DISA

When? Friday April 5th 14-16
Where? Onsite: B1006 at Linnaeus University in Växjö and online
Registration: Please sign up for the PhD-seminar via this link https://forms.gle/ptgACTxKcaQ5VASr8 by April 3rd (especially important if you plan on attending onsite so we have fika for everyone)

14.00-14.10 Welcome and practical information from Welf Löwe
14.10-14.55 Presentation and discussion: Bridging temporal gaps in vegetation indices satellite data and forecasting the future – Tibo Bruneel, PhD student at Softwerk AB
14.55 – 15.05 Coffee break
15.05 – 15.50 Presentation and discussion – PhD administration, information and course related issues – Gisela Isgren, Ewa Püschl and Milena Rusic
15.50 -16.00 Sum up and plan for our next seminar on May 3rd

Abstract

Bridging temporal gaps in vegetation indices satellite data and forecasting the future – Tibo Bruneel, PhD student at Softwerk AB
In this research seminar I will present the machine learning work I have done so far in my PhD studies with some intermediate results. The research addresses the challenge of inconsistent temporal data gaps in Sentinel-2 vegetation indices time series and the forecasting of the vegetation index future.

Temporal data gaps occur in sentinel-2 optical data, mostly arising from factors such as cloud cover and adverse weather conditions. Having very inconsistent time series prevents further analysis such as detection, quality control and forecasting. Therefore we first address the data continuity challenge to obtain a consistent daily estimation of the vegetation index.

Given a consistent timeline of daily estimates, we further attempt to predict the near future of the vegetation index. Here we attempt to achieve both a high forecasting accuracy, and a strong generalisability over different locations in the world, as models trained on a single location often are hard to generalise on different environments.