Climate API + Heat Map

Enter a street address to see climate projections for that area.

Variables included in the API:

• Average yearly temperature
• Total yearly precipitation
• Relative humidity
• Specific humidity
• Maximum yearly temperature
• Minimum yearly temperature

Why it matters

Typically, climate models predict macro trends for our climate, at a continent or region-scale resolution. From a local perspective, these models are not useful. As a result, scientists have developed downscaled models, which take these global climate projections and compute values at a local (up to 0.25 deg by 0.25 deg) resolution.

Downscaled climate data has only recently become good enough to start informing regional climate adaptation strategies. However, downscaled climate datasets are hard to access. They typically require you to download terabytes of data hidden behind outdated fileservers.

The goal of this project is to develop an easy-to-use API for accessing downscaled climate projections. This API serves as a layer of abstraction over downscaled datasets, making it easy for developers to access data and derive actionable insights.

A secondary goal is to create a web app demonstrating the usage of this API.

Demo

• The Climate API is deployed at https://2nnrmg.deta.dev/. For docs, see https://2nnrmg.deta.dev/docs.

Note: the web app interface is not currently deployed because the API spec is still changing.

Installation

To install python dependencies, set up a virtual environment in server and run pip install -r requirements.txt.

The Jupyter notebooks contain a set of dependencies separate from the main API, so you should set up a second virtual environment if you want to run those.

How it works

The Climate API is built using Python with FastAPI and deployed on Deta. Climate data comes from the NASA Earth Exchange Global Daily Downscaled Projections dataset, using the CMIP6 scenarios.

File Structure

• The client folder contains code for the web app demo (incomplete)
• The server folder contains code for serving the climate API:
  • main.py serves the main API layer
  • lib contains modules for business logic, primarily loading and processing the data.
  • notebooks contains some experiments with downloading and processing climate data.

Next Steps

Currently, the API is very slow for some types of requests. This is because we are computing yearly statistics from daily projections. We could speed up the API greatly by computing yearly statistics (averages/sums/extremes) ahead of time and storing them in a separate database.

Perhaps more importantly, we would like to serve more than just the basic variables listed above. Implementing the xclim (https://github.com/Ouranosinc/xclim) library for this API will let us estimate other important variables, like frequency of heat waves, intensity of cold spells, and drought codes.