A Novel Instrument for Bed Dynamics Observation Supports Machine Learning Applications in Mangrove Biogeomorphic Processes

Abstract

Short-term bed level changes play a critical role in long-term coastal wetland dynamics. High-frequency observation techniques are crucial for better understanding of intertidal biogeomorphic evolution. Here, we introduce an innovative instrument for bed level dynamics observation, that is, LSED-sensor (Laser based Surface Elevation Dynamics sensor). The LSED-sensors inherit the merits of the previously introduced optical SED sensors as it enables continuous high-frequency monitoring with relatively low cost of labor and acquisition. As an iteration of the optical SED-sensors, the LSED-sensors avoid touching the measuring object (i.e., bed surface), and they do not rely on daylight by adapting laser-ranging technique. Furthermore, the new LSED-sensors are equipped with a real-time data transmission function, enabling automatic observation networks covering multiple (remote) sites. During a 22-day field survey in a mangrove wetland, good agreement (R2 = 0.7) has been obtained between the automatic LSED-sensor measurement and an accurate ground-truth measurement method, that us, Sedimentation Erosion Bars. The obtained LSED-sensor data were subsequently used to develop machine learning predictors, which revealed the effect of vegetation is a main driver in the accumulative and daily bed level changes. We expect that the LSED-sensors can further support machine learning applications to extract new knowledge on coastal biogeomorphic processes.