Urban rivers and protected areas emerged as key cultural hotspots, crucial for inclusive and effective river management.

All code and data are fully open & reproducible: github.com/francesc30/C...

We trained a ResNet-152 deep learning model (>90% accuracy) to classify Flickr images of river-related activities and combined them with biophysical variables via XGBoost (>85% accuracy) to identify the main drivers of river recreation and aesthetics.

The framework integrates biophisical data, deep learning, and social media data theory to identify CES hotspots

🔬 Publications are underway (J. Env. Management & MethodsX)!

💧🌿I developed an AI-based framework to model freshwater cultural ecosystem services at large spatial scales using social media data.

Key Findings

There is a strong correlation between the naturalness of freshwater ecosystems and Aesthetic Value.

More CES-related tweets are posted on weekends (when people have more leisure time), and during spring and summer when the weather is favourable.

Model training
The Ter River basin was selected to train the AI model using the Multinomial Naive Bayes (MNB) algorithm, and the Forth, Scarce, and Fluvià basins were used to test the model’s performance.
The model achieved an accuracy of over 80%.

Allow geolocated publications!!