Use of image processing for estimation of snow water characteristics

Snow water equivalent and snow characteristics are critical in many applications. Traditionally, water and snow depths are measured manually in snow lines or recently using e.g. locally embedded sensory technologies such as sonar and ultrasound, respectively, which call for development of alternative technologies to deal with remote or inaccessible areas. Satellite images have been suggested as an alternative approach to ultrasound or sonar sensing modalities. Nevertheless, the spatial resolution of satellite imaging and diversity of topographical features in the scene often make the estimation very vulnerable. The prospect of integrating images from various modalities (i.e., satellite, UAV, remote observation camera, phone camera) to compensate for the above limitations has been highlighted in several research studies; although the results in this field are so far still very sparse, which call for further research both from theoretical, empirical and practical perspectives.

This PhD position contributes to this challenge by developing both theoretical frameworks and practical test fields to contribute to the estimation of both water depth and snow parameters. The initial conceptual framework for this estimation will develop appropriate mechanisms and algorithms for automatic image alignment, pre-processing in a way that extracts useful features from various images (satellite, UAV, surveillance camera and mobile camera) that can be fed to appropriate deep learning model (s) for water depth and snow parameter estimation. The candidate will scrutinize the state-of-the-art deep learning-based estimation models, which will be fine-tuned using available data collected at experimental sites. In parallel to deep-learning based approach, analytical solutions using ensemble Kalman filtering and its variants will be carried out for comparison and baseline model development. The candidate is expected to collaborate with both computer science and hydrology research teams to advance in this field. This collaboration offers access to various pilot sites where extensive in-situ observations are available and enable access to relevant practitioners that build bridge to effective and friendly implementations that can impact the research in this field. This position offers a unique opportunity to contribute to a critical area of environmental research, with significant implications for climate adaptation strategies, disaster preparedness, and sustainable development.

The ideal candidate for this PhD project will possess a comprehensive academic background in fields of computer science, computer vision, machine learning and environment science. Essential qualifications include advanced programming skills in languages like Python, R, or MATLAB, crucial for data analysis and environmental modelling. Additionally, extensive experience with Geographic Information Systems (GIS) and familiarity with Google Earth Engine are required for the processing and analysis of large-scale geospatial datasets and satellite data. Exceptional communication and collaboration skills are vital as the project demands active participation in interdisciplinary research efforts, aiming to innovate and push the boundaries of scientific understanding in environmental monitoring.

FMI

Centre for Machine Vision and Signal Processing, Faculty of ITEE, University of Oulu

Mourad Oussalah

Ali Torabi Highighi

Pertti Ala-aho

Kari Luojus FMI