An unexpected negative feedback between climate change and eutrophication: higher temperatures increase denitrification and buffer nitrogen loads in the Po River (Northern Italy)

The Po River is the longest river in Italy, flowing eastward across Northern Italy for over 650 km (figure 1), and is also the largest river, with an average discharge of ∼1500 m³s⁻¹ at its closing section [33]. The Po drainage basin extends over an area of ∼75 000 km², a large portion of which constitutes the widest and most fertile lowland in Italy (∼47 000 km²). The Po River is supported by both Alpine and Apennine's tributaries, fed mainly by snowmelt and rainfall, respectively, resulting in an annual flow regime that is characterized by two flood periods (in spring and late autumn) and two low-water periods (in summer and winter) [34]. The basin covers the transition zone between the sub-continental climate of Central Europe and the Mediterranean climate, with an average annual precipitation of approximately 1200 mm [35]. The Po River basin is densely urbanized and an intensely exploited area, accounting for 40% of Italy's gross domestic product and 35% of national agricultural production. With some of the highest rates of N losses to surface water and groundwater [19, 36, 37], this region is responsible for approximately two-thirds of the total nutrient inputs to the Northern Adriatic Sea [38–40].

Zoom In Zoom Out Reset image size

Accurate and continuous water temperature datasets, representative of the middle-lower reach of the Po River, were acquired from two monitoring stations, operated by energy companies, located at the cooling water intake of the power plants. Daily average water temperature data were recorded near the city of Piacenza (Emilia-Romagna Region, stream kilometer 330) at La Casella Power Station by the ENEL group (Italian multinational manufacturer and distributor of electricity and gas; www.enel.com/it/media/esplora/ricerca-foto/photo/2020/03/italia-centrale-la-casella) from 1992 to 2005, and at Piacenza Power Station by the A2A Life Company Group (www.a2a.eu/en/group) from 2006 to 2019, giving a complete dataset for the period 1992–2019. Water temperature measurements were carried out using resistance temperature detector (RTD) probes with platinum Pt100 resistance thermometers having a nominal resistance of 100 Ω at 0 °C defined according to IEC 751 (EN 60751) .Other sensor characteristics: measuring range 0–40 °C; accuracy ±0.1 °C at 0° C; 4-wire connection; signal conversion electronics with 4–20 mA output in measuring range 0 °C–40 °C. The validation procedure to reconstruct a continuous three-decade time series is reported in the supplementary material 1. From temperature daily data, the annual and seasonal trends in average values were analyzed for the spring (April–June) and summer (July–September) periods.

Monthly ${\text{NO}}_3^ -$ and ammonium (${\text{NH}}_4^ +$) loads and total nitrogen (TN) exported to the Adriatic Sea were calculated using discharge and concentration datasets for the study period (1992–2019) at the closing section of the Po River basin, which is conventionally located at Pontelagoscuro (44°53'19.34''N, 11°36'29.60''E) near the city of Ferrara (Emilia-Romagna Region; stream kilometer 586). Daily average discharge was acquired from the permanent records of a gauge operated by the Environmental Agency (ARPAE) of the Emilia-Romagna Region and retrieved from the 'Hydrological Annals—Second Part' published by ARPAE, the electronic versions of which are available on the Regional Open Data Portal (https://simc.arpae.it/dext3r/). Nitrogen species concentrations were obtained from fortnightly (or monthly) sampling campaigns carried out by ARPAE under the framework of the environmental monitoring program (https://dati.arpae.it/group/acqua). Sample collection and analysis were performed in accordance with standard methods and analytical protocols adopted by regional environmental agencies [41]. When not provided, TN concentrations were calculated from the concentrations of DIN (${\text{NO}}_3^ -$ + ${\text{NH}}_4^ +$) according to the formula TN = 0.93 × DIN + 0.75 (r² = 0.54; p < 0.001), obtained by relating time series including simultaneous TN and DIN measurements.

Nutrient loads were calculated as the product of the daily discharge and nutrient concentration (measured fortnightly or monthly and interpolated to daily intervals) and aggregated into monthly means. The method employed for the monthly load calculation was based on the linear interpolation of concentration values between two subsequent sampling events [42, 43], as follows (1):

$$\begin{equation}L = k \cdot \sum\limits_{j = 1}^n C_j^{\text{int} } \cdot { }{Q_j}\end{equation} \tag{ 1 }$$

where C_j ^int is the daily N species concentration (g N m⁻³) linearly interpolated between two measured samples, Q_j is the daily discharge (m³ s⁻¹), n is the number of days in each month, and k is a conversion coefficient to take the recorded period into account (e.g. 365 d for annual loads). Seasonal load trends in the spring and summer periods (t N season⁻¹) were evaluated according to the following monthly clustering: April–June (spring) and July–September (summer). Annual loads (t N yr⁻¹) were computed by summing up all the monthly contributions. To validate the annual loads calculated by the interpolation concentration method, the obtained values were compared to those calculated by flow-adjusted concentration method. Flow-adjusted concentrations are commonly employed for assessing annual loads and are recommended in monitoring guidelines [44] and international conventions (e.g. OSPAR-Convention for the protection of the marine environment of the North-East Atlantic) [45], but they are not valid for calculating monthly (and thus seasonal) loads because the environmental monitoring quality programs typically carry out just one sampling per month. A very good correlation between the annual values calculated by the two methods was found (r² = 0.99, p < 0.001) and a discrepancy of about 5% on average (see supplementary material 2).

With the aim of assessing if long-term nutrient load trends might be mediated by the Po River water temperature trends, monthly flow-normalized loads (Ln) were calculated according to [46] to remove the effects of varying inter-annual hydrological conditions on N transport:

$$\begin{equation}{\text{Ln}} = L \cdot Kh\end{equation} \tag{ 2 }$$

where Kh (hydrological coefficient). The hydrological coefficient was obtained as the ratio of the long-term (period 1992–2019) average outflow of a specific month to the monthly outflow of a particular year. The annual normalized loads were computed by summing up the monthly normalized loads. Similarly, the seasonal normalized loads were calculated by summing up the normalized loads from April to June and from July to September, for spring and summer period, respectively.

Because the Po River basin is among the most agriculturally productive and densely populated areas in Italy, changes in agricultural practices and populations could result in changes in riverine N loading. The temporal evolution of diffuse and point N sources in the watershed was checked by collecting census data at an almost 10 year time interval for agricultural land occupied by different crop types and production systems, numbers of farmed animals, synthetic fertilizer application practices, and human population. Statistics were integrated in a N budgeting approach previously applied to several sub-basins of the Po River system [37, 47, 48]. Details regarding the data sources, computational methods, and uncertainty assessment of the diffuse and point N sources are presented in supplementary material 3.

Annual and seasonal time series of temperature, riverine N loads, and water flow were analyzed using parametric (linear regression) and non-parametric tests (Mann–Kendall, Sen's slope, and Pettitt's test). Pearson correlation analysis was used to investigate the relationship between temperature and riverine N loads. All statistical tests were performed using the software R (Core Team, 2021) with the Kendall package for the Mann–Kendall test and the Trend package [49] for the other analyses. The tested factors and trends were considered statistically significant at p < 0.05. Details of the statistical tests are presented in supplementary material 4.

During the period 1992–2019, the annual TN loads at the closing section of the Po River basin showed a significant negative trend (p< 0.05, figure 2(a)), decreasing by nearly 33%, corresponding to a reduction of approximately 2000 t yr⁻¹. Depending on outflow variations linked to precipitation, the TN export varied greatly among years, ranging between ∼68 000 t N yr⁻¹ (2007 and 2017) and ∼237 000 t N yr⁻¹ (1996). As is commonly found in agricultural settings [15, 16], the nitrate load accounted, on average, for >75% (range = 62%–86%) of the TN load, whereas the contribution of ${\text{NH}}_4^ +$ was comparatively minor (range = 1%–5%) (figure 2(a)). Compared to the early 1990s, the ${\text{NO}}_3^ -$ load declined over the study period by more than 30% (p < 0.05, figure 2(a)), showing inter-annual variations that coincided with those detected in the TN load. The highest annual ${\text{NO}}_3^ -$ export (∼160 000 t yr⁻¹) occurred in 1996, while the lowest amount (∼50 000 t N yr⁻¹) occurred in both 2007 and 2017. Over the study period, the annual ${\text{NH}}_4^ +$ load decreased by approximately two-thirds (p< 0.001, figure 2(a)) from ∼6300 t N yr⁻¹ in the early 1990s to less than 2000 t N yr⁻¹ in recent years. The hydrological conditions have also varied significantly during this period, although there has been no significant long-term trend in the annual outflow. For example, 2007 and 2017 were extremely dry, with outflow values 42%–45% lower than the long-term average and corresponding to lower N transport. Conversely, 2014 was an extremely wet year with an annual outflow >50% higher than the long-term average, and consequently higher N transport. In the Po River, early signals of climate change effects have been reported over the last three decades, when hydrological extremes have become progressively amplified [19, 33, 34], with large floods followed by persistent drought conditions [50, 51].

Zoom In Zoom Out Reset image size

The trajectories of riverine N loads were not related to human pressures, productive sectors and the associated generation of N loads from diffuse and point sources. Indeed, the N balance across the croplands of the Po River basin revealed a steadily constant surplus during the 1990–2019 period, averaging ∼180 kt N yr⁻¹ (figure 3). The total N input during this period was estimated to exceed 600 kt N yr⁻¹, mostly derived by manure spreading (36%), synthetic fertilizers (33%) and biological fixation (26%). The total N output during the study period was estimated to exceed 430 kt N yr⁻¹, mainly associated with crop harvesting (74%). Total watershed N inputs to croplands showed a slight decline in 2010 (∼14%) with respect to the previous two decades, but this was coupled to a decrease also in total N outputs (∼15%) resulting, if the associated uncertainty is considered, in a net budget (i.e. surplus) not significantly different over the studied period. While the human population in the Po River basin has remained relatively constant over the last three decades at ∼17 million, important legislative acts aimed at improving urban wastewater treatment plants (e.g. Directive 91/271/EEC) were followed by an appreciable reduction in the direct discharge of untreated or poorly treated domestic wastewater [28]. Nitrogen loads from point sources decreased by nearly 45% between 1990 and 2000 and then remained almost constant until 2019 (figure 3) and this may have been partly responsible for the clear decrease of riverine ${\text{NH}}_4^ +$ loads. Despite this, the decrease was not in the order of magnitude to explain the decrease recorded for the riverine TN loads. Overall, over the entire investigated period, N loads from urban areas accounted for less than 5% of the total N input from diffuse agricultural sources. Since the early 1990s, ${\text{NO}}_3^ -$ pollution has become the main concern for surface water and groundwater in the Po River basin because the measures introduced by the European Directives for controlling widespread agricultural and livestock sources (i.e. 91/676/EEC, 2000/60/EC) have been largely ineffective [27, 52]. Recent studies have shown that in agricultural landscapes, artificial water bodies such as irrigation canals and drainage ditches may act as natural wetlands in terms of provision of biogeochemical services, i.e. the mitigation of N excess via denitrification [47, 53]. The capillary network of artificial waterways crossing the Po River plain was implemented over the centuries, from the Etruscan age to the 1960s, with multiple purpose, i.e. irrigation, drainage, and flood control [54–56]. It is reasonable to hypothesize that the N amount removed via denitrification by the whole canal network remained stable along the three decades analyzed in the present study and thus it is very unlikely to explain the major reduction observed in the Po River ${\text{NO}}_3^ -$ loads, whose cause is to be found elsewhere.

Zoom In Zoom Out Reset image size

Spring and summer nutrient loads represented on average 19%–24% and 13%–14% of the corresponding annual values, respectively (figures 2(b) and (c)). Summer TN and ${\text{NO}}_3^ -$ loads exhibited high inter-annual variations, ranging from ∼8000 (2003) to ∼3000 t N season⁻¹ (2002), and from ∼5500 (2003) to ∼27 600 t N season⁻¹ (2002), respectively. The analyzed dataset contained years with rather extreme summertime hydrological conditions; the summers of 2002 and 2014 were very wet, with outflow 56%–66% higher than the long-term summer average. In contrast, the summers of 2003 and 2007 were extremely dry, with outflow 39%–53% lower than the 1992–2019 average. The period from 2003 to 2007 was characterized by frequent and persistent summer drought that culminated in daily discharge frequently <300 m³s⁻¹. Of the six most-prolonged drought events recorded during the last century, four occurred between 2003 and 2007, with the lowest daily discharge of ∼170 m³s⁻¹ occurring in July 2006 [33, 57, 58]. The time series of summer loads exhibited a negative trend for TN and ${\text{NO}}_3^ -$ (p < 0.01, figure 2(c)), decreasing on average by 42%–47%, while a significant downward trend, if tested by linear regression, was not detected in spring when load variations among years were more erratic (figure 2(b)). Differently to TN and ${\text{NO}}_3^ -$, ${\text{NH}}_4^ +$ loads decreased by nearly 62% in spring (p < 0.01; figure 2(b)), whereas linear regression was no statistically significant in summer (figure 2(c)).

Summer outflow decreased by nearly 34% (1.3% per year), highlighting that drought events have been exacerbated during the more recent decades as previously demonstrated by hydrological studies [25, 33, 34]. The calculation of flow-normalized loads showed that the annual transport of TN, ${\text{NO}}_3^ -$, and ${\text{NH}}_4^ +$ at the Po River closing section decreased by 15%, 14%, and 61%, respectively, along the entire investigated period (figures 4(a), (d), and (g)). The results of the Mann–Kendall and Sen's slope analyses on flow-normalized nutrient loads showed negative Z values, confirmed by a negative slope, indicating downward trends since 1992 both at the annual and seasonal scale (table 1). The Pettitt's test showed that the decline in seasonal nutrient loads began in 2006 (figures 4(b), (c), (e) and (f)), except for ${\text{NH}}_4^ +$ for which trends began in 2008 for spring (figure 4(h)) and in 2009 for summer (figure 4(i)), resulting in annual loads started to decrease around 2010.

Zoom In Zoom Out Reset image size

Table 1. Results of the statistical analyses.

		Linear regression	Mann–Kendall			Sen's slope	Pettitt
	Period	p-value	p-value	S	Z	Q	K	Year
Flow-normalized TN loading	Annual	0.001	<0.001	−132	−2.59	−744.77	134 316	2010
	Spring	0.001	<0.001	−170	−3.34	−326.40	33 282	2006
	Summer	—	<0.001	−74	−1.44	−66.61	20 519	2006
Flow-normalized ${\text{NO}}_3^ -$ loading	Annual	0.05	<0.001	−92	−1.80	−630.06	104 556	2011
	Spring	0.001	<0.001	−152	−2.98	−221.59	24 275	2006
	Summer	—	<0.001	−26	−0.49	−19.93	15 024	2006
Flow-normalized ${\text{NH}}_4^ +$ loading	Annual	<0.001	<0.001	−214	−4.21	−126.44	5493	2005
	Spring	0.01	<0.001	−160	−3.14	−22.02	700	2008
	Summer	—	<0.001	−36	−0.69	−3.37	745	2009
Temperature	Annual	<0.001	<0.001	236	4.64	0.12	14.0	2002
	Spring	0.01	<0.001	160	3.14	0.09	17.1	2002
	Summer	<0.001	<0.001	192	3.77	0.14	21.1	2002
Outflow	Annual	—	<0.001	−62	−1.20	−0.39	60.45 × 109	2002
	Spring	—	<0.001	−122	−2.39	−0.14	12.60 × 109	2002
	Summer	0.05	<0.001	−20	−0.37	−0.03	17.29 × 109	2002

Significant positive trends in the annual, spring, and summer water temperature series of the Po River were identified for the 1992–2019 period (figure 5), as demonstrated by the positive Sen's slope values (table 1). The annual average temperature increased during this period by ∼3 °C, corresponding to an overall warming rate of 0.11 °C yr⁻¹, although the pattern of change showed two moments: the annual series from 1992 to 2002 were characterized by relative stability with an average temperature of 13.87 ± 0.22 °C and low inter-annual variability; while an abrupt increase occurred after 2002 with a slope of more than 0.18 °C yr⁻¹ and high fluctuations among the years (average 15.88 ± 0.88 °C) (figure 5). The highest annual temperatures (up to ∼17 °C) were recorded in 2007 and 2015, 2 years marked by significant thermal (high air temperature) and meteorological (low precipitation) signals [26, 32]. Seasonally, the average spring and summer water temperatures increased by nearly 2 °C (0.07 °C yr⁻¹) and 3.5 °C (0.13 °C yr⁻¹) over the monitoring period, respectively (figures 5(b) and (c)), with the most marked warming trends and inter-annual variability starting in 2002 (table 1). These temperature increases resulted to be faster than the average increases observed in other large European and American rivers in temperate zones during similar periods [59, 60]. However, the present outcomes agree with previous studies indicating a major contribution to warming from the hottest period of the annual cycle with stronger positive trends for late spring–summer months and a significant advance of spring warming [61–65].

Zoom In Zoom Out Reset image size

Meteorological stations located nearby the Po River course showed a significant positive trend for air temperature, recording an increase of about 2 °C in annual and summer average values and an increase of about 1 °C in spring average values over the last three decades (figures S2 and S3, supplementary material 1). The present data was confirmed by previous meteorological studies that have demonstrated how the air temperature in Po River basin has been affected by warming in the period 1952–2002, recording an increase of over 1 °C for average annual values [66] and detecting stronger positive anomalies in the mountain areas compared to the lowlands and the delta region [67]. Further studies have demonstrated an increase in annual maximum temperatures with linear and constant trends of about 0.5 °C every 10 year and predicted a raise of 3 °C–4 °C by the end of the last decade, as it happened [68] and an even higher temperature anomaly for the next decades [66].

Pettitt's test on the Po River water temperature highlighted a positive trend starting in 2002 (table 1) and this was consistent with the most marked increase in air temperature detected from the beginning of the 2000s (figure S3, supplementary material). Despite long-term increases in river water temperatures being correlated to increases in air temperatures, surprisingly, the warming trend of the Po River water was stronger than the atmosphere, when the latter is supposed to contribute to the warming of the former. These unusual data may be ascribed to the joint effect of rising air temperature and reduced outflow on river temperature trends [65].

In parallel to the upward temperature trends, the annual occurrence of warm days (i.e. the number of days with water temperatures above the long-term average) increased by more than 50% for both the spring and summer periods (figure 6). This condition was in agree with previous studies reporting, for the Mediterranean area, a significant increase of the days with warm temperature extremes [69–71], suggesting that the growing season length is increasing. The occurrence of warm days in summer is often related to low-flow conditions, as was the case for the period from 2003 to 2007, which was characterized by prolonged drought in the Po River basin. However, this has not been the case in the last decade, indicating that the Po River is becoming more sensitive and vulnerable to such extreme temperature events with ongoing climate change, as demonstrated for other large European rivers [64].

Zoom In Zoom Out Reset image size

The present outcomes demonstrated that the Po River water is steadily warming, with the number of warm days increasing over time and higher water temperatures corresponding to lower N loads during the entire spring–summer period, the time of year when the risk of coastal zone eutrophication is greatest [72, 73]. Indeed, highly significant negative (p < 0.0001) correlations were detected between average water temperature and monthly loads of TN and ${\text{NO}}_3^ -$ (figures 7(a)–(d)). When the temperature increased by 1 °C, TN and ${\text{NO}}_3^ -$ loads decreased by approximately 7% and 4% in summer and spring, respectively. A weaker but still significant negative correlation (p < 0.05) was also found between the average water temperature and monthly ${\text{NH}}_4^ +$ loads in spring (figure 7(e)). The inverse relationship observed between temperature and TN loads (mainly ${\text{NO}}_3^ -$) strongly indicates that the higher water temperatures recorded during the last few decades have stimulated ${\text{NO}}_3^ -$ removal via denitrification in the river sediments along the lowland reaches (figure 7). This likely act to partially buffer the eutrophication risk in the coastal waters. While several studies suggest that water temperature increases may alter the biodiversity and biological structure and functioning of rivers [59, 74], the resulting effects on ecosystem functions (i.e. N removal) and, ultimately, the regulation of ecosystem services (i.e. self-depuration capacity) remains unclear and warrant greater attention. Experimental laboratory studies have shown that warming boosts nitrification and denitrification rates alongside enzymatic reactions in freshwater sediments [17, 18, 75], but there is a lack of systematic research forecasting global warming effects on N cycling in rivers and expected changes in N loads [76]. When a suitable substrate, ${\text{NO}}_3^ -$, and labile carbon are available, denitrification generally responds positively to increases in water temperature. At the closing section of the Po River, dissolved organic carbon during the spring–summer months average 1.8 mg l⁻¹, indicating that organic carbon is balanced with respect to ${\text{NO}}_3^ -$ availability (averaging 1.7 mg N l⁻¹, 1992–2019 period) according to a theoretical ratio of ∼1 based on denitrification stoichiometry [77]. The dissolved organic carbon concentrations in the lower reaches of the Po River tally with those measured in other agricultural rivers [78, 79], which demonstrates that denitrification is not likely limited by the organic carbon supply. Higher water temperatures decrease oxygen solubility and increase sediment oxygen respiration, thereby limiting the oxygen penetration depth and resulting in a synergistic indirect effect that strengthens the denitrification capacity [17, 75]. The inverse relationship between water temperature and ${\text{NH}}_4^ +$ loads in spring also suggests that warming may stimulate nitrifying activity (figure 7(e)). Po River water column is indeed thoroughly mixed, thus dissolved oxygen concentrations are typically at or near 100% saturation, and the oxygenation of surface sediments is likely sufficient to support coupled nitrification–denitrification. However, as is widely reported, when water ${\text{NO}}_3^ -$ concentrations exceed 0.5 mg N l⁻¹, denitrification is expected to be fueled mainly by ${\text{NO}}_3^ -$ diffusing from the water column to the anoxic sediment layers [15, 16].

Zoom In Zoom Out Reset image size

All biogeochemical ${\text{NO}}_3^ -$ dissimilative pathways, including denitrification and DNRA (dissimilatory ${\text{NO}}_3^ -$ reduction to ${\text{NH}}_4^ +$), may be affected by water warming, both as a direct temperature effect on enzyme activity and as indirect temperature effect on sediment redox conditions (i.e. oxygen shortage because of decreased oxygen solubility or enhanced consumption rates). Organic carbon availability generally determines whether denitrification or DNRA will dominate in ${\text{NO}}_3^ -$ reduction, with organic enrichment and reducing (sulfidic) conditions under persistent stratification shifting ${\text{NO}}_3^ -$ reduction towards more pronounced DNRA, with internal ${\text{NO}}_3^ -$ recycling to ${\text{NH}}_4^ +$ [80, 81]. However, this is not the case in the Po River where sediments are sandy and organic matter content is generally low [82]. Stimulation of DNRA by increased water temperature cannot be completely excluded, but this would have contributed to ${\text{NH}}_4^ +$ accumulation in water, a condition not evidenced. On the contrary, the inverse relationship between water temperature and ${\text{NH}}_4^ +$ loads suggested that warming might also have stimulated nitrifying activity as, in the Po River, the water column is constantly mixed and oxygen saturated, a condition favoring ${\text{NH}}_4^ +$ consumption via nitrification–denitrification coupling. Despite direct measurements are still lacking, on the base of the evidence reported here, DNRA is likely a negligible pathway of N cycling in the Po River sediments.

The links between climate change and eutrophication are being debated and outcomes of many previous studies pointed towards an aggravation of eutrophication due to warming lentic water bodies [83]. Differently, warming and an increase in the duration of low-flow conditions might enhance the denitrification capacity of the river as a whole and partially reduce the risk of eutrophic conditions in the coastal zones. As temperatures are projected to increase in temperate regions over the coming decades, the present outcomes suggest an enhanced future denitrification, representing a natural way to counteract the harmful effects of eutrophication. Air temperatures are expected to rise across the entire Po River basin during all seasons and water temperatures will likely track this trend with the most significant changes occurring in summer alongside reductions in discharge [84]. A decrease in eutrophication phenomena in the Po River delta and nearby coastal zones may be expected, in the medium term, due to negative feedback between climate change and eutrophication in association with a potential water quality improvement.