We can't find the internet
Attempting to reconnect
Something went wrong!
Hang in there while we get back on track
Flux to Flow: a Clearer View of Earth’s Water Cycle Via Neural Networks and Satellite Data
Summary
This dissertation developed neural network methods to enhance the spatial resolution of satellite measurements of Earth's water cycle, enabling finer-scale monitoring of hydrological processes such as precipitation, evaporation, and runoff across diverse environments.
Earth’s weather and climate changes dynamically, and humans must monitor the evolution of our environment or fail to plan and adapt. Water is our most valuable resource, and as such it is of the utmost importance to have a continuous view of its movements as water cycles over, in, and under land, in the atmosphere, and in the sea. One way that we observe the water cycle is with satellite measurements. Unfortunately, the resolution of these instruments is typically too coarse to visualize many transient phenomena of interest. If their vision is fine enough, the instrument captures only a small space of Earth at any given time. Furthermore, it is valuable to connect satellite-derived water cycle measurements of large watershed basins with ground truth observations of the rivers flowing through these basins. These connection operations, though, are time and cost prohibitive or limited in performance, with efficient functions veiled behind black box closed-source solutions. Herein, all of the above facts are investigated under the veil of a single name, Flux to Flow (F2F). The name Flux to Flow encapsulates how the work takes several fluxes of geospatial data and transforms them into more coherent flows of knowledge. F2F performs metrologically well in the operation of streamflow forecasting when focused solely on a single or a few hydrological monitoring nodes at a time. Based on the results, future work might entail scaling up of the system to many computing nodes running in tandem, observing more adjacent outputs, or in its application as a standard device in connected municipal systems.
Sign in to start a discussion.
More Papers Like This
The Utilization of Satellite Data and Machine Learning for Predicting the Inundation Height in the Majalaya Watershed
This paper is not about microplastics; it uses satellite rainfall data, HEC-RAS flood modeling, and artificial neural networks to predict flood inundation heights in the Majalaya Watershed of Indonesia.
Quantifying the Geomorphic Effect of Floods Using Satellite Observations of River Mobility
This paper is not about microplastics; it uses satellite imagery and machine learning to study how flood magnitude, duration, and hydrograph shape determine lateral erosion and channel change in rivers.
Seasonal and annual tropical river pattern change detection using machine learning
Researchers applied machine learning to Sentinel-2 satellite imagery to detect seasonal and annual changes in tropical river channel patterns in a region with strongly seasonal rainfall, successfully classifying active channel landforms including water, bare sediment, and vegetated bars. The approach provides a scalable method for monitoring dynamic tropical river systems.
Modeling of daily groundwater level using deep learning neural networks
Researchers applied a CNN-biLSTM deep learning model to predict daily groundwater levels, finding it outperformed conventional modeling approaches by capturing both spatial patterns and temporal dependencies in the data. The method offers improved accuracy for groundwater monitoring, which is critical for managing increasingly stressed freshwater resources.
Advancing hydrological monitoring using image-based techniques: challenges and opportunities
This paper is not about microplastics — it reviews image-based techniques (including remote sensing and computer vision) for hydrological monitoring of water bodies, discussing challenges and opportunities in measuring water flow, flood events, and water quality.