Florian Hartig

← Publications

Machine learning and deep learning—A review for ecologists

Maximilian Pichler & Florian Hartig

Methods in Ecology and Evolution, 14(4), 994–1016 (2023)
Cite this
@article{pichler2023machine,
  author = {Maximilian Pichler and Florian Hartig},
  title = {Machine learning and deep learning—A review for ecologists},
  journal = {Methods in Ecology and Evolution},
  volume = {14},
  number = {4},
  pages = {994--1016},
  year = {2023},
  doi = {10.1111/2041-210X.14061},
}

DOI: 10.1111/2041-210X.14061
Citation statistics: 638 (Google Scholar) · 469 (OpenAlex), as of 29 October 2026

Google Scholar citations per year, as of 22 October 2026 (yearly counts are fetched separately from the total)

Funding

This study was supported through funding by Bavarian Ministry of Science and the Arts (Bavarian Climate Research Network, bayklif). Open access funding: Projekt DEAL.

View article (DOI) Preprint

Abstract

The popularity of machine learning (ML), deep learning (DL) and artificial intelligence (AI) has risen sharply in recent years. Despite this spike in popularity, the inner workings of ML and DL algorithms are often perceived as opaque, and their relationship to classical data analysis tools remains debated. Although it is often assumed that ML and DL excel primarily at making predictions, ML and DL can also be used for analytical tasks traditionally addressed with statistical models. Moreover, most recent discussions and reviews on ML focus mainly on DL, failing to synthesise the wealth of ML algorithms with different advantages and general principles. Here, we provide a comprehensive overview of the field of ML and DL, starting by summarizing its historical developments, existing algorithm families, differences to traditional statistical tools, and universal ML principles. We then discuss why and when ML and DL models excel at prediction tasks and where they could offer alternatives to traditional statistical methods for inference, highlighting current and emerging applications for ecological problems. Finally, we summarize emerging trends such as scientific and causal ML, explainable AI, and responsible AI that may significantly impact ecological data analysis in the future. We conclude that ML and DL are powerful new tools for predictive modelling and data analysis. The superior performance of ML and DL algorithms compared to statistical models can be explained by their higher flexibility and automatic data‐dependent complexity optimization. However, their use for causal inference is still disputed as the focus of ML and DL methods on predictions creates challenges for the interpretation of these models. Nevertheless, we expect ML and DL to become an indispensable tool in ecology and evolution, comparable to other traditional statistical tools.

What the paper shows and why it matters

(AI-generated)

Machine learning (ML) and deep learning (DL, neural networks with many layers) are increasingly used in ecology and evolution, yet many ecologists see them as opaque, and their relation to classical statistical models is debated. Earlier reviews concentrated on deep learning (Borowiec et al. 2022; Christin et al. 2019), although not everything that holds for deep networks holds for older methods such as boosted regression trees. And ML is often assumed to be purely predictive, an assumption with a reasonable core, since the best predictive model need not be the causal one. What was missing was an account of ML as a whole and of why these methods work, and this is what the paper sets out to give.

The paper reviews machine and deep learning for ecologists as one field with shared principles. Its central argument is that ML gives up an explicit statistical model of how the data arise in favour of flexible algorithms whose effective complexity (how closely the model can follow the data) adjusts to the data. Ensembles of many simple models, such as random forests and boosted trees, offset individual errors. Penalties that shrink coefficients and settings tuned by cross-validation limit flexibility. In deep learning, overparameterisation combined with stochastic gradient descent is hypothesised to act as an implicit brake, which the paper presents as still debated. The paper offers this control of complexity as the explanation of the predictive advantage over statistical models, and of why classical ML does well on small datasets and data in table form, while deep learning excels on images and other unstructured data. A simulation shows where good prediction and causal insight part ways: with a strongly correlated confounder, the true causal model had the highest prediction error, while models with wrong effect estimates predicted better. The paper warns against reading feature importances or explainable-AI outputs as causal effects.

Figure 3. Development usage of ML algorithms (RF, DL, kNN, BRT, Ridge/Lasso, SVM) in literature from the E&E field (see Appendix S1.1 for more details about the trend analysis). From Pichler & Hartig (2023), Methods in Ecology and Evolution, licensed under CC BY 4.0. Click the image to enlarge.

The results mean that ML and DL are powerful tools for prediction, data processing and data analysis, while causal conclusions need care, extrapolation is as hard as for statistical models, and simpler models remain preferable where they suffice. The paper expects ML to become an indispensable tool in ecology and evolution, comparable to linear regression models or analysis of variance. Garcia-Quintas et al. (2023) cite it when discussing the opacity of deep learning and the interpretation of random forests.

References

  • Borowiec et al. (2022). Deep learning as a tool for ecology and evolution. Methods in Ecology and Evolution 13, 1640–1660. doi:10.1111/2041-210x.13901
  • Christin et al. (2019). Applications for deep learning in ecology. Methods in Ecology and Evolution 10, 1632–1644. doi:10.1111/2041-210x.13256
  • Garcia‐Quintas et al. (2023). Machine and deep learning approaches to understand and predict habitat suitability for seabird breeding. Ecology and Evolution 13, e10549. doi:10.1002/ece3.10549