Skip to main content
Dryad logo

A meta-analysis of biological impacts of artificial light at night

Citation

Sanders, Dirk et al. (2020), A meta-analysis of biological impacts of artificial light at night, Dryad, Dataset, https://doi.org/10.5061/dryad.wpzgmsbjn

Abstract

This is a database of published studies, that measuered how the exposure to artificial light at night impacts the physiology, daily activity patterns and life-history traits. The data were collected using a systematic review with searches in Web of Science and Scopus. We also provide the R-code that was used to analyse the dataset with meta-analytic models in MCMCglmm.

Methods

Literature search. We identified relevant literature using keyword searches in Web of Science (we used “All databases” including Web of Science Core Collection, BIOSIS Citation Index, KCI-Korean Journal Database, MEDLINE, Russian Science Citation Index and SciELO Citation Index) and Scopus, finding any available papers published until 22 October 2019. We used the terms: "TS= (("Artificial light* at night" OR "Light* pollution" OR "Light* at night" OR "night time light*") AND ("species" OR "ecosystem*" OR "ecological commun") AND ("abundance" OR "behaviour" OR "richness" OR “reproduction" OR "mating" OR "*diversity" OR "composition" OR "predation" OR "herbivory" OR "activity" OR "timing" OR "physiology" OR "flight to light*" OR "melatonin" OR "development" OR "trophic" OR "biomass" OR "pollination"))”. After removing 352 duplicates, combining the searches resulted in 614 publications that were screened for inclusion criteria. To be included in the meta-analysis, studies needed to (1) test for ALAN effects on organisms either in the field or the lab; (2) have a control group that was exposed to natural light levels at night (or a dark control) and treatment groups with exposure to ALAN up to 100 lux - studies with higher levels were excluded as these are unlikely to occur in the field; (3) have at least 2 replicates per treatment; and (4) contain data on means, an estimation of variation and sample size. If only box plots were presented, we extracted the median and interquartile range. This resulted in 126 papers, with a total of 1304 effect size measures.

Effect size categorizing. We categorised the effect size measures into five different main groups: response to exposure to artificial light at night of (i) organismal physiology, (ii) phenology, (iii) organismal life history traits, (iv) activity (e.g. daily diurnal, nocturnal activity), or (v) populations and communities. We selected subcategories within each of the five major categories that we think describe the dataset best. For each subcategory to be included in the analysis it needed to have data that were extracted from at least five different studies. The subcategories were gene expresssion, hormones, immune response, stress response, gland stucture, phenology, sea finding in turtles, predation risk, body size, cognition, feeding, predation, reproductive output, activity cessation, activity onset, diurnal activity duration, nocturnal activity duration, abundance, bat activity and diversity. 

Analysis. The meta-analysis was conducted in R version 3.6.0 using the packages metafor and MCMCglmm. To account for study level non-independence due to multiple measurements per study, “Study” was included as a random effect. The MCMC chain ran for 150,000 iterations, and it was sampled every 50 iterations with the first 50,000 removed as burn-in to prevent autocorrelation among subsequent iterations. As we did not have any a priori knowledge on the distribution of our data, we used a flat prior: the inverse-Gamma prior (V = 1, nu = 0.002). Hedges’ d was used to compare measures of the variables between treatment and control. We assessed publication bias with testing for funnel plot asymmetry and p value hacking.

Funding

Natural Environment Research Council, Award: NE/N001672/1