5:15 PM - 6:45 PM
[ACG43-P04] Developing an Algorithm for Estimating Suspended Sediment Concentration in Global Rivers using Sentinel-2 Satellite Images and Flux Conservation at Confluences
Keywords:Suspended sediment concentration, remote sensing
Suspended sediment in rivers plays an important role in ensuring the diversity of riparian environments in rivers, maintaining shorelines, and supplying nutrients associated with sediment to the ocean. On the other hand, in-situ measurements of suspended sediment concentrations (SSC) are both financial and time consuming and making it difficult to obtain the actual form of global sediment transport. To compensate for inadequate observations and to clarify global sediment dynamics, remote sensing has recently become an effective tool for estimating SSC. Previous studies have been focused on developing a relationship between SSC and remote sensing reflectance, which satellite observes. However, while these methods require parameter estimation based on in-situ observations, its parameters differ from river to river. Therefore, the relationship obtained from one river cannot be applied to other rivers. In addition,a few studies aim for global application are based on field observations and have low accuracy.
In this study, we developed a model that does not require in-situ observation of SSC. The model combines bio-optical model and suspended sediment flux preservation at the confluence.
Firstly, the model that once particle size distribution of suspended sediment is known, SSC is obtained, is developed. The Junge function (N=KD^-j, where N is the number of particle for each size, D is the diameter and K,j are coefficients) is assumed as the particle size distribution. This particle size distribution gives the backscattering coefficient per unit mass of suspended sediment, which is the input value for the bio-optical model, by using Mie scattering theory. The bio-optical model is simplified by the assumption that the reflectance at near-infrared wavelengths in rivers containing high SSC is dominated by suspended sediment and water. Then, bio-optical model tells the relationship between SSC and particle size distribution. Supposing that particle size distribution is stationary in each river, the optimal SSC would minimize the sum of all the error between the reflectance estimated from the model and the satellite reflectance.
Next, the parameter j for the particle size distribution is determined. At the confluences, there are two ways to determine the sediment concentration in the river after the confluence: using flux conservation at the confluences or using satellite reflectance. The optimal grain size distribution should minimize the error in the sediment concentration determined by these two methods for all satellite images acquired. This yields the parameters of the optimal grain size distribution.
By applying this algorithm, the paraticle size distribution of sediment in each river is estimated, and the sediment concentration can be determined without using parameters that require field observations.
In this study, we developed a model that does not require in-situ observation of SSC. The model combines bio-optical model and suspended sediment flux preservation at the confluence.
Firstly, the model that once particle size distribution of suspended sediment is known, SSC is obtained, is developed. The Junge function (N=KD^-j, where N is the number of particle for each size, D is the diameter and K,j are coefficients) is assumed as the particle size distribution. This particle size distribution gives the backscattering coefficient per unit mass of suspended sediment, which is the input value for the bio-optical model, by using Mie scattering theory. The bio-optical model is simplified by the assumption that the reflectance at near-infrared wavelengths in rivers containing high SSC is dominated by suspended sediment and water. Then, bio-optical model tells the relationship between SSC and particle size distribution. Supposing that particle size distribution is stationary in each river, the optimal SSC would minimize the sum of all the error between the reflectance estimated from the model and the satellite reflectance.
Next, the parameter j for the particle size distribution is determined. At the confluences, there are two ways to determine the sediment concentration in the river after the confluence: using flux conservation at the confluences or using satellite reflectance. The optimal grain size distribution should minimize the error in the sediment concentration determined by these two methods for all satellite images acquired. This yields the parameters of the optimal grain size distribution.
By applying this algorithm, the paraticle size distribution of sediment in each river is estimated, and the sediment concentration can be determined without using parameters that require field observations.