Executive Summary : | The Netravati river is the major freshwater source for the highly populous districts of Dakshina Kannada & Udupi, with a net yield of over 1240 tmc. Despite this, a severe scarcity of water is faced by most cities in these districts during the peak summer season. The solution lies in enhancing per capita storage capacity for efficient conservation and utilization of river resources as well as rainwater to prevent excess water discharge into the sea. A multitude of vented dams and check dams have been constructed spanning the Netravati river, however, a critical issue that threatens them is the phenomenon of saltwater intrusion, common in West-flowing rivers. The problem is further exacerbated by human activities like deep drilling for borewells, large-scale irrigation, industrial pollution etc. Incidence of higher salinity raises significant concerns in qualitative water management, agricultural activities, industrial usage and increased damage to infrastructures/machinery and aquatic/land biodiversity. Recharge techniques like check dams can help in groundwater management to reduce salinity, which is traditionally managed with extensive sampling and in situ surveys. However, the Netravati catchment area is more than 4,409 sq km in size, and conventional periodic surveying of this entire region across varying seasons is a tedious process that requires extensive equipment, manpower and financial resources. Designing effective measures to model & predict salinity characteristics leveraging multiple sources of data, like satellite (direct & indirect salinity measurements), UAVs (e.g., drones), and in situ measurements (such as water quality, basic minerals (chloride, iron), turbidity, sediment load at different water depth, bathymetry (depth, soil sampling)), can play an important role in optimal water resource management. In this project, an AI-based end-to-end framework for automating the salinity mapping process using multimodal data is proposed. We utilise multiple satellite datasets (with the lowest resolution of 30m), images captured using GOI DigitalSky approved UAVs (with a lowest resolution of 3-1cm/pixel and flying height altitude of 50-150m), and in situ measurements in the river catchment area (with an approximate resolution of 100m gridwise or customized to capture the stream width, as per the requirement). The multi-source datasets will be preprocessed using intelligent techniques designed for superimposing varied input data modalities on each other and for upscaling them in similar resolutions. The resultant data is used to train deep neural models for representative feature learning, to spatially predict river salinity maps for qualitative and quantitative assessments of saltwater intrusion and contamination in dammed-up areas. These can be used to measure and evaluate changes in the extent of saltwater intrusion over time, and also enable future predictions for improved freshwater management for sustainable development. |
Co-PI: | Dr. Sowmya Kamath S, National Institute Of Technology (NIT) Karnataka-575025, Dr. Pruthviraj Umesh, National Institute Of Technology (NIT) Karnataka-575025, Dr. Gangadharan KV, National Institute Of Technology (NIT) Karnataka-575025, Prof. Soumya K Ghosh, Indian Institute Of Technology (IIT) Kharagpur, West Bengal-721302 |