Devis Tuia

Biographie

I come from Ticino and studied in Lausanne, between UNIL and EPFL. After my PhD at UNIL in remote sensing, I was postdoc in Valencia (Spain), Boulder (CO) and EPFL, working on model adaptation and prior knowledge integration in machine learning. In 2014 I became Research Assistant Professor at University of Zurich, where I started the 'multimodal remote sensing' group. In 2017, I joined Wageningen University (NL), where I was professor of the GeoInformation Science and Remote Sensing Laboratory. Since 2020, I joined EPFL Valais, to start the ECEO lab, working at the interface between Earth observation, machine learning and environmental sciences.

Official source

https://people.epfl.ch/devis.tuia?lang=fr

À propos de ce résultat

Cette page est générée automatiquement et peut contenir des informations qui ne sont pas correctes, complètes, à jour ou pertinentes par rapport à votre recherche. Il en va de même pour toutes les autres pages de ce site. Veillez à vérifier les informations auprès des sources officielles de l'EPFL.

Unités associées (7)

Unités associées

Chargement

Cours enseignés par cette personne (4)

Cours enseignés par cette personne

Chargement

Domaines de recherche associés (48)

Domaines de recherche associés

Chargement

Publications associées (124)

Publications associées

Chargement

Personnes menant des recherches similaires (97)

Personnes menant des recherches similaires

Chargement

GeoAI as Collaborative Effort. Interview with Devis Tuia

Devis Tuia

2023

Chargement

Short-term runoff forecasting in an alpine catchment with a long short-term memory neural network

Corinna Barbara Frank, Devis Tuia

The governing hydrological processes are expected to shift under climate change in the alpine regions of Switzerland. This raises the need for more adaptive and accurate methods to estimate river flow. In high-altitude catchments influenced by snow and glaciers, short-term flow forecasting is challenging, as the exact mechanisms of transient melting processes are difficult to model mathematically and are poorly understood to this date. Machine learning methods, particularly temporally aware neural networks, have been shown to compare well and often outperform process-based hydrological models on medium and long-range forecasting. In this work, we evaluate a Long Short-Term Memory neural network (LSTM) for short-term prediction (up to three days) of hourly river flow in an alpine headwater catchment (Goms Valley, Switzerland). We compare the model with the regional standard, an existing process-based model (named MINERVE) that is used by local authorities and is calibrated on the study area. We found that the LSTM was more accurate than the process-based model on high flows and better represented the diurnal melting cycles of snow and glacier in the area of interest. It was on par with MINERVE in estimating two flood events: the LSTM captures the dynamics of a precipitation-driven flood well, while underestimating the peak discharge during an event with varying conditions between rain and snow. Finally, we analyzed feature importances and tested the transferability of the trained LSTM on a neighboring catchment showing comparable topographic and hydrological features. The accurate results obtained highlight the applicability and competitiveness of data-driven temporal machine learning models with the existing process-based model in the study area.

2023

Chargement

Social media has created new opportunities to map cultural ecosystem services (CES) related to biodiversity at large scales. However, using these novel data to understand people's preferences in relation to these CES remains a challenge. To address this, we trained a deep learning model to capture people's interactions with selected flora and fauna on Flickr as a cultural service related to biodiversity and compared this with citizen science data on iNaturalist, with photos of individual species considered as human–species interactions. After mapping the distribution of people's interactions in Great Britain on Flickr and iNaturalist, we find significant spatial differences in people's preferences on the two platforms. Using a second, pretrained deep learning model, we were also able to identify different preferences for species groups such as birds on social media versus citizen science. To better understand people's preferences, we also compared peoples' interactions with species richness and abundance for a group of 36 bird species, sometimes finding large differences between people's interactions and these ecological measures. Our findings demonstrate that social media can be used to include a wider range of preferences in CES assessments along-side citizen science data. However, these preferences reflect only a limited first-hand experience of biodiversity.

2023

Chargement