DryAlps-IT

The DryAlps-IT project last for 12 months from January 2025. It is composed of 6 work packages:

WP1: Project management

WP2: Improved drought news collection

In this WP we aim to enhance the existing scraper of past drought news (both in number and spread) to overcome some of its limitations, such as targetization and search-location biases. In particular, we aim to make the algorithm suitable for an operational news collection pipeline by improving its efficiency and making the code modular and natively ready to integrate other aggregators for a wider coverage.

WP3: Filtering and classification

WP3 deals with filtering and classification of the collected texts using NLP (Natural Language Proicessing) methods. We aim to improve the existing methodological workflow using Large Language Models to filter non-relevant texts and to enrich relevant texts with metadata on general impact.

WP4: Validation and testing

In WP4, we aim to push activities related to the validation of the proposed methodology with other existing or annotated datasets. We will test the developed code on data collection, filtering and classification methodology and evaluate filtering and classification reliability of different approaches in order to make the pipeline robust and ensure the quality of its dynamically growing output.

WP5: Near-real time data visualization and sharing

In this WP, we aim at integrating the existing ADO (https://ado.eurac.edu/) and EDP (https://edp-portal.eurac.edu/) platforms with the newly developed drought impacts database. In particular, the ADO platform will provide temporally and spatially dynamic data visualization while EDP will be used to host the database, provide metadata service and data access through and API or web service.

WP6: Writing and dissemination

Communication activities related to the dissemination of the project will be developed in WP6. We are aiming to reach out different type of public through Eurac magazine and blogpost article as well as the development of a scientific manuscript presenting the final database and the applied methodology.

The main outputs of DryAlps-IT are:

1. Enhanced and open-source algorithm for past drought impact data collection transferable to other areas and languages of the world;
2. Established open-source methodology integrating NLP, linguistic and drought knowledge for the classification of impacts, their location and affected sectors;
3. Established an automatic pipeline for harvesting, classifying and mapping Italian news in near-real time for the Italian Alps
4. A dataset for interdisciplinary studies ranging from social science, such as media coverage and sentiment analysis, the humanities (e.g. discourse analyses, history scholarship), into calibration of statistical impact models for natural science applications.
5. A peer-reviewed publication in an international scientific journal presenting the developed database, visualization dashboard and the followed approach for its creation, annotation and possible usages.
6. The creation of a database of drought impacts for the Alps (with a test for the whole of Italy) which will be integrated in the ADO platform, automatically updated every week and made publicly available with information on occurrence dates, locations and impacted sectors.