A scalable workflow for ecosystem condition assessment from Earth observation data

Roozbeh Valavi , Kristen Jennifer Williams, Eric A Lehmann, Simon Collings, Shaun Levick, Kate Gilijohann, Ning Liu, Stephanie Johnson, Samantha EM Munroe, Thomas G Van Niel, Matthew Paget, Cassandra Malley, Peter Lyon, Thomas D Harwood, Simon Ferrier

1. Quantifying ecosystem condition consistently across space and time is a central challenge in biodiversity monitoring, ecosystem accounting, and ecological restoration. Earth observation (EO) data provide extensive coverage, but translating remotely sensed ecosystem characteristics into ecologically meaningful measures of condition requires approaches that account for ecological context, and natural variability across space and time.
2. The Habitat Condition Assessment System (HCAS) addresses this challenge through a reference-based benchmarking framework that evaluates ecosystem characteristics relative to ecologically comparable high-integrity reference ecosystems. However, the absence of a general, reproducible implementation has limited broader application of this approach across regions and datasets.
3. Here, we present a scalable workflow that formalises the HCAS framework as an analytical system, in which the ClassicHCAS R-package implements core ecosystem condition calculations. The workflow integrates remotely sensed ecosystem condition variables, environmental covariates, and reference ecosystem data within a benchmarking framework. Ecological context is addressed through reference ecosystem modelling, while ecosystem condition is inferred by benchmarking observations against multiple ecologically comparable reference ecosystems. The modular workflow supports alternative modelling approaches, analytical configurations and calibration strategies while maintaining consistency within a common reference-based framework.
4. We demonstrate the workflow using a continental-scale application for Australia (Supporting Information), generating spatially explicit ecosystem condition estimates and multi-decadal time series from Landsat data aggregated at 90 m resolution. Results capture broad gradients of ecosystem modification, fine-scale spatial structure and temporal dynamics relative to a consistent long-term reference framework. By providing open-source implementations of the HCAS workflow, including the ClassicHCAS R package and accompanying workflow code, we provide a reproducible and scalable approach to ecosystem condition assessment from EO data that can be readily adapted to different regions, datasets and applications.