Eukaryotic genomes are typically full of long terminal repeat-retrotransposons (LTR-RTs). The mobility of LTR-RTs alters genome structure and can thereby generate phenotypic novelty by altering gene function. LTR-RTs have a well-defined structure which makes them particularly suitable for computational detection. This fact has permitted the development of a diverse range of scientific softwares and pipelines for their de novo annotation through structural identification. We have tested some of the existing softwares and found a high discordance between their outputs. Moreover, visual inspection of the datasets showed that these tools identify a certain amount of false positive sequences. To overcome these limitations we introduce LTRpred2, which combines several methods for de novo structural detection of LTR retrotransposons to enhance annotation sensitivity and select only the most confident candidates using protein domain search. LTRpred2 also implements functions to perform a vast array of analyses including metagenomic studies comparing retrotransposon family activity across kingdoms, copy number quantification, retrotransposon family clustering, etc. We use LTRpred2 to re-annotate and explore the LTR-RT landscape across genomes of the entire Eukaryotic Tree of Life in order to uncover the structural biodiversity of these TEs and find novel correlations with genomic and environmental traits.