Soil degradation monitoring; Data analysis; Machine learning; Data collection; Biodiversity monitoring; Environmental monitoring; Citizen science; Data integration
ZweifelL, SamarinM, MeusburgerK, AlewellC (2019), Identifying Shallow Landslides on Swiss Alpine Grasslands using Machine Learning
, MLEG, Dübendorf.
Zweifel L., Meusburger K., Alewell C. Spatio-temporal analysis of soil degradation in Swiss alpine grasslands based on Object-Based Image Analysis, (2018), Spatio-temporal analysis of soil degradation in Swiss alpine grasslands based on Object-Based Image Analysis
, EGU, Vienna 20, 7162-7162.
ZweifelL, SamarinM, MeusburgerK, AlewellC (2018), Spatio-temporal Patterns of Soil Degradation in Swiss Alpine Grasslands revealed by Object-Based Image Analysis
, SGM, Bern.
Monitoring and analysing data streams is crucial for understanding many real-world phenomena. Sometimes such data streams can be obtained automatically by high-throughput grid-based sensor devices, but in complex and dynamic environments it is frequently necessary to complement such sensor data with specific high-precision field observations. In the weObserve project we analyse exactly such combined data situations where citizen observers provide semantically rich but typically biased information directly from the field, and where this information needs to be jointly analysed together with large-scale low-resolution sensor data. We address all individual steps of a suitable information processing pipeline ranging from data collection - where we focus on novel devices and apps for citizen scientists - to data integration - with all its associated problems of dealing with heterogeneous data records that differ in resolution, in reliability, precision and coverage - to data analysis - where we develop novel machine learning methods that can detect patterns in Big Data and exploit such patterns for making predictions. We evaluate this new processing pipeline in two carefully selected applications with complementary requirements and different ways to gather data. Whilst the first application scenario concerns environmental monitoring in the context of solid degradation and landslides in Swiss Alpine regions, the second case study focuses on biodiversity monitoring with specific emphasis on radar-based detection of bird migration patterns in Switzerland and its neighbouring countries. Both applications have in common that high-throughput grid-based sensor information (areal photographs, radar signals etc.) have to be combined with high-precision observational data from citizen observers. We expect that intelligent data analysis will lead to new insights to soil degradation processes (including extreme events like landslides), and - in the second application - to a better understanding of bird migration patterns. Detailed knowledge about such migration patterns will be indispensable for developing monitoring and forecast systems which are particularly pertinent for flight safety, renewable energy and control of pests and diseases.