Integrating sensor data and machine learning to advance the science and management of river carbon emissions

Abstract

Estimates of greenhouse gas emissions from river networks remain highly uncertain in many parts of the world, leading to gaps in global inventories and preventing effective management. In-situ sensor technology advances, coupled with mobile sensors on robotic sensor-deployment platforms, will allow more effective data acquisition to monitor carbon cycle processes influencing river CO2 and CH4 emissions. However, if countries are to respond effectively to global climate change threats, sensors must be installed more strategically to ensure that they can be used to directly evaluate a range of management responses across river networks. We evaluate how sensors and analytical advances can be integrated into networks that are adaptable to monitor a range of catchment processes and human modifications. The most promising data analytics that provide processing, modeling, and visualizing approaches for high-resolution river system data are assessed, illustrating how multi-sensor data coupled with machine learning solutions can improve both proactive (e.g. forecasting) and reactive (e.g. alerts) strategies to better manage river catchment carbon emissions.