Spatio-Temporal Variations of Discharge and Sediment in Rivers Flowing into the Anzali Lagoon

Abstract

In the last few years, trend identification has become an important issue in hydrological time-series analyses; it is also a difficult task, due to the variety of models and the impact of climate change on the river flow regime. Due to the vital importance of the Anzali Lagoon to the environment of the region, and the threat to its health caused by the volume or amount of inlet sediments, we decided to study the changes in flow and sediment in the rivers flowing into the Anzali Lagoon. For the present study, the long-term monthly, seasonal, and annual sediment and discharge data of seven stations were obtained during the period 1985–2019. According to the available information, the trend of sediment load variation was investigated at different time scales. In this study, the Mann–Kendall statistical test, the double-mass curve, and performance fitting were used to assess the seasonal and annual trends in sediment and river flow. The results showed that at Aghamahale station, the low relationship between discharge and sediment compared with that at other stations was due to the low slope and constant water of the Behmbar River, which caused the sediments to settle and decreased their carrying rate. Moreover, Nokhaleh station had the largest share of sediment transfer to the lagoon during 2002–2012. Sediment details also show that the highest amount of sediment in all stations occurred in non-crop seasons—i.e., from October to January—and was directly dependent on the amount of rainfall in these areas. The results of the sediment analysis also indicate that the discharge and the subsequent sediment loads from upstream to downstream were high over the summer. Furthermore, the rivers downstream demonstrated springtime peaks in the sediment loads and discharge, probably owing to snow melting.

1. Introduction

Erosion and sediment studies are very important in watershed management. The rate of erosion and the amount of sedimentation vary in each watershed because they are affected by various factors, such as climate, topography, and human activities [1]. Human-induced environmental changes on a global scale have increased the activity of geomorphic processes in many parts of the world. For example, exploitation of vegetation intensifies soil erosion, and large volumes of sediment enter the rivers. Annually, between 20 and 52 billion tons of sediment is transported by rivers and settles in stagnant water bodies worldwide [2]. Compared with the rest of the world’s rivers, rivers in Iran carry a higher amount of sediment due to climatic, hydrological, and geological conditions and excessive pressure on watersheds, which always lead to extensive damage [3].

Determining the rate of erosion and sedimentation is crucial for prioritizing the management of the watersheds. For this reason, estimating discharge and sediment in watersheds—especially ungauged ones—is very important.

Awareness of the sedimentation status and the amount of sediment output of a watershed provides an understanding of the erosion phenomenon and its consequences [4,5,6]. Changes in sediment transportation and its volume in rivers are often due to the changes in discharge. In other words, it can be inferred that the relationship between changes in discharge and suspended sediment load can be used in calculations related to the sediment load in rivers and dam reservoirs, as well as sediment issues in mathematical models and watershed management problems. Increasing the accuracy of these relationships can improve simulations of sediment transport phenomena [7,8]. Additionally, the suspended sediment load undergoes temporal and spatial changes. A large volume of sediment can be transferred during a single storm through the variability of the suspended sediment load due to discharge changes. Therefore, despite the difficulty and expenses of regular and frequent sampling of discharge and sediment (flood events), estimating the annual sediment load through data collection requires a more accurate estimation of the sediment [9].

The suspended load estimation method, which is based on measuring the suspended sediment concentration and discharge rate, is a method that requires continuous measurement, and is usually only possible for permanent rivers [10]. In this regard, the sediment computation equation is commonly used to estimate the sediment transport of rivers [11]. The sediment computation equation was developed by Fleming [12] for ungauged watersheds using the Walling [13] and Asselman [14] hypotheses for examining the statistics related to the suspended sediment of 250 rivers around the world. Based on the work of Robert [15], estimating the deposition using this equation always has some errors. However, this method provides a correct understanding of spatial and temporal changes in sediment, leading to realistic decisions about erosion and sediment behavior in the planning and implementation of watershed projects.

The trends in sediment loads have changed in most rivers, which could be attributed to many reasons, including the conversion of forest lands to agricultural farms, the increase in population density, the development of water resources, the construction of dam structures, climate change, and human activities [15,16,17,18,19,20,21,22,23].

Incorrect agricultural practices, high annual precipitation, and land-use conversions in northern Iran can be considered to be significant environmental contributors that accelerate soil erosion in this region [24]. In this case, the sediment yield caused by soil erosion can directly affect the plain areas such as Anzali Lagoon, which is located on the southern shore of the Caspian Sea in Iran. This lagoon faces many environmental problems; erosion is one of them, due to its proximity to the sea and the discharge of many rivers. On the other hand, due to the land use in different parts of this watershed, it is not safe from the harmful effects of river sediments that cause severe changes in its morphology [25,26].

In Anzali Lagoon and its watershed, many attempts have been made to assess the parameters related to the water quality of the lagoon [27,28], or to create chronologies of the sediment accumulated in the lagoon [29,30]. However, there has been no study of the sedimentation rate or an assessment of the effects of the major rivers’ discharge and erosion at the seven selected stations in the study area.

Studies on suspended sediment in Iranian rivers are very difficult due to the scattering or lack of data. Although studies conducted in this field have presented valuable results, they are insufficient, and more studies need to be conducted on the variations in the suspended sediment in order to achieve sustainable river management [3,30,31,32]. This requires us to estimate the past and present loads of suspended sediment in order to assess the amount of sediment and the rate of degradation in the water quality of rivers leading to lagoons. Since there are no clear data related to the discharge and sediment characteristics of the main-stream discharge into Anzali Lagoon, which can be attributed to insufficient sediment data, this study aimed to investigate the spatio-temporal variability in the discharge and the sediment loads discharged along the river from 1985 to 2019.

2. Materials and Methods

2.1. Study Area

The Anzali Lagoon watershed is situated along the coast of the Caspian Sea, between approximately 36°55′ and 37°32′ N and 48°45′ and 49°42′ E, in Guilan Province, northern Iran. The maximum elevation of the watershed is ~3105 m, while the elevation of the Caspian Sea’s coast is around −25 m. The watershed of the Anzali Lagoon is bordered by the Sefidrud River, the Alborz Mountains, and the Caspian Sea to the east, the southwest, and the north, respectively; it is divided into plains and mountainous landforms, with slopes of less than 1% and more than 25%, respectively. The altitude is below 100 m in the paddy area and up to 2500–3000 m in the mountainous region.

The area of the Anzali Plain is approximately 1200–2400 km² (width = 60 km, length = 20–40 km), while the mountainous area is ~1750 km² (width = 70 km, length = 25 km) (Figure 1). The Anzali Lagoon area is located in the Gorgan–Rasht geological zone of the Alborz Mountains in terms of zoning. This zone is a part of the North Alborz sedimentary region in the Middle and Upper Jurassic in terms of sedimentation. The high thickness of destructive and shallow Neogene–Quaternary sediments is one of the most notable features of this region, which was caused by water and mechanical erosion of the Alborz Formation.

Because of the geographical location of the lagoon on the coast of the Caspian Sea, it is constantly threatened by pollutants originating from various sources, and it is strongly affected by human activities such as industry, shipping, agriculture, Caspian oil operations, and the tourism industry. Advancement of coastal plant communities and shallow water areas towards the center of the lagoon, as well as the growth and development of plants, have led to the drying of the major parts of the lagoon—especially in the eastern watersheds and the Behmbar River (the presence of willow and alder trees and other types of shrubs on the margins of reeds on both sides). The transformation of plant communities in the region of Handkhaleh and the land development of the western watershed are visible. On the other hand, vegetation and forests in the upstream watersheds have been destroyed, and many sediments have entered the lagoon; consequently, its depth has been greatly reduced.

2.2. Data Collection

There are 23 hydrometric stations in the study area. Based on the length of the statistical period and the completeness of the data of daily discharge at the stations, seven hydrometric stations were selected, which are located on the rivers in different parts of the watershed. However, the length of the statistical period must be the same in all stations in order to make a correct judgment of the sediment load status, so two criteria were considered in the selection of these reference stations: First, the older stations with more than 20 years of sediment and discharge data were identified; thus, the number of stations was reduced to 17 stations. Secondly, on the rivers with several stations, stations located downstream of the rivers flowing into the lagoon were selected in order to provide more comprehensive information on the sedimentation conditions of the area. Thus, the number of stations was reduced to seven stations. The seven hydrometric stations and their computational index values over 29 years are presented in

Table 1. The characteristics of the rivers and representative stations in the Anzali Lagoon during 1985–2019 (data courtesy of Guilan Regional Water Company).

Rivers	Stations	Longitude	Latitude	Elevation (m)	Average Discharge (m3/s)	Average Sediment Load (t/day)
Behmbar	Aghamahale	343,802	4,147,503	−15	2.9	31.9
Masoulde Rudkhan	Chemseghal	355,172	4,137,526	−22	6.5	265.01
Kolsar	Kolsar	351,742	4,140,028	−23	8.12	118.2
Shakhrez	Laksar	360,196	4,135,259	−20	14.8	332.8
Pasikhan	Nokhaleh	362,962	4,134,794	−20	27.9	874.8
Chafrud	Rudbarsara	331,846	4,152,616	−20	2.6	88.5
Siahrud	Golsar	374,668	4,128,805	−20	6.44	263.57

.

There are several methods for estimating the suspended sediment load, most of which consider the statistics of different equations. Based on the available data in this study, the power relationship between discharge and sediment—which is called the sediment ratio curve—was used. In practice, according to the corresponding discharge and sediment data, both datasets were transferred to the logarithmic coordinate plane, and the line of best fit was passed through them based on the least squares method. Finally, the discharge–sediment diagram was obtained with the relevant equation and its correlation coefficient for the study area, and an equation was established between the two variables (Equation (1)):

$$Q_s=a\times Q_w^b$$

(1)

where Q_s is the sediment rate (t/day), Q_w denotes the discharge rate (m³/s), and a and b are the model coefficients showing the distance between the intersection of the line of best fit with the vertical axis and the origin, and the slope of the line of best fit, respectively.

Since, in this research, the ultimate goal was to estimate the sediment load transferred from the rivers of the Anzali Lagoon watershed into the lagoon, in line with the objectives of this study, the regression relationships of the discharge variables and the sedimentation rate, as well as the relationship between discharge and sediment, were investigated, and sediment measurement curves were plotted. The procedure consisted of several basic steps: In the first step, the discharge and sediment load data were arranged in order of year. In the second step, the water discharge data and the corresponding sediment load were evaluated by a graphical method (R²) and statistical indices such as the root-mean-square error (RMSE) [33], the normalized root-mean-square error (nRMSE) [34], the mean error (MAE) [35], and model efficiency (EF) [36]. For the graphical method, both datasets were transferred to the coordinate planes. In the next step, the line of best fit was calculated by four methods (linear, power, logarithmic, and exponential), and the fitting performance among the observed relationships was calculated. For statistical indices, the simulated and observed Q_s values were analyzed.

2.3. Data Analysis

2.3.1. Mann–Kendall Test

The Mann–Kendall (MK) test was used to analyze the trends in the discharge and sediment load [37,38]. The nonparametric Mann–Kendall test was adopted for the identification of trends in the time-series data. This compares the relative magnitudes of the sample data rather than the data values. The major benefit of this test is that the data do not need to conform to any distribution [39,40]. Moreover, data reported as non-detects can be included in the dataset by allocating them a common value smaller than the smallest measured value. Based on the assumptions of the procedure for trends, one value of the data is considered per time unit. In the cases in which there are multiple data points per time unit, the median value is applied. The values are considered and evaluated as an ordered time series, and then each data point is compared with all of the following data values. When there is no trend, zero is assumed to be the initial value of the Mann–Kendall statistic (S). If a data value from one timepoint is lower than the data from the following timepoint, S will be increased by 1. Conversely, if the data value from a later timepoint is lower than a data value sampled earlier, then S will be decreased by 1. The final value of S can be obtained by the net result of all of the increments and decrements yielded. The Mann–Kendall trend analysis can be expressed as Equations (2) to (5).

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

(2)

Let X₁, X₂, …, X_n represent n data points, where X_j represents the data point at time j; the Mann–Kendall statistic (S) is given [37,38] as follows:

$$\mathrm{Sgn}=+1 if \left(X_j-X_i\right)>0, 0 if \left(X_j-X_i\right)=0,-1 if \left(X_j-X_i\right)<0$$

(3)

A very high positive value of S indicates an increasing trend, while a very low negative value shows a decreasing trend. However, in order to statistically determine the significance of the trend, it is crucial to determine the probability associated with S and the sample size n [41,42,43,44]. The variance of S can be obtained using Equation (4):

$${{\displaystyle \sum }}_{}^2=\frac{1}{18}\left[n\left(n-1\right)\left(2n+5\right)-{{\displaystyle \sum }}_{p=1}^g\left(t_p-1\right)\left(2t+5\right)\right]$$

(4)

where n is the number of data points, g is the number of tied groups (a tied group is a set of sample data with the same value), and t_p represents the number of data points in the pth group. The normalized test statistic Z_s is computed by Equation (5):

$$Z_s=\{\begin{array}{c}\begin{array}{cc}\left(S+1\right)/\sigma & for S<0\end{array}\\ \begin{array}{cc}0& for S=0\end{array}\\ \begin{array}{cc}\left(S-1\right)/\sigma & for S>0\end{array}\end{array}$$

(5)

The trend significance is measured by the test statistic Z_s. In fact, the null hypothesis, H₀, is tested by this analysis, which denotes the lack of a monotonic trend in the data: if |Z_s| is greater than Z, then the null hypothesis is invalid, and the trend is significant at the chosen level of significance (usually 5% with Z_0.025 = 1.96). The significance shows that there is an effective factor in the trend and it has not occurred randomly. The Mann–Kendall test has been identified as an excellent method for statistical analysis to detect trends in different applications [45].

2.3.2. Double-Mass Curve

To identify the changes in hydrological regimes caused by human activities, double-mass curves have been widely used. These represent the changes in the cumulative data related to one variable versus the cumulative data of another variable for a simultaneous period [46,47,48,49].

If the double-mass curve is a straight line for two variables, there will be no change in the proportionality between them, or they will be affected only by climatic variability. The effect of human activity on the turning point becomes apparent, and the hydrological regime is shown by the change in the slope of the double-mass curve [50,51].

According to the studies on land-use changes in the study area, the area of the Anzali Lagoon has decreased because of increased human activities during the past 10 years [52], which have transformed 65% of the total land use to agricultural land [52]. Therefore, the impacts of the rivers in the Anzali Lagoon watershed and of the land-use changes on sediment transport were evaluated through the double-mass curve related to the river discharge versus the sediment load. In this regard, it should be noted that climate change in a region does not affect the slope of the obtained curves, because it affects all stations in the region equally [52]; for this reason, it was not included in this part of the calculations.

3. Results

In

Table 2. Statistical values of the Mann–Kendall Z test for monthly discharge and sediment data of the Anzali Lagoon watershed.

Station	Parameter	Mar	Apr.	May	Jun.	Jul.	Aug.	Sep.	Nov.	Oct.	Dec.	Jan.	Feb.
Rudbarsara	Discharge	0.68	1.92	−1.04	1.16	−2.05 *	−0.18	0.61	−0.31	−0.4	0.6	−1.53	0.93
	Sediment	0.96	0.01	0.63	0.65	−3.19 **	−0.72	−0.76	0.17	−1.15	−1.29	−2.2 *	0.56
Aghamahale	Discharge	2.65 **	2.06 *	3.08 **	0.87	0.45	−0.06	0.65	−0.42	−1.17	−1.68	0.58	−0.75
	Sediment	1.92	0.93	2.2 *	−1.19	0.08	−0.48	−0.16	−0.44	−0.75	−1.14	−0.03	−0.9
Kolsar	Discharge	1.10	1.40	−2.53*	0.98	0.62	1.09	1.05	1.34	1.15	−0.05	0.56	−1.61
	Sediment	1.04	0.87	−3.43 **	−0.6	0.9	1.06	0.42	−0.47	0.12	−0.61	1.15	−1.33
Chemseghal	Discharge	−0.19	0	−0.19	1.86	−0.3	1.15	0.8	1.23	0.77	0.04	−1.07	−2.21 *
	Sediment	−0.37	−0.25	−0.77	1.65	−0.62	0.76	0.52	0.01	0.21	−0.25	0.83	−2.42 *
Golsar	Discharge	0.27	−0.15	0.3	1.36	−0.53	0.99	0.43	1.18	0.49	−0.18	−1.26	−2.2 *
	Sediment	0.28	0.25	−0.09	1.65	−0.62	1.03	0.16	0.01	−0.06	−0.53	−0.7	−2.39 *
Laksar	Discharge	0.22	2.42 *	0.09	−1.08	−1.28	−1.04	−1.1	−2.06 *	−3.29 **	−0.91	−2.11 *	−0.52
	Sediment	−0.03	2.28 *	−0.34	−2.14 *	−1.6	−1.65	−0.27	−2.31 *	−1.03	−1.01	−2.28 *	0.04
Nokhaleh	Discharge	−1.47	−0.06	−2.25 *	−0.71	−1.92	0.51	1.16	0.46	1.43	−0.14	−1.19	−1.55
	Sediment	−2.18 *	−0.03	−1.82	−1.15	−2.01 *	1.63	1.19	−0.08	1.16	−0.48	0.06	−1.58

* For a given variable, Z-statistic is significant at the level of 5%. ** For a given variable, Z-statistic is significant at the level of 1%.

and

Table 3. Statistical values of the Mann–Kendall Z test for seasonal and annual discharge and sediment data of the Anzali Lagoon watershed.

Station	Parameter	Annual	Spring	Summer	Fall	Winter
Rudbarsara	Discharge	−1.1	−0.19	−0.71	−0.28	−0.42
	Sediment	−3.2 **	0.5	−2.5 *	−1.63	−2.67 **
Aghamahale	Discharge	0.14	2.48	−0.82	−0.33	−0.93
	Sediment	−2.65 **	0.88	−2.3 *	−0.79	−0.96
Kolsar	Discharge	−0.64	−1.5	0.65	−0.37	−0.34
	Sediment	−0.64	−1.5	0.65	−0.37	−0.34
Chemseghal	Discharge	−0.71	−0.58	−0.34	1.47	−0.85
	Sediment	−1.45	−0.52	−0.85	−0.04	−0.79
Golsar	Discharge	−0.65	−0.37	−0.73	0.92	−1.14
	Sediment	−1.45	0.13	−0.82	−0.49	−1.02
Laksar	Discharge	−0.92	0.4	−1.57	−1.62	−1.5
	Sediment	−1.84	−0.12	−2.34 *	−1.5	−1.23
Nokhaleh	Discharge	0.08	−2.03 *	−0.76	1.52	−0.59
	Sediment	−1.38	−2.06 *	−1.8	0.99	−1.32

* For a given variable, Z-statistic is significant at the level of 5%. ** For a given variable, Z-statistic is significant at the level of 1%.

, the Mann–Kendall Z-statistic values for sediment variables are shown at monthly, seasonal, and annual time scales for the stations of Anzali Lagoon. These values indicate that out of 84 (7 stations and 12 monthly data series) data series, only one data series has a significant negative trend at a significance level of 10%, which is related to the Kolsar station in June, with a Z-statistic of −3.43. The rest of the data series related to this station have no significant trends, and there were no sudden negative or positive changes. No significant trends were observed in the annual and seasonal scales of this station’s data. Evaluation of the Z-values and figures shows that out of 84 series of monthly sediment data, an insignificant trend was observed in only 1 data series, while 83 data series had significant trends at a significance level of up to 1%. Out of these 83 series, 1 Z-statistic was significant at the 10% level, 4 statistics were significant at the 5% level, and 19 Z-statistics were significant at the 1% level. Out of 24 data series with a significant trend, 6 series showed a positive trend, which were approximately in the range of 2.06 to 3.08. Except for cluster-related sediment series, sudden changes from positive to negative, and vice versa, were observed in six time series related to Aghamahale, Chemseghal, Golsar, Laksar, and Nokhaleh, probably due to the sudden changes in sediment reported at these stations.

Figure 2A–G show the trend of changes in annual discharge for the river stations of Anzali Lagoon. As is clear from these graphs, the annual data of all stations had a negative trend; the largest trend was related to Rudbarsara station, which had a statistic of Z = −1.1. According to the graphs, it is not possible to make an explicit judgment about the increase or decrease in discharge in the watershed area of Anzali Lagoon; thus, more appropriate statistical analysis is necessary. On this basis, a statistical analysis of the Z-statistics was performed. However, the study of the Z-statistics for all 84 data series shows that no discharge data series was significant at the level of 10%, while at the significance levels of 5% and 1%, 9 and 1 monthly discharge data series were significant for Aghamahale station, respectively. This significant variation happened in April and June. The significant monthly discharge series at the level of 1% indicate the occurrence of seasonal floods for the Rudbarsara (August), Aghamahale (May), Kolsar (June), Laksar (May and October), and Nokhaleh (June) stations. Examination of the seasonal Z-statistics among the selected stations also showed that the Z-statistic of Aghamahale station is significant at the level of 2.48 in the spring. Seasonal analysis of the Z-statistics also showed that Nokhaleh station is significant at the level of 1% with a Z-statistic of −0.03 in spring. A seasonal overview of the selected stations in terms of the discharge and sediment series also showed that Nokhaleh station has a significant negative trend in spring. Spring Z-statistics related to discharge and sediment were obtained as −0.03 and −0.06, respectively, which were significant at the level of 1%.

The annual Z-statistic was not significant at any station for the discharge series, as was the case for the sediment series. Of course, the changes in the Z-statistic of annual sediment were significant at the level of 5%, and this statistic was estimated to be −3.2 and −2.65 at Rudbarsara and Aghamahale stations, respectively. Moreover, a study of the Z-values on an annual scale for different stations showed a change from positive to negative for the Laksar station, which could indicate sudden changes in the amount of sediment caused by discharge to this station.

As shown in Table 2 and Table 3, negative Z-values with confidence values lower than 0.01 show a significant decrease in discharge and sediment load in the upper river. For several months, absolute Z-values related to the sediment load greater than those related to the discharge were obtained, showing that the reduction in sediment load was more remarkable. The minimum absolute Z-value related to the discharge was obtained at the Nokhaleh station, while the maximum value was obtained at the Aghamahale station, which indicates that the reduction in discharge in the lower part of the Anzali Lagoon watershed was more notable than that in the other parts. Considering the water and sediment trapped by different land uses upstream from the Nokhaleh station, with less supplemental water along the river and greater water diversion, discharge decreased from upstream to downstream, and the rate of decrease was most notable at the downstream station (Aghamahale). Additionally, Nokhaleh station had the maximum absolute Z-value in terms of sediment load, which indicates that the clear or low-sediment-laden water released upstream scoured the riverbed. Sediment concentration was supplemented downstream by the addition of eroded bed materials to the water discharge; hence, the rate of increase upstream was higher than that downstream.

The annual sediment load data related to the seven gauging stations were recorded during 1985–2019. Unlike the river discharge data, suspended sediment data were not regularly included in these data. At each station, data collection related to the suspended sediment concentration was conducted with 24 repetitions per year, so data including both low and high rates of discharge were used to plot a sediment rating curve, which represented the sediment features for each year. The availability of daily discharge data, which were consistently measured, facilitated the estimation of daily sediment discharge using the sediment rate. The sediment rate curves, which show the relationship between the sediment and the daily discharge collected from the gauging stations, are presented in Figure 3.

A comparison of correlation values reveals that the correlation between the monthly sediment load and discharge was strong at Rudbarsara, Kolsar, Chemseghal, Golsar, Laksar, and Nokhaleh stations (R² > 0.75). However, the monthly sediment load at Aghamahale station had a poor relationship with the monthly discharge (R² = 0.59) in the watershed closer to Anzali Lagoon. Based on the power regression analysis, the relationships between the daily sediment load and monthly discharge for each station can be expressed as shown in

Table 4. Relationships between Q_s and Q_w and statistical indices.

Hydrometry Station	Polynomial Equation	R2 (%)	RMSE (t/d)	nRMSE (%)	MAE (%)	EF (%)
Rudbarsara	Qs = 1.80Qw1.77	0.7	6.37	0.95	0.21	−0.31
Aghamahale	Qs = 2.71Qw1.21	0.59	22.51	0.98	0.94	−16.3
Kolsar	Qs = 1.44Qw1.3	0.72	6.9	0.21	1.58	2.78
Chemseghal	Qs = 2.19Qw1.6	0.76	1.55	0.14	0	−0.11
Golsar	Qs = 2.33Qw1.56	0.77	6.71	0.16	0.03	−1.53
Laksar	Qs = 1.66Qw1.46	0.8	6.93	0.12	0.01	1.43
Nokhaleh	Qs = 0.90Qw1.58	0.79	9.11	0.17	0.02	1.49

. Moreover, the performance value of the model (EF) also indicates the accuracy of the data’s fit, and varies from infinitely negative in the worst case to 1 for a full data fit. The RMSE analysis of the sediment values showed that for the power equation of all stations, except Aghamahale, the values were in the acceptable range, ranging from 1.55 to 9.11 tons per day. Because the nRMSE level was less than 1, this model showed good performance for estimating the sediment load. Moreover, according to the classification by Jamieson et al. [53], an nRMSE less than 10% is in the excellent category, and an nRMSE between 10 and 20% is in a good category. Furthermore, the closer the EF, RMSE, and nRMSE values are to zero, the better the simulation model performs.

Based on the R² calculation, and as proven by several studies, the power equation was identified using the error test method as the best equation. The equations in Table 4 were used to estimate the missing suspended sediment data at Rudbarsara, Kolsar, Chemseghal, Golsar, Laksar, and Nokhaleh stations, but not Aghamahale. In addition, since only 59% of the data related to the suspended sediment could be explained by the Aghamahale station’s equation, variation in the sediment load was analyzed on the basis of the data obtained by RWC at this station.

At Aghamahale station, the weak relationship between discharge and sediment compared with those at other stations can be attributed to the low slope and static water of the Behmbar River (0.2%), resulting in the sediments of this river settling [51,52], so the sediment-carrying rate decreases.

An examination of the statistical index values in

Table 5. Function of the sedimentation curves of the stations, and the values of the statistical indices (R², RMSE, nRMSE, EF, and MAE).

Station	Statistical Indices	Polynomial	Linear	Exponential	Logarithmic
Rudbarsara	Function	Qs = −2.01Qw2 + 142.63Qw − 235.79	Qs = 72.25Qw − 100.8	Qs = 1.9e0.2Qw	Qs = 218.82ln(Qw) − 10.9
	RMSE (t/d)	180.62	96.69	12.35	67.68
	nRMSE (%)	−1294.63	584.24	443.53	1271.9
	EF (%)	167.49	47.99	0.78	23.52
	MAE (%)	20.23	−10.15	3.56	1.03
	R2	0.3	2	0.3	0.1
Aghamahale	Function	Qs = −0.13Qw2 + 21.24Qw − 13.92	Qs = 3.05Qw + 23.2	Qs = 2.9e0.04Qw	Qs = 34.972ln(Qw) + 25.9
	RMSE (t/d)	61.71	23.75	17.6	26.19
	nRMSE (%)	214.37	80.77	559.47	71.37
	EF (%)	0.84	0.96	1	0.97
	MAE (%)	2.1	5.28	−20.77	2.82
	R2	0.3	0.01	0.04	0.03
KOLSAR	Function	Qs = −0.29Qw2 + 25.4Qw − 76.5	Qs = 11.8Qw − 13.51	Qs = 3.94e0.11Qw	Qs = 84.221ln(Qw) − 36.8
	RMSE (t/d)	83.49	53.9	12.29	34.64
	nRMSE (%)	−109.13	−398.99	311.86	521.56
	EF (%)	7.35	7.67	2920.14	1.7
	MAE (%)	−32.31	−55.24	−165.09	16.79
	R2	0.21	0.1	0.3	0.11
Chemseghal	Function	Qs = −0.2Qw2 + 64.5Qw − 122.5	Qs = 53.14Qw − 79.05	Qs = 4.24e0.19Qw	Qs = 225.47ln(Qw) − 9.21
	RMSE (t/d)	81.15	65.93	8.96	93.26
	nRMSE (%)	−300.93	2447.12	148.28	198.81
	EF (%)	125.63	59.23	0.14	15.54
	MAE (%)	31.75	3.11	0.79	11.03
	R2	0.12	0.1	0.5	0.04
Golsar	Function	Qs = −0.27Qw2 + 64.3Qw − 120.76	Qs = 53.1Qw − 78.139	Qs = 4.02e0.20Qw	Qs = 203.35ln(Qw) + 23.1
	RMSE (t/d)	138.54	123.84	11.18	104.47
	nRMSE (%)	−114.72	−158.36	278.14	368.37
	EF (%)	9.27	7.63	0.45	3.22
	MAE (%)	−49.22	−63.39	3.53	−12.32
	R2	0.12	0.1	0.5	0.04
Laksar	Function	Qs = −0.21Qw2 + 55.6Qw − 359.8	Qs = 33.5Qw − 164.08	Qs = 9.51e0.07Qw	Qs = 332.53ln(Qw) − 317.92
	RMSE (t/d)	279.92	181.56	49.6	260.17
	nRMSE (%)	−107.74	−110.65	521.55	−163.48
	EF (%)	9.11	3.7	0.27	7.73
	MAE (%)	0.56	−8.4	−7.75	−190.14
	R2	0.37	0.3	0.4	0.14
Nokhaleh	Function	Qs = 0.4Qw2 + 0.97Qw + 1.7	Qs = 69.4Qw − 1064.6	Qs = 15.9e0.04Qw	Qs = 1099.1ln(Qw) − 2028.3
	RMSE (t/d)	79.72	740.02	53.71	1364.89
	nRMSE (%)	4689.48	−69.51	337.83	−110.58
	EF (%)	1.38	122.7	0.67	417.94
	MAE (%)	26.73	−536.09	−10.8	−1268.69
	R2	0.58	0.4	0.57	0.13

shows that the lowest values are related to the linear graphs; among these, Aghamahale station has the lowest correlation. Furthermore, a study of all of the R² values shows that the power functions have the highest R² values among all functions; thus, the correlation obtained from the power functions is higher, which is the basis of the calculations.

According to a comparison between Table 4 and Table 5, and according to the results of the proposed statistical indices, a power equation has the best fit compared with the other functions, and is more or less the opposite of the other functions in this study. This means that the other functions do not have a very acceptable ability to predict the amount of sediment corresponding to the discharge values. According to the USBR recommendations, according to the classification of the volumetric discharge of rivers, and if the data distribution status requires it, two or more lines can be passed through the measured data instead of one regression line, which is the line of best fit based on the least squares method. Two or more regression equations are used to calculate the long-term suspended load according to the different applications of volumetric passage. Fitted lines must have an acceptable correlation coefficient.

Figure 3A–G were plotted to investigate the coordination of changes in the cumulative discharge and annual sediment statistics for the studied stations during the current statistical period. Usually, this curve—which is the same as the double-mass curve of water discharge versus sediment—also showed that there were three distinct time phases in the data, with the second change point representing the logical changes arising from sudden changes in water discharge and sediment in the time series. This change in the discharge rate increased steadily at the Kolsar station, which may indicate human activity along the river, resulting in sudden changes in discharge and sediment volume.

The double-mass curves (DMCs) of the cumulative annual river discharge and the cumulative annual sediment load at each hydrological station are shown in Figure 4A–G. The impacts of human activities on sediment loads of rivers in the Anzali Lagoon watershed can be investigated using the DMCs. The plots related to the DMCs of stations relating to the pre- and post-construction periods showed relatively straight lines. In the other words, the proportion between river discharge and sediment load has been constant over the past 35 years [41].

The points on the coordinate system used to draw double-mass curves do not form a straight line but, on average, a straight line can cross between them with a slight deviation. These slight deviations change the slope of the curve. In general, double-mass curves reflect the changes in the two components under study, which is the basis for analyzing the change trend. Based on the results, the trend of curve-slope changes during the study period was not constant in any of the selected stations, and the amount of sediment was not proportional to the amount of discharge. In other words, the trends of change in discharge and sediment were not uniform, and the data had increasing or decreasing trends such that in the first few years, the data had no changes, meaning that the amount of sediment was proportional to the amount of discharge. However, since 2001, this trend has increased in six of the seven stations, which could be due to more sediment or discharge in this period. This result confirms that the sediment data are affected by other factors.

Without a sudden break in the DMC slope, the impacts of the construction of some structures on the sediment load were identified as non-significant in the middle of the watershed in which the stations were situated. However, according to the plots, a temporary decrease was indicated in the sediment flux in the middle of the watershed during 1994–2010 compared the prior period. Similarly, the DMCs related to the stations indicate that the changes in land use had no effect on the sediment flux in the middle reaches of the Anzali Lagoon watershed in 1987.

Consequently, there were no significant changes in the sediment loads caused by the construction of the structures along the river. Because of the small barrage, only a small amount of sediment might be trapped. Moreover, the increase in agricultural land owing to the construction of the barrage might have resulted in an increase in the sediment load. The sediment load of the Anzali Lagoon watershed can be significantly increased by the extreme natural events—especially severe flooding—while the decrease in sediment load in 1994–2010 in the middle of the watershed was likely a result of the low-discharge years.

The different colors in Figure 4 indicate spatial and temporal variations in the sediment and discharge data series. In fact, a break in the slope of the DMC indicates the positive or negative effect of watershed management [54]; for example, for the Behmbar and Golsar stations, at approximately 0.1 m³/sec, the slope is broken. For Rudbarsara, Kolsar, Chemseghal, Nokhaleh, and Laksar, this break happens in the range of 0.1–1 m³/sec, which is shown in blue. Every break in the dataset is defined by an equation. The orange color in the graphs represents another break in the slope—for Nokhaleh station, this happened at 100 m³/sec; for the rest, the slope broke at approximately 10 m³/sec. The third break, which is shown in gray, happened at more than 100 m³/sec for Nokhaleh station, while for the rest of the stations, it happened at approximately 10 m³/sec. At every breakpoint, the suspended sediment load could be calculated by the defined equation.

The amount of suspended sediment loads depends on several factors but, due to the limitations in the recorded data, the daily discharge was considered to be the best factor for estimating the sedimentation. From the discharge–sediment ratio obtained from Figure 3A–G, and with the daily discharge of the stations, the annual sediment yield was obtained for a 10-year period (2002–2012), which used a one-line sediment measurement curve for the calculations. The amounts of suspended sediment at Rudbarsara, Aghamahale, Chemseghal, Kolsar, Golsar, Laksar, and Nokhaleh stations were calculated as 32,788.9, 194,659, 97,471.3, 112,689.88, 326,223.7, and 648,864 tons, respectively, totaling 1,437,302 tons from all seven stations.

As can be seen, Nokhaleh station made the largest contribution to the sediment transfer to the lagoon. The sediment details also show that the highest amount of sediment occurred in non-crop seasons at all stations—i.e., from October to January—and was directly dependent on the amount of rainfall in these areas. It is also recommended to study the location of Nokhaleh station in order to be able to comment with more certainty on the increased sediment load at this station. The amount of rainfall, along with human activities in the vicinity of this station, may play a decisive role in increasing the sediment load at this station.

4. Conclusions

The sediment load and discharge have decreased significantly during the flood seasons since 1988 for all stations. Since 1990, changes have occurred in the natural distribution of discharge during flood seasons, and the discharge of non-flood seasons has dominated, so that its contribution to the annual discharge accounts for more than 54%. According to these results, sediment loads have also significantly decreased in the flood season—especially downstream.

A review of the JICA [25] report on land use showed that land use has changed from 1991 to 2019. The urban area is increasing with the development of the watershed. Moreover, the area of paddy fields decreased by ~100 km² from 1991 to 2002, which can be attributed to the increase in tree plantations in the plain areas. The rangeland (mountain grasses) and bare land areas have mainly varied due to the changes in climatic conditions over the years. There are also areas in which Populus spp. and alder trees have been planted in rangelands around the lagoon. On the eastern side of the lagoon, there are rangelands with an area of ~100 km². There are no considerable industrial areas around the lagoon. Anzali city is located in the northern part of the lagoon, around the outlet and along the Caspian Sea shoreline.

In general, according to the statistical analysis mentioned in the previous sections, changes in the time series of suspended sediment load in the area may be related to various land-use activities—the soil morphological characteristics of the study area, which have usually been very fragile over the years; aggression and other human activities; and rainfall reduction are among the parameters affecting the parameters of the sediment measurement curve, in agreement with the results of studies conducted by other researchers [20,48].

It can be concluded that at Aghamahale station, the weak relationship between discharge and sediment compared with those at other stations is due to the low slope of the Behmbar River, which results in continuously moist soil; consequently, most of the sediments of this river settle and their carrying rate decreases. Moreover, we can conclude that Nokhaleh station had the largest share of sediment transfer to the lagoon. The sediment details also show that the highest amount of sediment in all stations occurred in non-crop seasons—i.e., from October to January—which was directly dependent on the amount of rainfall in these areas. The amount of suspended sediment at Rudbarsara, Aghamahale, Chemseghal, Kolsar, Golsar, Laksar, and Nokhaleh stations was 32,788.9, 194,659, 97,471.3, 112,689.88, 326,223.7, and 648,864 tons, respectively, totaling 1,437,302 tons. A simultaneous study of sediment variables with a shorter time scale and sampling of the sub-watershed output, based on changes in land use and geological formations, and their relationships with other discharge components in the form of water formations and measuring rings, could be used in the analysis of hydrological conditions in the watershed and the effects of human activities on the changes in the amounts of discharge and sediment.

The low water discharge and sediment load occurred more frequently than full water discharge and a high sediment load, and the relationship between discharge and sediment was not always a direct relationship. Since most of the sediment is transferred at high discharges, it is recommended that sufficient sampling should be carried out during flood flows. The sediment load–peak discharge relationship can be used as a tool for sustainable water and soil management in estimating sediment loads.