A study applying random forest, generalized linear models, and hybrid geostatistical methods to predict sponge species richness (SSR). The research addressed variable and model selection issues, mapping non-linear relationships and generating spatial distributions of SSR. The dataset is associated with a 2017 publication in Environmental Modelling & Software and is hosted by the Australian Ocean Data Network.
Use Cases
- Modeling non-linear relationships between sponge richness and environmental predictors based on the described methods.
- Comparing predictive accuracy of random forest and generalized linear model hybrids for count data.
- Generating spatial distribution maps for marine species richness based on the study's high-accuracy results.
- Investigating the effect of highly correlated predictors on model performance as mentioned in the findings.
Strengths
- Methodology is detailed in a peer-reviewed publication (Environmental Modelling & Software, Volume 97, 2017).
- Study explicitly addresses variable and model selection challenges for predictive accuracy.
- Findings include the association of high sponge species richness with hard seabed features.
Limitations
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count and dataset size are unknown, which may limit suitability assessment.
- Data freshness should be verified; metadata shows a last updated date of 2026-06-27.
Provenance
- Source
- Australian Ocean Data Network, associated with a published research article.
- Collection Method
- Likely contains model outputs and environmental predictors from a spatial modeling study.
- Time Range
- null
- Freshness
- Last updated 2026-06-27 19:02:55.728714; freshness should be verified.
- Geography
- Likely covers marine areas studied by the authors, specific location not stated.