KBA Insight streamlit app

[01painadam]

Goal(s)

Build a simple streamlit app that demonstrates how agent/LLM features can provide insights for a portfoio of Key Biodiversity Areas (KBA). These insights pull from rich descriptive metadata about the KBA, and support statements with calculations using GFW/LCL data.

Data and metadata here

Example insights

The app should:

• Select countries of interest (KBAs are samples from these)
• Show KBAs on map
• LLM should provide a general summary of the portfolio as a whole, highlighting:
• description of common habitats & species
• presence of indigenous communities, conservation coverage, management practices
• Offers insights about threats & pressures to nature, ecosystem health (recovery/restoration), and possible actions (organisations to engage with)
• LLM can comment on at risk species
• Insights are backed up by relevant data, and data layers are shown on the map where possible (loss, gain, extent)

Scope

Define the limits of the work to be carried out; what is included or not included?

• Simple analysis tools to calculate data to support insights e.g.
  • Tree cover loss trends
  • Tree cover loss intensity (tree cover in 2000 / total loss)
• Use static dataset for KBA metadata and zonal stats
• Sample portfolio down to some reasonable number of areas (e.g. 10-20)
• Priority datasets:
  • Tree cover Loss (annual), Gain, Extent
  • Total Tree cover loss drivers
  • Forest Carbon emissions (annual), removals, stocks
  • Grasslands
• Streamlit app

User Stories & Acceptance Criteria

What does the user want to achieve? For each, define list of pass/fail criteria that must be met for the user story to be considered complete, from the users' perspective (if known)

• When I am selecting a portfolio of KBAs, I want to filter by multiple countries, so that I can see insights for areas relevant to me (often across border e.g. Amazon, Congo Basin etc)
• When I select a portfolio of KBAs, I want to see a summary description of that Portfolio (habitats/ecosystems, species etc), so that I can quickly understand some characteristics about the areas in general
• When I select a portfolio of KBAs, I want to see a description of the threats and pressures to biodiversity that Portfolio faces (deforestation trends, drivers of land cover change), so that I can quickly understand the context of those areas e.g. which species or habitats are being affected and how
• When I select a portfolio of KBAs, I want to see a summary of the state of ecosystem health across the Portfolio (the amount of conservation coverage, tree cover and tree cover gain), so that I can quickly understand some characteristics about the areas in general
• When I am trying to understand an insight, I want to see related data layers on the map, so that I can contextualise statements
• When I am trying to understand an insight, I want to see supporting data and analysis, so that I can trust in the descriptive statements
• When I am planning actions, I want to know the key actors (e.g. those managing the land, so that I can engage with those organisations or communities to protect and restore nature.

Hints

Suggestions or hints on how to get started or go about achieving the goal of the epic.

Future work

What does this unlock later (that is out of scope here)?