Leaftime

About

Leaftime builds technology to help people appreciate, protect, and restore nature. See leaftime.net for scientific and technical consulting on natural climate solutions projects, using remote sensing and machine learning.

The Antelope Valley project aims to connect people with the nature experiences they love, by letting them know when and where to be. Please stay on trails and don't walk among the flowers. Soil compaction from foot traffic reduces the loose, aerated ground that poppies need to germinate — damage that can suppress blooms for multiple years.

If you enjoy this project and want to help keep it running, consider buying us a coffee.

How does it work? Technical details

This project uses AI in combination with satellite imagery and weather data to forecast orange California poppies and yellow goldfields in the Antelope Valley, north of Los Angeles. The forecast comes from a deep learning model, leveraging vegetation and flower indices from satellite remote sensing (e.g. Chen et al., 2019, Zang et al., 2020) and weather data such as accumulated seasonal precipitation, growing degree days, and wind speed.

The forecast model provides a modest improvement over a naive model predicting that “the future looks like the past”: the forecast improves on a cloud-free mosaic of orange and yellow flower classes from the current year's imagery, previous to the forecast date, by 7% in flower-weighted F1 score. Because of this, the forecast greatly benefits from low-latency access to Sentinel-2 satellite imagery, which is freely available from Amazon Web Services as Cloud-Optimized GeoTIFFs, managed by Element 84. Leaftime plans to increase forecast accuracy as additional data becomes available.