Published in Scientific Papers. Series E. Land Reclamation, Earth Observation & Surveying, Environmental Engineering, Vol. XIV
Written by Adina NITULESCU
Forests are among Earth’s most important ecosystems, but unfortunately are vanishing at alarming rates. Thus, deforestation is a global challenge with serious impacts. This paper utilizes Google Earth Engine Landsat data, together with climatic and land-cover data in order to predict the annual deforestation rate in Suceava County. This is being accomplished by using deep learning techniques, more precisely by using Long Short-Term Memory Neural Networks in order to forecast the land changes. These offer distinct advantages in handling time-series data with long-term dependencies, thus making them particularly suitable for tracking changes in vegetation over time. The results have shown an average prediction of 0.45% tree cover loss every year for the next 10 years timeframe. Deforestation studies in Romania have largely focused on spatial analysis but predictive temporal modelling using LSTM remained underexplored. By leveraging the collected data, the paper demonstrates how deep learning techniques can enhance the monitoring and forecasting of forest loss, supporting more sustainable forest management and policy-making. Integrating such approaches could pave the way to anticipating risk and protecting high risk zones before deforestation occurs.
[Read full article] {simplepopup link="Citation97"}[Citation]
{/simplepopup}
{simplepopup popup="false" name="Citation97"}NITULESCU A. 2026, DEFORESTATION PREDICTIONS IN NORTHERN ROMANIA USING LONG SHORT-TERM MEMORY NEURAL NETWORKS. Scientific Papers. Series E. Land Reclamation, Earth Observation & Surveying, Environmental Engineering, Vol. XIV, Print ISSN 2285-6064, 946-951. {/simplepopup}
