Project

Machine learning to identify good practices with Nature-based Solutions

This study aims to support scaling of Nature-based Solutions (NbS) for adaptation by using machine learning tools. These tools can improve the knowledge gained from good practices and identify key enablers and barriers for implementation. The outcomes of this study will: 1) advance evidence synthesis methods for climate action, and 2) construct a global repository of more and less successful cases to allow for dissemination of NbS that work across regions and contexts.

Many pilot projects and case studies have demonstrated that Nature-based Solutions can contribute to climate change adaptation, create economic value, improve biodiversity, and contribute to human health and wellbeing. Exemplary cases are found across various places and regions. However, insights into ‘what works’ and what the challenges and drivers are for implementation of Nature-based Solutions are lacking. This study aims to support the widespread use of Nature-based Solutions (NbS) for adaptation by using machine learning tools. These tools can improve the learning from good practices and identify key enablers and barriers for implementation.

The outcomes of this study will: 1) improve methods for finding evidence of successful climate action, and 2) create a global collection of more and less successful cases to show NbS that work. By exploring a combination of technologies from the realm of Natural Language Processing (NLP) and named entity recognition (NER), we will aim to extract and classify NbS as well as their associated barriers/enablers for implementation from text, and store this in a database.

Project description

We will take the following steps:

  • Identify suitable and accessible (e.g. via file system, database or API) natural language sources (English text). We start by targeting scientific papers and existing repositories (e.g. Climate-Adapt);
  • Develop a vocabulary and taxonomy of NbS using reference papers, chapter 4 of the IPCC WG2 report (which discusses NbS), and online discussions among experts in the field;
  • Explore and develop suitable NER models for (1) NbS and (2) barriers/enablers;
  • Explore and develop a method for relation discovery between NbS and barriers/enablers;
  • Store the outcomes in a structured database that is publicly available;
  • Validate and interpret results at an international climate conference where participants are asked to contribute and reflect on the findings.

Results

Planned results include:

  • NLP models for recognition and classification of (1) Nature-based Solutions and (2) their barriers & enablers;
  • Database containing good practices in NbS for adaptation per region, including barriers/enablers extracted from literature
  • Results published in a conference paper, for example, at Adaptation Futures 2023.