Using Mixing Models and Tracers to Determine the Contribution of Land Use in Erosion and Sediment Yield (Case Study: Tange Bostanak Watershed, Fars Province)

IntroductionSediment yields effects are often termed on-site and off-site impacts, increasingly in watersheds (Palazon, Gaspar, Latorre, Blake and Navas, 2015). The sediment fingerprinting method has provided a direct and successful approach to quantify sources of sediment (Chen, Fang and Shi, 2016). Sediment fingerprinting approaches offer the potential to quantify the contribution of different sediment sources, evaluate catchment erosion dynamics, and develop management plans to tackle, among other problems, reservoir siltation (Palazon et al, 2015). Land use changes are crucial to understanding the evolution of sediment load and discharge dynamics (Wang, Chen, Fu and Lu, 2014). The present study aims to compare the results of several fingerprinting models, soil erosion features, and map and direct field measurements of soil erosion.
Study area:Our study was conducted in the Tange Bostanak catchment (30°16′ to 30°25′ N and 52°03′ to 52°13′ E), in the Southern Zagros mountains, 80 km Northwest of Shiraz, Iran.
Material and MethodsSampling and Data Collection
Potential sediment sources were identified by examining the main land use types within the study catchment, which are dominated by four main groups: rangelands, forests, moderate forests, farming and gardens. Forty three representative samples were collected from these potential sources at different locations within the study catchment. In order to remove bias associated with grain-size effects, only the 63 μm soil and sediment fraction, obtained by dry sieving, was taken for tracer analysis. Total concentrations of Ba, Cd, Co, Cr, Cu, Li, Mn, Ni, P, Si, Sr, Ti, C, N, Zn, 87Sr. 86Sr, 143Nd. 144Nd, Sr and Nd were measured by ICP-Mass after digestion of 3 gr of the soil samples with aqua regia (HCl–HNO3; 3:1) for 2 hours. In order to assess the validity of the analytical results, accuracy
and precision were calculated.
Source Discrimination/Discriminant Analysis
The statistical analysis employed to identify a composite fingerprint which is capable of discriminating between potential sources was done in two steps. The first step involved testing the discrimination of potential sources by the fingerprint properties using the Kruskal–Wallis test. Second, we applied stepwise multivariate discriminant function analysis (DA) to select the optimum subset composition of geochemical tracers for maximizing discrimination (Collins, Walling, Webb and King, 2010).
To identify outliers, was calculated the squared Mahalanobis distance, or the multivariate distance of each sample from the centroid of all samples belonging to a given category in the multidimensional space.
Sediment Sources Apportionment and Mixing Models
In geochemical tracing studies the relative contribution of source material to suspended sediment is usually estimated using a multivariate mixing model. The literature describes many different mathematical forms of mixing models. In all mixing models, the objective is to determine the source component proportions (x) in the suspended sediment samples by minimizing the errors. Five mixing models and their modifications to estimate contribution of sources were used.
(Slattery, Walden and Burt, 2000),
(Collins, Walling and Leeks, 1997),
(Devereux, Prestegaard, Needelman and Gellis, 2010),
(Collins et al., 2010),
(Motha, Wallbrink, Hairsine and Grayson, 2004)
where: ci = concentration of fingerprint property (i) in sediment samples; Sij = concentration of fingerprint property (i) in source category (j); X j = percentage contribution from source category (j); Z j = particle size correction factor for source category (j); Oj = organic matter content correction factor for source category (j); Wi = tracer discriminatory weighting or tracer specific weighting; SVji = weighting representing the within-source variability of fingerprint property (i) in source category (j); VARij = variance of the measured values of tracer i in source area j; mj = the total number of samples for an individual source; n = number of fingerprint properties; and m = number of sediment source categories. In sediment fingerprinting studies, to find the best optimum sediment contribution minimizing mixing model errors, Collins et al. (2012) proposed a revised modeling approach comparing the results of local optimization to determine the uncertainties with the following goodness of fit GOF equation and also used RMSE index to select the best mixing models.
Results and DiscussionSoil erosion and sediment yield are among the most destructive phenomena that cause a lot of damage in different regions. However, in order to combat this phenomenon and soil conservation projects, it is needed to be aware of the location of sediment sources in the region. Sediment fingerprinting technique based on geochemical tracers, organic, isotopic ratios, as well as useing various mixing models lead to the recognition and the contribution of different sediment sources in an area. In this study, using the optimum combination of organic and rare tracers to separate the different sources and then determine contribution of this erosion and sediment yield resource using Collins (1997), Collins Modified (2010), Motha (2004), Landwehr (2010), and Slattery (2000) models. In this case, to determine the best model GOF index and matching results with actual measurements of the sediments and erosion amount based on BLM model were used. The result were analyzed by discriminant analysis showed compounds of C, Cu Si, and Ti were considered as tracer and M Collins mixing model with indices GOF, 99.9 was selected as the best model. Measurement results showed that sedimentation rates of different land uses, pastures with relative importance of 3.4 is the priority in management of soil conservation. Mixing models showed the highest proportion in sediment and erosion basins on weak and degraded rangelands area is by 57.04 area percentage and 64.9 contribution percent in field measurements. Cultivations have minimal role in sediment yield of Tange Bostanak watershed. Results showed Motha model with a 0.924 correlation coefficients had the smallest difference with actual values of sediment contributions. The differences in the results of fingerprinting technique and field measurement techniques referred to their quantitative or qualitative approaches and not considering sediment delivery ratio concept and other sediment transition processes. It clearly confirmed the necessity of simultaneous use of all aforesaid techniques to get access to reliable results and to choose the better mixing models in catchments.
ConclusionSoil erosion and sediment yield are among the most destructive phenomena that cause a lot of damage in different regions. However, in order to combat this phenomenon and soil conservation projects, it is needed to be aware of the location of sediment sources in the region. Sediment fingerprinting technique based on geochemical tracers, organic, isotopic ratios, as well as using various mixing models lead to the recognition and the contribution of different sediment sources in an area. In this study, using the optimum combination of organic and rare tracers to separate the different sources and then determine contribution of this erosion and sediment yield resource using, Collins (1997), Collins Modified (2010), Motha (2004), Landwehr (2010), and Slattery(2000) models. In this case, to determine the best model GOF index and matching results with actual measurements of the sediments and erosion amount based on BLM model were used. The results were analyzed by discriminant analysis and showed compounds of C, Cu Si, and Ti were considered as tracer and M Collins mixing model with indices GOF and RMSE, 99.9 and 2.07 were selected as the best models. Measurement results showed that sedimentation rates of different land uses, pastures with weak and moderate cover percentage and gardens with relative importance of 3.04 and 7.72 are prior in the management of soil conservation. Mixing models showed the highest proportion in sediment and erosion basins in the rangelands area is about 16.75 percentage and 57.04 contribution percent and the cultivation has minimal role in sediment yield of Tange Bostanak watershed. Moreover, results showed that Motha model with a 0.924 correlation coefficients had the smallest difference with actual values.