Fables and foibles: a critical analysis of the Palaeoflora database and the Coexistence approach for palaeoclimate reconstruction
The \"Coexistence Approach\" is a mutual climate range (MCR) technique combined with the nearest-living relative (NLR) concept. It has been widely used for palaeoclimate reconstructions based on Eurasian plant fossil assemblages, most of them palynofloras (studied using light microscopy). The results have been surprisingly uniform, typically converging to subtropical, per-humid or monsoonal conditions. Studies based on the coexistence approach have had a marked impact in literature, generating over 10,000 citations thus far. However, recent studies have pointed out inherent theoretical and practical problems entangled in the application of this widely used method. But so far little is known how results generated by the coexistence approach are affected by subjective errors, data errors, and violations of the basic assumptions. The majority of Coexistence Approach studies make use of the Palaeoflora database (the combination of which will be abbreviated to CA+PF). Testing results produced by CA+PF studies has been hindered by the general unavailability of the contents in the underlying Palaeoflora database; two exceptions are the mean-annual temperature tolerances and lists of assigned associations between fossils and nearest-living relatives. Using a recently published study on the Eocene of China, which provides the first usable insight into the data structure of the Palaeoflora database, we compare the theory and practice of Coexistence Approach using the Palaeoflora database (CA+PF). We show that CA+PF is riddled by association and climate data error. We reveal the flaws in the application of CA, which is often in stark contrast to the theory of the method. We show that CA+PF is highly vulnerable against numerous sources of errors, mainly because it lacks safeguards that could identify unreliable data. We demonstrate that the CA+PF produces coherent, pseudo-precise results even for artificially generated, random plant assemblages. Alternative MCR-NLR methods can surpass the most imminent deficits of CA, and may be used as a stop-gap until more accurate bioclimatic and distribution data on potential Eurasian NLRs, and theoretically- and statistically-robust methods will become available. Finally, general guidelines are provided for the future application of methods using the mutual climatic range with nearest living relatives approach when reconstructing climate from plant fossil assemblages.