Monitoring through many eyes: Integrating disparate datasets to improve\n monitoring of the Great Barrier Reef

Erin E Peterson, Edgar Santos–Fernández, Carla Chen, Samuel Clifford, Julie Vercelloni, Alan R. Pearse, Ross Brown, Bryce Christensen, Allan James, Kenneth R. N. Anthony, Jennifer Loder, Manuel González‐Rivero, Chris Roelfsema, M. Julian Caley, Tomasz Bednarz, Kerrie Mengersen

Abstract

Numerous organisations collect data in the Great Barrier Reef (GBR), but they are rarely analysed together due to different program objectives, methods, and data quality. We developed a weighted spatiotemporal Bayesian model and used it to integrate image based hard coral data collected by professional and citizen scientists, who captured and or classified underwater images. We used the model to predict coral cover across the GBR with estimates of uncertainty; thus filling gaps in space and time where no data exist. Additional data increased the models predictive ability by 43 percent, but did not affect model inferences about pressures (e.g. bleaching and cyclone damage). Thus, effective integration of professional and high-volume citizen data could enhance the capacity and cost efficiency of monitoring programs. This general approach is equally viable for other variables collected in the marine environment or other ecosystems; opening up new opportunities to integrate data and provide pathways for community engagement and stewardship.

Introduction

The Great Barrier Reef (GBR) spans approximately 348,000 km², presenting a monumental challenge for comprehensive monitoring. While professional programs like the AIMS Long-term Monitoring Program (LTMP) provide high-quality "gold standard" data, their spatial coverage is inherently limited by logistics and funding. Conversely, a wealth of data is collected by universities, NGOs, and citizen scientists, but these disparate datasets are rarely integrated into official reporting due to variations in methodology, scale, and perceived quality.

This paper addresses the core challenge of "data integration": how to combine precisely measured professional transects with diverse, often noisier, data from other sources without compromising statistical rigor. By developing a mechanistic weighting system, the researchers demonstrate how fragmented data can be unified to provide a more accurate and comprehensive map of reef health.

Methodology and Data Integration

The researchers employed a weighted spatio-temporal Bayesian Beta regression model using a Stochastic Partial Differential Equation (SPDE) approach (implemented via the r-INLA package in R). This framework allows for efficient modeling of large-scale spatial correlations, effectively "filling in the gaps" between survey sites based on proximity and environmental covariates.

A core innovation of this study is the weighting scheme used to integrate datasets from five primary sources: professional surveys (LTMP, MMP, UQ-RSRC), large-scale philanthropic efforts (XL Catlin Seaview Survey), and citizen science (Reef Check Australia). Weights were not arbitrary but were calculated based on specific sample characteristics.

Table 1: Summary of data sources and the parameters used to calculate weights for the integrated model, including image extent, annotation method, and sample size (N).

As shown in the table above, weights ($w_{js}$) were assigned based on:

  • Image Extent: The physical area captured in a sample.
  • Annotation Method: Distinguishing between manual and automated classification.
  • Observer Accuracy: For citizen scientists, accuracy was calibrated against professional standards using validation images.
  • Sample Volume: Up-weighting cells where many images were aggregated.

The model also accounted for environmental pressures, including cyclone exposure, bleaching history, sea surface temperature anomalies, and reef zoning (e.g., "No Take" zones).

Key Results and Insights

The integration of "non-gold standard" data significantly enhanced the model's performance without diluting its scientific integrity. Key findings include:

  • Predictive Boost: The "All Data" model outperformed the "Professional Only" model by 42.76% in terms of Root Mean Square Prediction Error (RMSPE).
  • Correlation Jump: The correlation between observed and predicted coral cover rose from essentially zero (-0.056) in the professional-only model to 0.7431 in the integrated model.
  • The "Power of Many": While a single citizen scientist's classification might have low weight, high volumes of such data eventually provide a very strong, reliable signal. Adding this data helped the researchers see where the coral was, without misidentifying why the coral was changing (e.g., bleaching or cyclones).
  • Realistic Uncertainty: Interestingly, adding more data revealed high spatial variability that professional surveys had missed, leading to more honest, albeit sometimes higher, estimates of uncertainty in specific regions.

Discussion and Future Work

This framework provides a ready-made "plug-and-play" system to integrate diverse data types, including future AI-derived classifications. By accounting for the specific biases and error rates of different algorithms or observers, researchers can leverage the speed of automation alongside the nuance of human observation.

Future iterations could refine the model by treating weights as stochastic variables (priors) within the Bayesian framework or using Item Response Theory (IRT) to account for varying task complexity—recognizing that identifying coral in murky water is inherently harder than in clear conditions.

Conclusion

By effectively integrating professional and high-volume citizen science data, this approach enhances the capacity and cost-efficiency of large-scale monitoring programs. This general methodology is equally viable for other variables in marine or terrestrial environments, opening new opportunities for community engagement and improved ecosystem stewardship.

Summary updated

Abstract

Numerous organisations collect data in the Great Barrier Reef (GBR), but they\nare rarely analysed together due to different program objectives, methods, and\ndata quality. We developed a weighted spatiotemporal Bayesian model and used it\nto integrate image based hard coral data collected by professional and citizen\nscientists, who captured and or classified underwater images. We used the model\nto predict coral cover across the GBR with estimates of uncertainty; thus\nfilling gaps in space and time where no data exist. Additional data increased\nthe models predictive ability by 43 percent, but did not affect model\ninferences about pressures (e.g. bleaching and cyclone damage). Thus, effective\nintegration of professional and high-volume citizen data could enhance the\ncapacity and cost efficiency of monitoring programs. This general approach is\nequally viable for other variables collected in the marine environment or other\necosystems; opening up new opportunities to integrate data and provide pathways\nfor community engagement and stewardship.\n
BibTeX
@article{peterson2020monitoring,
  author = {Erin E Peterson and Edgar Santos–Fernández and Carla Chen and Samuel Clifford and Julie Vercelloni and Alan R. Pearse and Ross Brown and Bryce Christensen and Allan James and Kenneth R. N. Anthony and Jennifer Loder and Manuel González‐Rivero and Chris Roelfsema and M. Julian Caley and Tomasz Bednarz and Kerrie Mengersen},
  title = {{Monitoring through many eyes: Integrating disparate datasets to improve\textbackslash\{\}n monitoring of the Great Barrier Reef}},
  journal = {Environ. Model. Softw.},
  year = {2020},
  doi = {10.1016/j.envsoft.2019.104557},
}

Loading related content...

Figure