The Ministry for the Environment engaged Pattle Delamore Partners and Lynker Analytics to develop a computer-based model using machine-learning to establish a proof of concept for the Jobs for Nature environmental monitoring work programme. The aim of the work is to have an independent and cost-effective mechanism for assessing the survival of funded riparian planting.
The Ministry for the Environment engaged Pattle Delamore Partners and Lynker Analytics to develop a computer-based model using machine-learning to establish a proof of concept for the Jobs for Nature environmental monitoring work programme. The aim of the work is to have an independent and cost-effective mechanism for assessing the survival of funded riparian planting.
The project involved:
- testing the suitability of different types and resolutions of imagery (medium- and high-resolution satellite imagery, aerial and drone imagery) for remote sensing of riparian planting survival
- classification and analysis of riparian plants identified at multiple sites using multiple types of imagery
- field validation of plant classification outputs
- advice on upscaling the plant survival ‘chain of analysis’ to a regional and national level.
This technical report describes the project from development of plant classification and machine learning models, through pilot site testing and ground truthing, to application to many sites. The report makes key findings about the imagery and model strengths and limitations.