Jump to Main Content
A computationally efficient model for determining sound speed in shallow tropical freshwater systems, with field validation
- Sadalage, Jyoti, Das, Arnab, Joshi, Yashwant
- Lakes & reservoirs 2019 v.24 no.1 pp. 59-65
- acoustics, data collection, electrical conductivity, equations, freshwater, lakes, models, regression analysis, salinity, water temperature
- The optimum performance of any underwater system depends on the propagation characteristics of the acoustic signals in the local water medium. Shallow tropical freshwater systems suffer from sub‐optimal performance of sonar systems deployed for any acoustic sensing because of random fluctuations of the water medium. The propagation characteristics depend largely on the sound speed variations defined by the site‐specific physical parameters such as water temperature, salinity and depth. The present study focuses on analysing the sound speed profile of a typical shallow freshwater system (Khadakwasla Lake; 18.43°N, 73.76°E), using regression models with the goal of deriving a computationally efficient model. To this end, a linear and polynomial regression model was developed, and their performance compared with the results of the model of Chen and Millero (), based on root mean square error (RMSE). In situ measurements of electrical conductivity, temperature and density (CTD) were carried out using a Valeport 602 CTD meter. Approximately 125 CTD samples were obtained during a 2‐day experimental study conducted at Khadakwasla Lake from 11 October 2017 to 12 October 2017. The data collection was undertaken throughout the day at multiple locations in the lake over a spatial distance of ~16 km. The Valeport 602 CTD meter uses the Chen and Millero () formula, considered the most conventional sound speed equation. The computational complexity of the proposed models was measured in terms of the number of addition and multiplication operations required. The validation of both models was carried out by varying the model input parameters within defined limits. The model inputs have been derived from an in situ experimental data collection process in a typical shallow tropical freshwater system. The linear regression model exhibited an RMSE of 4.15 m/s, while the polynomial regression model exhibited a good agreement with an RMSE of 0.5 m/s.