Insect decline: a global database under the microscope shows the urgent need to review the evaluation by scientific journals

In the face of the ecological crisis, databases are multiplying to measure trends in biodiversity, but they are not systematically evaluated. Laurence Gaume of the Amap laboratory (UM, CNRS) and Marion Desquilbet (TSE, INRAE) took a closer look at InsectChange, published in Ecology, compiling time series of insect abundance and biomass on a global scale. Their comprehensive analysis highlights over 500 errors that call into question the results obtained from this database, particularly those of the meta-analysis published in Science in 2020. It also provides essential elements for improving InsectChange. Their study, recommended by Peer Community in Ecology and published in Peer Community Journal on October 8, 2024, points to the quality problem of large databases. While opening up methodological avenues, it calls on scientific journals to put in place protective measures against these deleterious effects on science and knowledge.

Nepenthes, a butterfly of the Lycaenidae family, on a Nepenthes flower bud in northern Borneo © Michaël Guéroult/INRAE.

Insect decline raises major environmental, economic and societal issues. A new scientific publication just published in Peer Community Journal identifies over 500 problems in the global insect temporal database InsectChange, published in Ecology in 2021.

This database was based on a meta-analysis published in Science in 2020, according to which insect decline was not as great as previously thought, and that agriculture was not one of the causes of this decline. Despite international criticism from 65 scientists, only a minimalist erratum was published, leaving the results of this meta-analysis unchanged and highly publicized, and conveying a reassuring message to the general public.

This publication by Laurence Gaume and Marion Desquilbet uncovers a multitude of new problems and deciphers their origin and nature (erroneous counts transmitted from one database to another, sampling bias, non-standardized units of measurement, data from experiments, inadequate geographic coordinates for sampling to measure the impact of agriculture or urbanization using satellite land cover data, etc.). It concludes that these 553 problems, categorized as errors, inconsistencies, methodological problems and information deficiencies, call into question any insect trends estimated from InsectChange, and make it impossible to test from this database whether land use is likely to explain the trends observed.

So, for example, many freshwater datasets actually include all aquatic invertebrates, and increases in "insects" turn out to be invasive mussel blooms. By correcting these data, the two scientists find that the freshwater insect trends published in Science have been grossly overestimated. Another type of problem is that half of the InsectChange datasets are not representative of insect dynamics under natural conditions. This is the case of dragonflies colonizing experimental ponds created by an English scientist to study them. These data are incorporated without mentioning the experimental context, artificially leading to an increase in these insects. Furthermore, the transformation of these non-standardized data, as done in the meta-analysis, compromises the comparison of time series slopes and the estimation of overall insect trends. Finally, the coverage of agricultural land around the sampling areas is grossly overestimated, partly due to inaccurate geographical coordinates in two-thirds of the datasets, leading to crop areas being ruled out as a possible cause of insect decline.

The need for systematic evaluation of ecological reference data

This observation raises a number of issues:

  • It is vital to ensure the quality of global ecological databases, which are multiplying in the context of declining biodiversity. In particular, this publication provides detailed information crucial to improving InsectChange and enabling its informed use.
  • This publication develops a reproducible analysis method that can inspire the development of systemic evaluation grids for database quality.
  • It calls on scientific journals to improve pre-publication peer review and post-publication feedback, especially for publications based on large datasets, and particularly for reputable journals, the preferred intermediaries for journalists.
  • Finally, she highlights the important role played by the non-profit organization Peer Community In, in particular their approach to open science and independent, transparent evaluation, which helps preserve scientific integrity and the quality of science.

References linked to the article :