abbreviations are used in this manuscript:

AI	artificial intelligence
AMJJ	April–May–June–July
DL	deep learning
DW	Durbin–Watson
J	July
JJ	June–July
JJA	June–July–August
LBDA	Living Blended Drought Atlas
MCM	million cubic meters
MJJ	May–June–July
ML	machine learning
OWDA	Old World Drought Atlas
PRB	Po River Basin
Q	observed streamflow
RCPs	Representative concentration pathways
scPDSI	self-calibrated Palmer Drought Severity Index
ST	sign test
SLR	stepwise linear regression
VIF	variation inflation factor

Footnotes

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Figures and Tables

Table 1

Observed streamflow (Q) from 1807 to 2022 for J, JJ, MJJ, and AMJJ with the filter end year being 2022. The highlighted “black cell” identifies periods (J Q, JJ Q, MJJ Q, AMJJ Q) in which the end year 2022 was the lowest streamflow. The numbers (1 to 30) represent the filter length (number of years) with the final year being 2022. For example, 18 represents an 18-year filter (2005 to 2022).

	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15
J Q
JJ Q
MJJ Q
AMJJ Q
	16	17	18	19	20	21	22	23	24	25	26	27	28	29	30
J Q
JJ Q
MJJ Q
AMJJ Q

Table 2

April–May–June–July (AMJJ) observed streamflow (Q), reconstruction period, $R^2$, $R_{\mathrm{Predicted}}^2$ ($R_{\mathrm{P}}^2$), variation inflation factor (VIF), Durbin–Watson (DW), sign test (ST), and regression equation for the regression model (scPDSI proxy cell number in parenthesis).

Period	${\mathit{R}}^2$	${\mathit{R}}_{\mathbf{P}}^2$	VIF	DW	ST	Equation
1889–1918	0.55	0.45	1.3	2.09	14/16	AMJJ Q = 20,954 + 1174 (1625) + 1270 (1029)
1923–1952	0.53	0.45	1.0	2.14	15/15	AMJJ Q = 17,528.7 + 2267 (1726)
1974–2003	0.50	0.42	1.0	2.16	15/15	AMJJ Q = 16,936 + 2494 (1363)
1977–2006	0.74	0.67	1.2	2.12	14/16	AMJJ Q = 18,149 + 2355 (1363) + 1128 (1262)
1978–2007	0.75	0.68	1.3	2.39	14/16	AMJJ Q = 17,822 + 2190 (1363) + 939 (1262)
1980–2009	0.65	0.57	1.2	2.50	15/15	AMJJ Q = 17,831 + 1943 (1215) + 1084 (1315)
1983–2012	0.76	0.71	1.2	2.17	14/16	AMJJ Q = 18,781 + 1120 (1362) + 1753 (1216) + 774 (1576)
1970–2009	0.61	0.54	1.5	2.23	21/19	AMJJ Q = 17,230 + 1497 (1216) + 1281 (1363)
1914–1963	0.50	0.43	1.0	2.30	26/24	AMJJ Q = 18,532 + 1599 (1442) + 1287 (1488)
1959–2008	0.50	0.43	1.6	1.93	26/24	AMJJ Q = 16,453 + 1226 (1363) + 1324 (1176)

Supplementary Materials

The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/hydrology11120219/s1. The Supplementary Materials are available as a single ZIP file, which contains the results of the comparison between the streamflow reconstructions conducted in [28] and those generated using PALEO-RECON. The ZIP file also includes a README.pdf document that provides a detailed explanation of the Supplementary Materials, including their organization and content.

References

1. Vezzoli, R.; Mercogliano, P.; Pecora, S.; Zollo, A.; Cacciamani, C. Hydrological simulation of Po River (North Italy) discharge under climate change scenarios using the RCM COSMO-CLM. Sci. Total. Environ.; 2015; 521-522, pp. 346-358. [DOI: https://dx.doi.org/10.1016/j.scitotenv.2015.03.096]

2. Puspitarini, Po River Basin Authority. Caratteristiche del Bacino del Fiume Po e Primo Esame dell’ Impatto Ambientale Delle Attivitá Umane Sulle Risorse Idriche (Characteristics of Po River Catchment and First Investigation of the Impact of Human Activities on Water Resources); Technical report Autorità di bacino del fiume Po: Parma, Italy, 2006; (In Italian)

3. Ninfo, A.; Ciavola, P.; Billi, P. The Po Delta is restarting progradation: Geomorphological evolution based on a 47-years Earth Observation dataset. Sci. Rep.; 2018; 8, 3457. [DOI: https://dx.doi.org/10.1038/s41598-018-21928-3] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/29472570]

4. Musolino, D.; Vezzani, C.; Massarutto, A. Drought Management in the Po River Basin, Italy. Drought; John Wiley & Sons, Ltd.: Hoboken, NJ, USA, 2018; Chapter 3.11 pp. 201-215. [DOI: https://dx.doi.org/10.1002/9781119017073.ch11]

5. Zanchetin, D. Monthly Discharge of Po River at Pontelagoscuro, 2022. Dataset. Available online: https://doi.org/10.5281/zenodo.7225699 (accessed on 3 July 2024).

6. Bozzola, M.; Swanson, T. Policy implications of climate variability on agriculture: Water management in the Po river basin, Italy. Environ. Sci. Policy; 2014; 43, pp. 26-38. [DOI: https://dx.doi.org/10.1016/j.envsci.2013.12.002]

7. Nadeau, B.L.; Borghese, L. ‘We Are in Extreme Crisis’. Italian Parmesan Producers Fear for Future Amid Drought. 2022; Available online: https://www.cnn.com/2022/07/01/europe/italy-drought-river-po-climate-intl-cmd/index.html (accessed on 31 October 2024).

8. Ashkenaz, A. World War 2 Mystery Solved as German Vehicle Rises from River After Severe Drought. 2022; Available online: https://www.express.co.uk/news/science/1587478/wold-war-2-archaeology-news-german-vehicle-found-drought-mystery-solved (accessed on 31 October 2024).

9. Kuta, S. Drought in Italy Reveals Sunken World War II Barge. 2022; Available online: https://www.smithsonianmag.com/smart-news/drought-italy-exposes-sunken-world-war-ii-barge-180980297/ (accessed on 31 October 2024).

10. Bonnel, O. In the Po Valley, a Historic Drought Is Threatening Italy’s Breadbasket. 2022; Available online: https://www.lemonde.fr/en/climate-change/article/2022/07/09/in-the-po-valley-a-historic-drought-is-threatening-italy-s-breadbasket_5989560_152.html (accessed on 31 October 2024).

11. Press, T.A. Italy’s Lake Garda Shrinks to Near-Historic Low Amid Drought. 2022; Available online: https://www.nbcnews.com/news/world/italys-lake-garda-shrinks-historic-low-drought-rcna42996 (accessed on 31 October 2024).

12. Montanari, A.; Nguyen, H.; Rubinetti, S.; Ceola, S.; Galelli, S.; Rubino, A.; Zanchettin, D. Why the 2022 Po River drought is the worst in the past two centuries. Sci. Adv.; 2023; 9, eadg8304. [DOI: https://dx.doi.org/10.1126/sciadv.adg8304] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/37556532]

13. Obertelli, M. A Data-Driven Approach to Streamflow Reconstruction Using Dendrochronological Data. Master’s Thesis; Politecnico di Milano, School of Civil, Environmental and Land Management Engineering: Milan, Italy, 2020.

14. Formetta, G.; Tootle, G.; Therrell, M. Regional Reconstruction of Po River Basin (Italy) Streamflow. Hydrology; 2022; 9, 163. [DOI: https://dx.doi.org/10.3390/hydrology9100163]

15. Ho, M.; Lall, U.; Cook, E.R. Can a paleodrought record be used to reconstruct streamflow?: A case study for the Missouri River Basin. Water Resour. Res.; 2016; 52, pp. 5195-5212. [DOI: https://dx.doi.org/10.1002/2015WR018444]

16. Ho, M.; Lall, U.; Sun, X.; Cook, E.R. Multiscale temporal variability and regional patterns in 555 years of conterminous U.S. streamflow. Water Resour. Res.; 2017; 53, pp. 3047-3066. [DOI: https://dx.doi.org/10.1002/2016WR019632]

17. Woodhouse, C.A. A 431-Yr Reconstruction of Western Colorado Snowpack from Tree Rings. J. Clim.; 2003; 16, pp. 1551-1561. [DOI: https://dx.doi.org/10.1175/1520-0442-16.10.1551]

18. Cook, E.R.; Seager, R.; Kushnir, Y.; Briffa, K.R.; Büntgen, U.; Frank, D.; Krusic, P.J.; Tegel, W.; van der Schrier, G.; Andreu-Hayles, L. et al. Old World megadroughts and pluvials during the Common Era. Sci. Adv.; 2015; 1, e1500561. [DOI: https://dx.doi.org/10.1126/sciadv.1500561] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/26601136]

19. Ramírez Molina, A.A. PALEO-RECON: An Automated Tool for Paleoclimate Reconstructions [Software] (v.1.0.1). Zenodo; 2024; [DOI: https://dx.doi.org/10.5281/zenodo.14089242]

20. Biondi, F.; Waikul, K. DENDROCLIM2002: A C++ program for statistical calibration of climate signals in tree-ring chronologies. Comput. Geosci.; 2004; 30, pp. 303-311. [DOI: https://dx.doi.org/10.1016/j.cageo.2003.11.004]

21. O’Brien, R.M. A Caution Regarding Rules of Thumb for Variance Inflation Factors. Qual. Quant.; 2007; 41, pp. 673-690. [DOI: https://dx.doi.org/10.1007/s11135-006-9018-6]

22. Durbin, J.; Watson, G.S. Testing for Serial Correlation in Least Squares Regression: I. Biometrika; 1950; 37, pp. 409-428. [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/14801065]

23. Robeson, S.M.; Maxwell, J.T.; Ficklin, D.L. Bias Correction of Paleoclimatic Reconstructions: A New Look at 1,200+ Years of Upper Colorado River Flow. Geophys. Res. Lett.; 2020; 47, e2019GL086689. [DOI: https://dx.doi.org/10.1029/2019GL086689]

24. Cook, E.R.; Seager, R.; Heim, R.R., Jr.; Vose, R.S.; Herweijer, C.; Woodhouse, C. Megadroughts in North America: Placing IPCC projections of hydroclimatic change in a long-term palaeoclimate context. J. Quat. Sci.; 2010; 25, pp. 48-61. [DOI: https://dx.doi.org/10.1002/jqs.1303]

25. Global Runoff Data Centre. Major River Basins of the World (Version 2nd ed.). [Dataset]. 2020; Available online: https://mrb.grdc.bafg.de/ (accessed on 31 October 2024).

26. Bedri, R.; Piechota, T. Future Colorado River Basin Drought and Surplus. Hydrology; 2022; 9, 227. [DOI: https://dx.doi.org/10.3390/hydrology9120227]

27. Madrigal, C.; Bedri, R.; Piechota, T.; Li, W.; Tootle, G.; El-Askary, H. Water Whiplash in Mediterranean Regions of the World. Water; 2024; 16, 450. [DOI: https://dx.doi.org/10.3390/w16030450]

28. Tootle, G.; Oubeidillah, A.; Elliott, E.; Formetta, G.; Bezak, N. Streamflow Reconstructions Using Tree-Ring-Based Paleo Proxies for the Sava River Basin (Slovenia). Hydrology; 2023; 10, 138. [DOI: https://dx.doi.org/10.3390/hydrology10070138]

29. Nasreen, S.; Součková, M.; Vargas Godoy, M.R.; Singh, U.; Markonis, Y.; Kumar, R.; Rakovec, O.; Hanel, M. A 500-year annual runoff reconstruction for 14 selected European catchments. Earth Syst. Sci. Data; 2022; 14, pp. 4035-4056. [DOI: https://dx.doi.org/10.5194/essd-14-4035-2022]