Estimation Total Dissolved Solids and Turbidity Concentration in Karkheh and Dez Dam and Great Karun River by Using Sentinel-2 Satellite Images

1-Introduction
Considering the importance of rivers as part of freshwater resources and their role in meeting the needs of agriculture, industry, urban populations, etc., monitoring and predicting the quality of these water resources is essential. These water sources are affected by numerous factors due to their different geological and environmental conditions and their qualitative status also undergoes dramatic changes. However, the quality monitoring of these abundant water resources on the planet's surface is not feasible and requires the use of advanced and powerful tools (Bagherian Marzouni et al., 2014). Due to its capabilities, satellite remote sensing can be used as one of these tools in monitoring water quality and will accurately detect the spatial and temporal changes of these water sources (Bonansea et al., 2015). So far, in many studies, the capabilities of remote sensing satellites to estimate surface water quality parameters has been evaluated, and in most of them, acceptable results have been obtained indicating the ability of this technology in the issue as mentioned above. Among these studies, we can mention laili et al. (2015), in which in a small section of Indonesian waters, have figured out a new regression algorithm between Landsat 8 and groundwater quality parameters. Toming et al. (2016) in a study using satellite images of Sentinel-2 on the water quality of the lakes in Estonia, could find a good correlation between the satellite band proportions and ground.
The purpose of this research is to establish a relation between satellite images of Sentinel-2 A and two quality water parameters with a suitable model along the Karun and Dez River. For this purpose, firstly suitable spectral indices were extracted from them by applying the necessary processing on satellite images. In the next step, optimal relationships between extracted indices and water quality parameters are established using different models. Finally, using models with higher accuracy in terms of modeling, the dispersion map of each parameter in the length of the Karun River is provided.The purpose of this research is to establish a relation between satellite images of Sentinel-2 A and 2 quality water parameters with a suitable model along the Karun and Dez River. For this purpose, firstly suitable spectral indices were extracted from them by applying the necessary processing on satellite images. In the next step, optimal relationships between extracted indices and water quality parameters are established using different models. Finally, using models with higher accuracy in terms of modeling, the dispersion map of each parameter in the length of the Karun River is provided.
 
2-Methodology
This study presented in eight steps as below:
Step 1: Preparation of ground data and satellite imagery:The ground data used in this study is the measured data at the water quality sampling stations. The data included information on these quality parameters that were used from 2015 to early 2017 in ten stations.
Step 2: Recording the value of the reflection bands at the ground measurement stations:In order to implement this research, satellite images of sentinel-2 and groundwater quality parameters were collected and measured at the same time from the study area. In this step, the values of measured water quality parameters were also sorted by date and sampling stations were prepared in separate files.
Step 3: Analyze the initial sensitivity and determine the bands that have a stronger connection with each water quality parameter
 
Table 1: result of sensitivity analysis for sentinel-2 bands





TDS


Turbidity


EC


pH


Hco3


So4


Cl


Na


K


Mg


Ca


Parameter Type
Band Number




0.376


0.472


0.296


0.384


0.493


0.219


0.338


0.279


0.312


0.217


0.294


B2




0.379


0.303


0.325


0.307


0.238


0.239


0.268


0.238


0.179


0.291


0.217


B3




0.352


0.237


0.283


0.278


0.260


0.232


0.225


0.269


0.165


0.196


0.269


B4




0.346


0.332


0.274


0.428


0.315


0.214


0.256


0.294


0.256


0.275


0.313


B5




0.401


0.208


0.248


0.322


0.294


0.278


0.253


0.249


0.210


0.268


0.239


B6




0.403


0.257


0.227


0.299


0.273


0.281


0.258


0.256


0.203


0.283


0.210


B7




0.263


0.285


0.301


0.346


0.198


0.245


0.231


0.227


0.184


0.209


0.224


B8




0.422


0.306


0.316


0.309


0.241


0.275


0.251


0.244


0.195


0.299


0.212


B8a




0.249


0.205


0.267


0.325


0.238


0.273


0.233


0.287


0.205


0.209


0.158


B11




0.391


0.265


0.214


0.310


0.282


0.293


0.254


0.247


0.270


0.244


0.178


B12





 
Step 4: Calculating spectral indices and selecting spectral indicators with higher correlation
Step 5: Secondary Sensitivity Analysis and Selection of Spectral Indicators with Stronger Connections
 
In the next step, by applying the sensitivity analysis method, the relationship between each spectral indicator and water quality parameters was calculated (Table 2).
 
Table 2. Result of sensitivity analysis for spectral indicator





TDS


Turbidity


EC


pH


So4


Hco3


Cl


Na


K


Mg


Ca


Parameter Type
 
Spectral Indexes




0.455


0.580


0.470


0.407


0.534


0.260


0.482


0.535


0.364


0.511


0.366


Single bans reflectance




0.465


0.659


0.563


0.516


0.599


0.501


0689


0.688


0.670


0.532


0.666


( 14BmaxBmin)">




0.436


0.740


0.452


0.633


0.562


0.681


0.701


0.598


0.600


0.485


0.677


( 14BminBmax)">




0.396


0.702


0.438


0.720


0.527


0.581


0.758


0.669


0.656


0.506


0.740


( 14Bmax-BminBmax+Bmin)">





 
Step 6: Normalization of data
Step 7: Modeling the relationship between satellite images and groundwater quality parameters:In order to model the relationship between satellite images and groundwater quality parameters, and based on the results of previous steps, the normalized values derived from the calculation of spectral indices were determined as inputs and water quality parameters were determined as outputs of ANN and ANFIS models.
Step 8: Providing water dispersion map for water quality parameters:At this step, the modeling process was repeated with the transformation of ANN and ANFIS models until each model was accurately mapped the relationship between water quality parameters.
 
3- Findings of the research
Table 3 shows the evaluation result of used model in this study.
 
Table 3. Evaluation result of ANN and ANFIS model for water quality parameters.





Hco3


So4


Cl


Na


K


Mg


Ca


WQPT




ANFIS


ANN


ANFIS


ANN


ANFIS


ANN


ANFIS


ANN


ANFIS


ANN


ANFIS


ANN


ANFIS


ANN


Error Type




0.497


0.315


0.0871


0.691


0.266


0.263


0.229


0.264


0.136


0.0709


0.127


0.397


0.120


0.279


RE




0.164


0.131


0.0587


0.311


0.0959


0.0748


0.102


0.079


0.126


0.0605


0.077


0.157


0.115


0.194


RMSE





 
Figures 1– 4 show the concentration map of TDS and turbidity parameters studied in this research in Karun River in Dez and Karkheh dam and the Karun River from Malasani section to the Farsiat station.
 
                               (a)
 
                            (b)
Figure 1: Concentration map of TDS parameter in a) Karkheh and b) Dez dam.
 
Figure 2. Concentration map of TDS parameter Karun River.
 
 
(a)
 
(b)
Figure 3. Concentration map of turbidity parameter in A) Karkheh and b) Dez dam.
 
Figure 4. Concentration map of turbidity parameter Karun River.
 
4- Conclusion
In this study, two models of ANN and ANFIS computational intelligence models were used to model the relationship between satellite images of Sentinel-2 and two quality parameters of water along the Karun River. The results of this study indicate the high level of remote sensing ability to monitor water quality, similar to other studies; as this is well understood in previous researches, remote sensing technology can be widely used to monitor other surface water resources of Khuzestan province.
 
References
Bagherian.Marzouani.M., Akhoundali.A.M., Moazed.H., Jaafarzadeh.N., Ahadian.J., Hasoonizadeh.H., 2014, Evaluation of Karun River Water Quality Scenaros Using Simulation Model Results, International Journal of Advanced Biological and Biomedical Research, Volume 2, Issue 2, 339-358.
Bonansea.M., Claudia Rodriguez.M., Pinotti.L., Ferrero.S., 2015, Using multi-temporal Landsat imagery and linear Models for assessing water quality parameters in Rio Tercero reservoir (Argentina), Journal of Remote Sensing of Environment, 158, 28–41.
Laili, N., Arafah, F., Jaelani, L., Subehi, L., Pamungkas, A., Koenhardono, E., & Sulisetyono, A. 2015. Development of water quality parameter retrieval algorithms for estimating total suspended solids and chlorophyll-a concentration using landsat-8 imagery at poteran island water. ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, 2(2), 55.
Toming, K., Kutser, T., Laas, A., Sepp, M., Paavel, P., Nõges T., 2016. First experiences in mapping lake water quality parameters with Sentinel-2 MSI imagery, Remote Sensing journal, 8, 640.