Redefining the concept of ecological health: Assessment and prediction of health as model parameters at multiple spatial scales across a continental-scale river basin

Background/Question/Methods

Governments worldwide continue to monitor and manage stresses that expanding human activities inflict on ecosystems. Recent research aims to facilitate evidence-based environmental policies. Focus has been moving away from descriptive tools, towards holistic and increasingly model-based approaches for ASSESSMENT and PREDICTION of ecosystem health in response to environmental changes. Assessment emphasizes the appropriate interpretation of a suite of metrics (e.g., richness/abundance of bioindicator species); prediction may involve more than modeling alone. Which stressors at what scale drive the biotic response, and what constitutes an appropriate collective interpretation of multiple metrics across scales? E.g., what health conditions are implicated by a large weighted average of biotic metrics alongside abiotic attributes of frequent stream floods with occasional basin-wide droughts? Should health prediction necessarily be a three-step exercise, whereby individual metrics are predicted, combined, then interpreted for ecosystem health?

     We take a one-step modeling approach that integrates assessment and prediction of health, defined as a model parameter. Thus, health is quantitative, not directly observable, but estimated by fitting the model to biotic and abiotic data collectively. Our approach builds on the nonlinear hierarchical regression technique of latent health factor index (LHFI) modeling, tailored to the Murray-Darling Basin (MDB) which covers 14% of Australia.

Results/Conclusions

The present benchmark assessment protocol for the MDB is the Sustainable Rivers Audit (SRA). It maps theme-based metrics (fish theme, vegetation theme, etc.) to health scores for 2-4 altitudinal zones per river valley, with 23 valleys across the MDB. Spatially, SRA scores are reported at the zone and valley levels, and ecologically, at the theme and overall levels. Studies have shown poor performance of health predictions at certain scales using SRA scores, which can be insensitive to changes in hydrological attributes despite ground evidence of flow-induced impact on biota. In comparison, a single LHFI model integrates biotic and abiotic data across spatial and ecological scales. We show higher sensitivity of LHFI health estimates to environmental changes across the MDB at various scales. Because health at any given scale is a spatially continuous model parameter, health can be statistically interpolated at locations where biotic information is missing (e.g., unsampled sites due to cost constraints), and simultaneously the influence of abiotic attributes as drivers can be identified at different scales. We conclude that integrated ecosystem health assessment and prediction via LHFI modeling is a viable and cost effective approach with a potentially vital role in the advancement of evidence-based environmental policies.