Mechanisms of Nitrogen Cycling Driven by Salinity in Inland Plateau Lakes, Based on a Haline Gradient Experiment Using Pangong Tso Sediment

Abstract

Pangong Tso, a typical plateau lake exhibiting an east-to-west gradient from freshwater to saline conditions, was used to simulate the migration and transformation of nitrogen compounds under different salinity conditions. This study systematically investigates the effects of salinity on nitrogen cycling and transformation in Pangong Tso sediments from 12 sites through controlled laboratory experiments and field monitoring across 120 sites. The data analysis method includes correlation analysis, ANOVA, structural equation modeling (SEM), and mixed-effects modeling (MEM). The results demonstrate that salinity significantly affects nitrogen cycling in plateau lakes. Salinity inhibits nitrification, resulting in an accumulation of ammonium nitrogen (NH₄⁺-N), while simultaneously suppressing gaseous nitrogen emissions through the inhibition of denitrification. Salinity has a significant negative effect on nitrite nitrogen (NO₂⁻-N), which is attributable to enhanced redox-driven transformations under hypersaline conditions. A salinity threshold of approximately 9‰ was identified, above which nitrite oxidation was strongly inhibited, consistent with the known high salinity sensitivity of nitrite-oxidizing bacteria (NOB). Higher salinity levels correlated positively with increased NH₄⁺-N and total nitrogen (TN) concentrations in overlying water (p < 0.01), and were further supported by observed increases in dissolved organic nitrogen (DON) and dissolved total nitrogen (DTN) along with rising salinity, and vice versa.

1. Introduction

The Tibetan Plateau, as the “water tower of Asia”, has a lake ecosystem that demonstrates exceptional environmental sensitivity [1,2]. In recent years, due to global warming, melting glaciers, and increasing precipitation, lakes in this region have expanded significantly and showed a decreasing trend of salinity [3,4,5]. In the meantime, the nitrogen cycle process is being influenced by biological and geological chemical adjustments; salinity is a key factor and has influential impacts on the nitrogen transformation pathway, microbial community structure, and endogenous release process [6,7].

Studies indicate that many lakes on the Tibetan Plateau have experienced extensive declines in salinity over the past few decades [3,4,5,8,9]. The main driving factor includes annual precipitation [4], glacier recession and melting water increase [10], surface runoff increase, and lake expansion [11]; these environmental changes act together to impact the water balance and dilution effect of dissolved ions in lakes, which induced the salinity decrease in general [3,5].

The nitrogen cycle includes ammoniation, nitrification, denitrification, dissimilatory nitrate reduction (DNRA), and anaerobic ammonia oxidation (anammox) [12,13,14]. In Tibetan Plateau lakes, salt reduction can promote autotrophic microbial diversity and activity [6], while high salinity and alkalinity may inhibit the expression of denitrification genes [15].

The nitrogen cycle encompasses the processes of nitrogen enrichment, transformation, and release, and is closely related to the overall health of the ecological environment [9,16,17,18]. Sediments represent a crucial reservoir of the nitrogen cycle and serve as an essential source of nitrogen release [19]. Sediment nitrogen release potentials are sensitive to certain environmental factors, such as temperature, salinity, and redox statues [20,21,22,23,24]. An increase in salinity results in a decreased reduction in nitrates and nitrites to nitrogen gas (N₂), thereby enhancing the nitrogen retention capacity [25]. Lake salinity rising inhibits denitrification in general. Saltwater pulses significantly inhibit denitrifying enzyme activity in freshwater peat soils as well [15,26]. The DNRA process may be less affected by salinity changes [27].

While numerous studies have explored the impact of salinity on the nitrogen cycle, experimental confirmation regarding the differentiation of nitrogen cycle pathways under varying salinity gradients in plateau lakes remains limited. Specifically, there is a scarcity of comparative investigations involving multi-point, long-term monitoring and laboratory validation in lakes with salinity gradients. Such comprehensive assessments are crucial for understanding the response of the nitrogen cycle and ecological risks associated with salinity changes in these environments.

To address the knowledge deficit regarding nitrogen cycling in plateau lakes, Pangong Tso (China) (Figure 1), a typical example of a plateau lake exhibiting an east-to-west gradient from freshwater to saline conditions, is assessed using controlled laboratory experiments under different salinity conditions, and the rates and pathways of nitrogen transformation in sediments are quantified. Simultaneously, water samples were collected from 120 locations, including overlying water and surface water, while monitoring salinity, NH₄⁺-N, TN, NO₃⁻-N, and NO₂⁻-N indices. Structural equation modeling (SEM) and piecewise functions were used in conjunction with experimental data, and regression analysis employing the mixed-effects model (MEM) was performed on the monitoring outcomes.

2. Materials and Methods

2.1. Experimental Data Modeling Approach

2.1.1. Experimental Design

Via the surface sediment simulation method, sediment samples were collected from 12 sampling points (Figure 2) in Pangong Tso in May, July, and September of 2023, and were transported to the laboratory where the migration and transformation of contaminants in the sediment were simulated. In addition to the salinity of Pangong Tso, different gradients were set in order to study the effects of changes in water mineralization on the transport and transformation of NO₃⁻-N, NO₂⁻-N, TN, NH₄⁺-N, and salinity at the water–sediment interface.

(1)

Sediment pretreatment protocol

The mud samples were sealed in polyethylene containers and kept cold until use. Prior to the experiment, the samples were brought to room temperature and then adjusted to the target experimental temperature. Visible impurities (e.g., plant debris, stones) were removed using a 1 mm nylon sieve. The sediments were then thoroughly homogenized by manual stirring with a sterile plastic spatula to ensure consistency. Various indicators were measured for use as the initial base values before the experiment started.

(2)

Setting up the experimental groups

For sampling, three simulation experiments were conducted in June, August, and October. The mud used for the experiments was a substrate collected in situ from 12 spots, and ultrapure water was used for the experiments. Three batches of experiments were conducted.

(3)

Haline gradient design

According to the salinity monitoring results obtained from 120 points in the Pangong Tso basin from May to September, the salinity range of the experimental gradient was appropriately adjusted to cover the existing salinity range of the water in Pangong Tso. Therefore, the haline gradient of the three batches of 12-point groups in the experiment was set to 0‰, 3‰, 6‰, 9‰, and 12‰, with different salinity experimental solutions using the “coarse sea salt (artificial sea salt) + ultrapure water mode” configuration. The artificial seawater was prepared by dissolving analytical-grade NaCl in ultrapure water to achieve the desired salinity gradients. The experimental apparatus was a cylindrical glass container with a volume of 1~2 L. The cylindrical glass container had an inner diameter of about 10 cm and a height of about 20 cm. And when the experimental sample was configured, ultrapure water and the substrate were added according to the volume ratio of 4:1, and the substrate distribution area was shaded.

(4)

Measurement indicators

In the experiment, five indices were measured: NO₃⁻-N, NO₂⁻-N, TN, NH₄⁺-N, and salinity. In water quality analysis, to assess nitrogen NO₃⁻-N, ultraviolet (UV) spectrophotometry is generally employed to determine concentrations by measuring absorbance at a specific wavelength with a UV–visible spectrophotometer [28,29]. NO₂⁻-N was determined through its characteristic of reacting with azo dyes to form colored complexes, and these were quantitatively analyzed by measuring absorbance at 540 nm using a spectrophotometer. TN was determined using the chemical digestion spectrophotometric method, where the nitrogen compounds in the samples were first converted to NO₃⁻-N under acidic conditions and were then calorimetrically determined using a UV spectrophotometer. NH₄⁺-N was determined through its reaction with salicylic acid under specific acid–base conditions, resulting in the formation of a yellow complex, and then was calorimetrically determined using a spectrophotometer, which is especially suitable for the detection of low concentrations of NH₄⁺-N [18]. NO₂⁻-N quantification was conducted via spectrophotometry, following the guidelines of Chinese National Standard GB 7493-87 [30]. TN quantification was carried out using the alkaline potassium persulfate digestion UV spectrophotometric method, in strict adherence to Chinese National Standard HJ 636-2012 [31]. For the determination of salinity, the electrical conductivity (EC) method was utilized, where a conductivity meter measures the EC of the water sample and then estimates its salinity. A Shanghai INESA (Leici) DDS-307A conductivity meter was used for salinity detection.

(5)

Measurement time and requirements

The measurement times were 12 h, 1 d, 2 d, 4 d, 6 d, and 8 d after the beginning of the experiment. For each sample, a syringe or pipette gun was used to extract the appropriate amount of overlying water from the mud–water interface to enable the measurement of indicators. After sampling was completed, ultrapure water was slowly injected along the container wall until the original volume was reached, testing and adjusting the salinity of the overlying water, so that the salinity of the experimental groups was always maintained within the gradient value design.

2.1.2. Analytical Indices

At each time point, an appropriate volume of the surface water was extracted from the water–sediment interface using a syringe or pipette for index determination. After sampling, ultrapure water was meticulously injected along the container wall to restore the original volume. The salinity of the surface water was then tested and adjusted to ensure that the salinity of each experimental group remained at the designed gradient value. In instances where values fell below the detection limit, they were replaced with a small number, i.e., 0.0001.

2.1.3. Data Analysis Methods for Experimental Data

(1)

Experimental data fitting and analysis

Experimental data were first fitted using a first-order kinetic equation [22]:

$$C_t=C_0+k\cdot t$$

This was to determine pollutant release rates, where${ C}_t$ and $C_0$ are the concentrations at time t and initial time, respectively (both in mg/L), t is time (hours), and k is the release rate constant (in mg·L⁻¹·d⁻¹). For each haline gradient (0‰, 3‰, 6‰, 9‰, and 12‰) and each of the nine pollutant indicators, linear regression analysis via the least squares method was employed to calculate k-values and evaluate model goodness of fit (R² > 0.8 and p < 0.05 considered valid) [23]. When identifying threshold effects, linear relationships in the panel data were explored in order to discover whether a turning point threshold existed in the analysis of the effect of salinity on the rate of conversion of different nitrogen forms.

(2)

Model Integration and Analysis

Based on the above responses, this can be further refined into a path diagram for SEM and analyzed using the semopy package in Python 3.0. To investigate the coupled effects of salinity, time, and experimental batch on contaminant release, a three-way ANOVA was performed using Stata 17.0 with the following model equation:

$$aov(concentration salinity\times time+batch)$$

The significance of the main effects and interactions were tested (α = 0.05). Stata 17.0 was used for mathematical model analysis and segmented linear regression, and Origin was used for data visualization.

2.2. Monitoring Data Modeling Approach

2.2.1. Monitoring Design

Surface water and overlying water samples were obtained from 120 locations (Figure 2), during the months of May, July, and September. The measurement methods are the same as Section 2.1.1 (4), following the national standards of China [30,31,32,33,34].

2.2.2. Data Analysis Methods for Monitoring Data

A comprehensive data analysis was performed. First, descriptive statistics were applied to the monitoring data to characterize their distribution, and boxplots were generated to visualize variability and identify potential outliers. Second, correlation analysis (CA) was conducted to evaluate the relationships between salinity and various nitrogen species (e.g., TN, NO₃⁻-N, NH₄⁺-N) in both surface and overlying waters. Finally, a mixed-effects regression model (MEM) was constructed in Stata 17.0, with salinity as the fixed-effects independent variable and nitrogen species as dependent variables. Random effects were included to account for heterogeneity across water layers, water types, and sampling months. The model results were subsequently compared with those derived from controlled experimental conditions to ensure consistency and robustness.

3. Results and Discussion

3.1. Experimental Data Modeling

3.1.1. Significance Test of Correlation Coefficient and ANOVA

In terms of variance effects and system stability, the variance estimates of all variables were significant (p < 0.05), indicating that the interaction between salinity and nitrogen forms was statistically robust and that the dynamics of the whole system were regulated by a variety of factors (e.g., microbial activity and redox conditions). In addition, the weak positive effect of salinity on ammonia and other forms of nitrogen (coefficient < 0.01) suggests that a long-term reduction in salinity may trigger nonlinear ecological risks (

Table 1. Significance test of correlation coefficients.

	NO3−-N	NO2−-N	TN	NH4+-N	Other_N	Salinity
NO3−-N	1
NO2−-N	0.018	1
TN	0.146	0.074	1
NH4+-N	−0.044	−0.056	0.634	1
other_N	−0.074	−0.084	0.869	0.272	1
salinity	0.091	−0.055	0.171	0.090	0.154	1

).

The ANOVA results indicated that the salinity, time, and batch factors have significant effects on the response variable, and these effects on factor concentrations can be examined separately without concern for confounding via interactions (

Table 2. Results of significance tests for interactions (ANOVA).

	Partial SS	df	MS	F	Prob > F
Model	7.474	89	0.084	3.6	0
salinity_~c	0.254	4	0.064	2.73	0.028
time_fac	1.247	5	0.249	10.7	0
salinity_~c#time_fac	0.19	20	0.01	0.41	0.99
batch_fac	4.043	2	2.022	86.74	0
salinity_~c#batch_fac	0.33	8	0.041	1.77	0.079
time_fac#batch_fac	1.095	10	0.109	4.7	0
salinity_~c#time_fac#batch_fac	0.314	40	0.008	0.34	1
Residual	23.076	990	0.023
Total	30.551	1079	0.028

Note: The values in the “Total” row are the sums of the respective columns from all sources of variation. The abbreviations are as follows: Partial SS = partial sum of squares; df = degrees of freedom; MS = mean square; F = F statistic; Prob > F = p-value.

).

3.1.2. Structural Equation Modeling (SEM)

In this study, direct and varying effects among salinity, NH₄⁺-N, NO₂⁻-N, NO₃⁻-N, and other forms of nitrogen (other_N) were analyzed using structural equation modeling (SEM). The variance estimates for all variables were statistically significant (p < 0.05), indicating there were significant interactions between these variables.

Salinity has different effects on the transport and transformation of different forms of nitrogen. The figure above shows statistically significant positive correlations between salinity and both NH₄⁺-N and other forms of nitrogen (other_N). Specifically, for every 1 unit increase in salinity, the level of ammonia nitrogen increases by 0.004 units, while the increase in other_N was more significant (correlation coefficient of 0.006). In contrast, salinity had a significant inhibitory effect on NO₂⁻-N with a correlation coefficient of −0.0038, suggesting that an increase in salinity may lead to a decrease in NO₂⁻-N concentration.

In terms of variance effects, the variance estimates for all variables were statistically significant (p < 0.05). Salinity (salinity) had a slight positive but statistically significant effect on NH₄⁺-N with a correlation coefficient of 0.003964, indicating that the concentration of NH₄⁺-N increased slightly with increasing salinity. Salinity had a slightly positive effect on other_N with a correlation coefficient of 0.006099. Salinity had a significant negative effect on NO₂⁻-N with a correlation coefficient of −0.003758, indicating that an increase in salinity may lead to a decrease in NO₂⁻-N concentration. NH₄⁺-N had a significant positive effect on other nitrogen (other_N), with a correlation coefficient of 0.45318, which is a relatively strong positive relationship, indicating that an increase in ammonia nitrogen significantly enhances the concentration of other forms of nitrogen. NO₂⁻-N had a significant negative effect on other nitrogen (other_N) with a correlation coefficient of −0.284583, indicating that an increase in NO₂⁻-N can lead to a decrease in other forms of nitrogen. NO₃⁻-N also had a significant negative effect on other_N, with a correlation coefficient of −0.262589, which also indicates that an increase in NO₃-N may lead to a decrease in other forms of nitrogen.

The direct and variance effects between salinity and different nitrogen forms were analyzed thoroughly using SEM (

Table 3. SEM analysis results.

lval	op	rval	Estimate
NH4+-N	~	salinity	0.004
NO2−-N	~	NH4+-N	−0.019
NO2−-N	~	salinity	−0.004
NO3−-N	~	NO2−-N	0.039
NO3−-N	~	salinity	0
other_N	~	NO3−-N	−0.263
other_N	~	NH4+-N	0.453
other_N	~	NO2−-N	−0.285
other_N	~~	salinity	0.006
NO2−-N	~~	NO2−-N	0.018
NO3−-N	~~	NO3−-N	0.009
NH4+-N	~~	NH4+-N	0.032
other_N	~~	other_N	0.083

Note:~ indicates a regression path, ~~ indicates a covariance or variance relationship.

, Figure 3), which revealed the multi-scale regulation mechanism of salinity in the nitrogen cycle. The results showed that the direct effects of salinity on nitrogen forms were mainly reflected in the following: Firstly, for every 1 unit increase in salinity, the concentration of NH₄⁺-N increased significantly by about 0.004 units (correlation coefficient of 0.003964), which can be attributed to the fact that salinity inhibited the activity of nitrifying microorganisms, thus slowing down the conversion process of ammonia nitrogen into nitrite and NO₃-N. Secondly, the positive effect of salinity on other forms of nitrogen (other_N) was more significant (correlation coefficient of 0.006099), suggesting that salinity may reduce the loss of gaseous nitrogen by inhibiting denitrification. In addition, salinity significantly decreased NO₂⁻-N concentration (correlation coefficient of −0.003758), which may be due to the rapid oxidation (to NO₃⁻-N) or reduction (to N₂O/N₂) of nitrite under highly saline conditions. In terms of interactions between different nitrogen forms, ammoniacal nitrogen had a significant positive effect on other nitrogen forms, with a significant increase of 0.453 units for every 1 unit increase in ammoniacal nitrogen concentration (correlation coefficient of 0.45318), emphasizing the position of ammoniacal nitrogen as a ‘core node’ in the nitrogen cycle, possibly through nitrification (NH₄⁺→NO₂⁻→NO₃⁻), thus driving the transformation of nitrogen forms. In addition, there was a strong negative effect (correlation coefficient of −0.284583) between NO₂⁻-N and other forms of nitrogen, which may be due to the rapid oxidation of nitrite to NO₃-N or its participation in the denitrification process, resulting in the depletion of other forms of nitrogen. NO₃⁻-N exhibited a significant negative feedback effect on other nitrogen forms (correlation coefficient of −0.262589), potentially inhibiting upstream nitrogen transformation.

3.1.3. Analysis of the Effect of Salinity on the Conversion Rate of the Nitrogen Cycle

A visualization of different nitrogen forms obtained from the different salinity levels at different time points is shown in Figure 4. The image shows various characteristics of the different nitrogen concentrations in the different forms.

Firstly, the TN concentration exhibited a consistent declining trend throughout the experiment, with salinity-dependent attenuation rates. Under freshwater conditions (0‰ salinity), the TN concentration decreased more slowly, whereas with increasing salinity, the decrease in the TN concentration accelerated, especially at a salinity of 12, where the TN concentration decreased more evidently in the late stage of the experiment.

Secondly, NO₂⁻-N concentrations were low at the beginning of the experiment but peaked at some salinity levels and then gradually decreased. At a salinity of 6, the NO₂⁻-N concentration increased rapidly to the highest value at the beginning of the experiment and then decreased sharply, while under other salinity scenarios, the concentration change was relatively smooth and the peak occurred at a lower time and magnitude.

Thirdly, the changes in NO₃⁻-N revealed complex salinity-mediated modulations. At a salinity of 0, the NO₃⁻-N concentration was relatively stable and low, while at a salinity of 12, the concentration showed a significant peak in the process of experiment, followed by a rapid decline. At other salinities, the magnitude of the change in NO₃⁻-N concentration was relatively small.

Fourthly, the NH₄⁺-N concentration was high at the beginning of the experiment and then gradually decreased, but the rate of decrease and the final stabilized concentration differed at different salinity levels. In the 0-salinity scenario, the ammonia nitrogen concentration decreased slowly and finally stabilized at a relatively high level, while as the salinity increased, the ammonia nitrogen concentration decreased faster, especially under conditions with salinities of 9 and 12. NH₄⁺-N concentration decreased more significantly in the late stage of the experiment and finally stabilized at a lower concentration level.

In order to describe the cyclic transformation rate of the different forms of nitrogen, the forms of nitrogen were first grouped according to the type of pollutant. For those forms that showed clear inflection points, segmented regression analysis was performed, and inflection points were set accordingly. Regression analysis was used to obtain the regression coefficients of each group and then to calculate the release rate constants (k-values) of the nitrogen forms. Based on these rate constants, the cyclic transformation rates of nitrogen forms at different salinities were further analyzed to determine the salinity thresholds. The specific analysis results of the changes in concentrations under different haline gradients are given below.

TN concentrations decreased over time, while salinity significantly affected the rate of decline. At a salinity of 0, TN decreased slowly because the pure water was unsuitable for the growth of associated bacteria. The absolute k-value of the TN concentration was highest at a salinity of 3‰, indicating a greater nitrogen loss to the atmosphere (

Table 4. Haline gradient segmented regression results (TN).

Haline Gradient (‰)	K-Value (First Segment)	R2 (First Segment)	K-Value (Second Segment)	R2 (Second Segment)	Turning Point (hours)
0	−0.002	0.595	−0.001	0.983	96
3	−0.013	1	−0.001	0.953	24
6	−0.005	0.497	−0.002	0.934	48
9	−0.004	0.890	−0.001	0.759	96
12	−0.004	0.767	−0.001	0.362	96

, Figure 5).

NO₃⁻-N and NO₂⁻-N were the two most critical intermediates. The change in NO₃⁻-N concentration consisted of an initial increase followed by a decrease at salinities of 0, 3, 9, and 12‰. The increase in NO₃⁻-N concentration at 12‰ salinity was characterized by the largest k-value. At this salinity, the higher NO₂⁻-N concentration in the water body facilitated the generation of NO₃⁻-N through the nitrification process. In the second phase at 12‰ salinity, the absolute value of k was the largest, indicating a faster reduction in NO₃⁻-N, with the highest NO₃⁻-N reduction rate in denitrification. At other salinities, the NO₃⁻-N concentration varied smoothly with lower peaks, suggesting that nitrite production and consumption reached equilibrium at low and intermediate salinities (

Table 5. Haline gradient segmented regression results (NO₃⁻-N).

Haline Gradient (‰)	K-Value (First Segment)	R2 (First Segment)	K-Value (Second Segment)	R2 (Second Segment)	Turning Point (hours)
0	0.0004	0.7706	−0.0001	0.9995	96
3	0.0006	1	−0.0001	0.6996	24
6	/	/	/	/	/
9	0.0004	0.6642	−0.001	1	144
12	0.0012	0.625	−0.0041	1	144

, Figure 6).

At 6‰ salinity, the NO₃⁻ concentration exhibited a distinct response, rapidly increasing to a peak at the start of the experiment, then sharply decreasing. This suggests a dynamic interplay between salinity inhibition and competition between NOB and ammonia-oxidizing bacteria (AOB). As the salinity increased to 12‰, AOB-dominated nitrification became predominant, while NOB-driven processes and anammox were inhibited. Instead, with fluctuations at 6‰ salinity and possibly more N₂O production at 12‰ salinity, the reaction shifted to the following:

$$NO₃⁻\to N₂O$$

The reaction that produced NO₂⁻-N is known as nitrification, where AOB oxidizes ammonia nitrogen to nitrite:

$$NH₄⁺+1.5O₂\to NO₂⁻+2H⁺+H₂O$$

NOB further oxidizes nitrite to NO₃⁻-N:

$$NO₂⁻+0.5O₂\to N{O}_{3^{-}}$$

NO₂⁻-N had the largest absolute k-value in the second segment at 3‰ salinity. The k-value in the first segment was the highest at 6‰ salinity, indicating the fastest increase in NO₂⁻-N concentration at this salinity. This observed peak in NO₂⁻-N concentration at around 6‰ salinity, followed by a sharp decline at higher salinities, may be explained by a trade-off between AOB tolerance and NOB inhibition. Specifically, AOB appear to maintain activity at moderate salinity levels, leading to nitrite accumulation, whereas elevated salinity levels increasingly inhibit NOB activity, reducing the oxidation of nitrite to nitrate and resulting in the observed pattern. At high salinities of 9‰ and 12‰, NO₂⁻-N concentrations were essentially below the detection limit. In contrast to other studies in wastewater treatment environments, where a significant decrease in AOB activity was found at salinities > 10‰, this experiment suggests that low salinities (9‰ and 12‰) may be sufficient to inactivate AOB and inhibit ammonia oxidation reactions in natural environments such as lake sediment discharges (

Table 6. Haline gradient segmented regression results (NO₂_N).

Haline Gradient (‰)	K-Value (First Segment)	R2 (First Segment)	K-Value (Second Segment)	R2 (Second Segment)	Turning Point (hours)
0	0.0003	0.9285	0	0.2904	96
3	0.0002	1	−0.0001	1	24
6	0.0077	1	−0.0038	1	24, 48
9	0	0	0	0	/
12	0	0	0	0	/

, Figure 7).

The NH₄⁺-N concentration first showed an increasing and then a decreasing trend, which was significantly influenced by salinity. The NH₄⁺-N concentration increased most rapidly at 6‰ salinity, while the decrease was fastest at 9‰ salinity, primarily due to nitrification or anammox processes (

Table 7. Haline gradient segmented regression results (NH₄⁺-N).

Haline Gradient (‰)	K-Value (First Segment)	R2 (First Segment)	K-Value (Second Segment)	R2 (Second Segment)	Turning Point (hours)
0	0.0014583	1	−0.0003359	0.9908	24
3	0.0017708	0.9258	−0.0007037	0.8661	48
6	0.0048843	1	−0.0005706	0.9632	24
9	0.0012897	0.8435	−0.0008478	0.8810	48
12	0.0006713	1	−0.0006605	0.9843	24

, Figure 8).

3.2. Monitoring Data Modeling

(1)

Descriptive Analysis Results

TN concentrations showed a more dispersed distribution. The central lake region exhibited the highest median TN level, while other regions maintained comparatively lower values. Notably, extreme outliers were detected in both the central lake and tributary zones, with values far exceeding the interquartile range, indicating the occurrence of unusually high nitrogen levels in localized areas (Figure 9a).

DTN was predominantly concentrated in the lower concentration range; however, the western lake region displayed a broader range of DTN values. Several extreme outliers were observed in this region, with values well beyond the upper quartile, making these anomalies particularly prominent (Figure 9b).

NO₃⁻-N exhibited a slightly right-skewed distribution, with the majority of samples falling within a low concentration range. The western lake region showed slightly elevated NO₃⁻-N levels. Extreme outliers were also present in the central lake and tributary zones, indicating abnormally high nitrate concentrations in some locations (Figure 9c).

NO₂⁻-N concentrations were generally low, with most data points clustered at the lower end of the distribution. Nevertheless, the central and eastern lake regions displayed higher median values. A few extreme outliers were recorded in the tributary estuary and central lake areas, suggesting localized nitrite enrichment (Figure 9d).

NH₄⁺-N concentrations were more dispersed in the tributary and western lake regions, while other regions showed a tighter concentration distribution. In particular, the central lake region exhibited distinct outliers within the upper quartile range, reflecting the presence of exceptionally high ammonium levels in certain areas (Figure 9e).

The salinity levels across the sampling sites exhibited a relatively concentrated distribution, with no evident outliers, suggesting overall consistency in salinity concentrations among different regions. Among the water body types, the western lake region had the highest median salinity, followed by the tributary estuarine zones (Figure 9f).

(2)

Correlation Analysis (CA)

In the overlying water, the correlation between salinity and nitrogen species (especially TN and NO₃⁻-N) was significantly positive. Changes in salinity may affect the distribution of nitrogen. In surface water, the negative correlation between salinity and NO₃⁻-N suggests that higher salinity may depress NO₃⁻-N concentrations. The relationship between salinity and NH₄⁺-N and NO₂⁻-N is weak or complicated (

Table 8. Correlation analysis (CA) results for overlying water and surface water.

(overlying water)	salini~g	NH4+-N~g	TN_ove~g	NO3−-N_~g	NO2−-N_~g
salinity_o~g	1
NH4+-N_over~g	0.098 *	1
TN_overlying	0.366 ***	0.496 ***	1
NO3−-N_over~g	0.238 ***	0.056	0.479 ***	1
NO2−-N_over~g	0.161 ***	0.134 **	0.313 ***	−0.056	1
(surface water)	salini~g	NH4+-N_~g	TN_ove~g	NO3−-N_~g	NO2−-N_~g
salinity_o~g	1
NH4+-N_over~g	0.083	1
TN_overlying	0.186 ***	0.470 ***	1
NO3−-N_over~g	0.155 ***	0.463 ***	0.933 ***	1
NO2−-N_over~g	−0.147 ***	−0.086	0.160 ***	0.206 ***	1

Note: * p < 0.1;** p < 0.05; *** p < 0.01.

).

(3)

Mixed-effects Model Regression

In order to further analyze the relationship between salinity and nitrogen forms, regression analysis was carried out using the overlying water data with a mixed-effects model, and compared with the conclusions in the experimental environment.

The coefficient of estimation was 0.004, the standard error was 0, the t-value was 9.71, and the p-value was 0, indicating that salinity had a significant positive effect on NH₄⁺-N (p < 0.01). The chi-square value of the model was 94.225 and the p-value was 0.000, indicating that the model was significant. AIC is −2569.587, which indicates that the model fits well (

Table 9. Mixed-effects REML regression of NH₄⁺-N and salinity.

NH4+-N	Coef.	St. Err.	t-Value	p-Value	[95% Conf	Interval]	Sig
salinity	0.004	0	9.71	0	0.003	0.005	***
Constant	0.076	0.014	5.38	0	0.049	0.104	***
Constant	0.002	0.001	.b	.b	0.001	0.005
Constant	0.001	0	.b	.b	0	0.002
Constant	0	0	.b	.b	0	0.108
Constant	0.005	0	.b	.b	0.005	0.005
Mean dependent var		0.080		SD dependent var		0.086
Number of obs		1080		Chi-square		94.225
Prob > chi2		0.000		Akaike crit. (AIC)		−2569.587

Note: *** p < 0.01.

).

The coefficient of estimation of salinity and TN was 0.124, standard error was 0.002, t value was 60.32, and p-value was 0, indicating that salinity had a significant positive effect on TN (p < 0.01). The model chi-square value was 3638.515 and the p-value was 0.000, so the model was significant. The AIC was 456.921, indicating a good model fit (

Table 10. Mixed-effects REML regression of TN and salinity.

TN	Coef.	St.Err.	t-Value	p-Value	[95% Conf	Interval]	Sig
salinity	0.124	0.002	60.32	0	0.12	0.128	***
Constant	0.405	0.082	4.94	0	0.244	0.566	***
Constant	0.027	0.034	.b	.b	0.002	0.328
Constant	0.109	0.04	.b	.b	0.053	0.224
Constant	0.009	0.005	.b	.b	0.003	0.026
Constant	0.09	0.005	.b	.b	0.081	0.1
Mean dependent var		0.642		SD dependent var		0.895
Number of obs		720		Chi-square		3638.515
Prob > chi2		0.000		Akaike crit. (AIC)		456.921

Note: *** p < 0.01.

).

The coefficient of estimation for salinity and DON was 0.121, the standard error was 0.003, the t-value was 46.45, and the p-value was 0, indicating that salinity had a significant positive correlation with dissolved organic nitrogen (p < 0.01). The increase in salinity significantly promoted the release of dissolved organic nitrogen. The model chi-square value was 2157.259 and the p-value was 0.000, so the model was significant. The AIC was 180.343, indicating a good model fit (

Table 11. Mixed-effects REML regression of DON and salinity.

DON	Coef.	St. Err.	t-Value	p-Value	[95% Conf	Interval]	Sig
salinity	0.121	0.003	46.45	0	0.116	0.126	***
Constant	0.028	0.094	0.30	0.763	−0.156	0.213
Constant	0.05	0.044	.b	.b	0.009	0.277
Constant	0.052	2.406	.b	.b	0	8.399 × 1037
Constant	0.052	2.406	.b	.b	0	8.399 × 1037
Constant	0.071	0.006	.b	.b	0.06	0.082
Mean dependent var		0.287		SD dependent var		0.855
Number of obs		360		Chi-square		2157.259
Prob > chi2		0.000		Akaike crit. (AIC)		180.343

Note: *** p < 0.01.

).

The coefficient of estimation for salinity and DTN was 0.123, the standard error was 0.002, the t value was 60.81, and the p-value was 0, indicating that salinity had a significant positive correlation with DTN (p < 0.01). The increase in salinity significantly promoted the release of dissolved total nitrogen. The model chi-square value was 3697.983 and the p-value was 0.000, so the model was significant. The AIC was 389.081, indicating a good model fit (

Table 12. Mixed-effects REML regression of DTN and salinity.

dtn	Coef.	St.Err.	t-Value	p-Value	[95% Conf	Interval]	Sig
salinity	0.123	0.002	60.81	0	0.119	0.127	***
Constant	0.349	0.085	4.13	0	0.184	0.515	***
Constant	0.038	0.035	.b	.b	0.006	0.239
Constant	0.09	0.034	.b	.b	0.043	0.187
Constant	0.01	0.005	.b	.b	0.004	0.029
Constant	0.081	0.005	.b	.b	0.073	0.091
Mean dependent var		0.574		SD dependent var		0.860
Number of obs		720		Chi-square		3697.983
Prob > chi2		0.000		Akaike crit. (AIC)		389.081

Note: *** p < 0.01.

).

The coefficient of estimation for salinity and NO₃⁻-N was −0.002, the standard error was 0.001, the t-value was −2.40, and the p-value was 0.016, indicating a significant negative correlation between salinity and NO₃⁻-N (p < 0.05). The model chi-square value is 5.774 and the p-value is 0.016, which means the model is significant. The AIC was −1429.831, indicating a good model fit (

Table 13. Mixed-effects REML regression of NO₃⁻-N and salinity.

NO3−-N	Coef.	St.Err.	t-Value	p-Value	[95% Conf	Interval]	Sig
salinity	−0.002	0.001	−2.40	0.016	−0.003	0	**
Constant	0.166	0.036	4.66	0	0.096	0.235	***
Constant	0.009	0.006	.b	.b	0.002	0.034
Constant	0.004	0.004	.b	.b	0.001	0.026
Constant	0.021	0.004	.b	.b	0.014	0.03
Constant	0.012	0.001	.b	.b	0.011	0.013
Mean dependent var		0.132		SD dependent var		0.199
Number of obs		1079		Chi-square		5.774
Prob > chi2		0.016		Akaike crit. (AIC)		−1429.831

Note: *** p < 0.01, ** p < 0.05.

).

The coefficient of estimation for salinity and NO₂⁻-N was 0, standard error was 0.001, t value was 0.19, and p-value was 0.851, indicating that salinity had no significant effect on nitrite nitrogen (p > 0.05). The chi-square value of the model was 0.035 and the p-value was 0.851, indicating that the model fit was not ideal. The AIC was −845.374, indicating a poor model fit (

Table 14. Mixed-effects REML regression of NO₂⁻-N and salinity.

NO2−-N	Coef.	St.Err.	t-Value	p-Value	[95% Conf	Interval]	Sig
salinity	0	0.001	0.19	0.851	−0.001	0.002
Constant	0.04	0.015	2.68	0.007	0.011	0.069	***
Constant	0	0	.b	.b	0	3.492 × 1071
Constant	0.003	0.009	.b	.b	0	0.997
Constant	0.005	0.009	.b	.b	0	0.21
Constant	0.015	0.001	.b	.b	0.013	0.017
Mean dependent var		0.052		SD dependent var		0.162
Number of obs		720		Chi-square		0.035
Prob > chi2		0.851		Akaike crit. (AIC)		−845.374

Note: *** p < 0.01.

).

4. Conclusions

With the intensification of global warming and the melting of glaciers, the salinity of plateau lakes is showing a decreasing trend [3,5]. This change in salinity may have significant impacts on the release and cycling of nitrogen in plateau lakes. According to the experimental data in our research, the correlation analysis confirmed the significant impact of salinity on nitrogen release and cycling in plateau lake sediments. ANOVA results further validated the robustness of this conclusion, indicating that the influence of salinity changes on sediment nitrogen release and cycling is not affected by water mineralization, experimental duration, or salinity levels. SEM analysis and segmented regression revealed the overall influence trends between salinity and nitrogen-related products. NH₄⁺-N acted as a central driver of nitrogen transformation through nitrification, while NO₂⁻-N and NO₃⁻-N exhibited competitive inhibitory effects on other nitrogen forms. A previous study showed that NOB is more sensitive to high salinity [35], and have been found to be most abundant at 8‰ salinity, where almost all NO₂⁻-N is oxidized to NO₃⁻-N [36]. This is partly in companion to the rapid approach of NO₂⁻-N concentrations to zero at salinity 9‰ and 12‰ in this experiment.

Based on the field monitoring data, the increase in salinity significantly promoted the increase in NH₄⁺-N and TN concentrations, a finding supported by the MEM analysis. It was highly consistent with the experimental results and SEM. This indicates that the experimental findings have extrapolation and can partly explain the nitrogen response mechanism in the actual environment. Similar descriptions and analyses have been made in other studies, such as inhibition of bacterial community-mediated nitrogen cycling in soils by high salinity [37] and rapid decrease in nitrogenase activity at elevated salinity, suggesting that salinity affects nitrogenase functional complexes and thus increases nitrogen release from sediments [38]. DON and DTN were monitored and analyzed only by MEM, and both results showed that salinity had a significant positive effect on their concentrations (p < 0.01), further supporting the conclusion that increased salinity drives the accumulation of dissolved nitrogen.