A Powerful New Tool for Iowa Climate Estimation from Flora Distribution

It is a self-evident fact that a plant’s potential for occupancy of a habitat is climatically limited. The thought of a palm tree in the desert is absurd - the sight of one would be bewildering. The variable distribution of plants is a phenomenon which all humans encounter in their daily lives, and leads to our basic understanding of the concept - a plant will only grow where it is able. Underlying this simple observation are the mechanisms of natural selection which allow species to diversify and occupy different niches, thus avoiding potential competitors. The connection is so strong between a plant and its climatic range that one could easily guess the climate of a place without ever physically feeling the climatic conditions - it only takes a plant or two for us to imagine what a place is like. Thus, pictures of palm trees conjure up thoughts of subtropical coasts and island habitats with warm year round conditions, while pictures of cacti might lead us to imagine an arid area with high mean annual temperature and little rainfall. Now think about this - if you were given a picture of a landscape which featured palm trees, and told that the picture was taken 10, 20, 50 thousand years ago, you would still suppose a specific climate to the area, despite the chronological distance (which I would like to coin as a neologism, “chronodistance”, because heck if that doesn’t sound cool.)

Such postulations reside under the heading of ECV, or Estimation of Climate from Vegetation.The general underlying principle behind ECV would state that plant species perform better in specific climates, resulting in a patterned correlation between species distribution and climatic variables such as temperature, precipitation, etc. Because of the complexity of floral climate ranges, the requisite mountains of data and the variability of climates across all scales, statistical models are a necessary medium for testing hypotheses regarding the connections between climate and vegetation in the real world. The use of statistical models to associate plant species distribution and climate variation has become increasingly common within scientific literature (at least common enough to have passed the acronym threshold). With fossil pollen data and plant fossils from packrat middens, such models can be (and have been) employed to estimate geographical climate variation on multiple scales across the past, and may easily be site specific (I’m lookin’ at you, Iowa).

Enter Harper and Nixon (2015) with CRACLE, Climate Reconstruction Analysis using Coexistence Likelihood Estimation. While CRACLE is but one of many ECV models, it is regarded (agreeably) by the authors as somewhat superior to its earlier counterparts for reasons which vary with comparison. One of the most appealing initial details is that CRACLE is compatible with publicly available data on plant distribution and climate variation. It is a simple recipe list which CRACLE requires: a site-specific species list and a climate profile for each species are, essentially, all that is needed. Most important, the former may easily be supplied by fossil data. This reduce the site-specificity to some degree, but not drastically enough, it seems, to limit the use of CRACLE for paleobotanical studies.

There are two basic types of ECV: taxonomic, which correlate taxa and climates, and physiognomic, which correlates leaf shape characteristics with climate. CRACLE lies in the family of taxonomic ECV, often allowing the use of free georeferenced species data in lieu of actual data sampling - a boon. While seemingly more accurate than previous taxonomic methods, physiognomic ECV is less practical as it requires extensive data acquisition regarding leaf form and abundance, and has been criticised because physiognomic correlations with climate may be geographically variable. Furthermore, for paleobotanical modeling, it is much harder to get an ample dataset of physiognomic characteristics than it is to simply see who was around.

Harper and Nixon’s study put young CRACLE through the ringer in order to test its accuracy in climate prediction compared with other ECV models. They used data on 4,000 species to infer the climate of 165 sites across the world (barring Antarctica and Europe). CRACLE’s hypothesis states that the probability of the occurrence of a single species within an area (ascribed a climate variable) can be predicted if the probability of occurrence in relation to that variable is known (See “Figure 1”). Initial climate parameters were on a rasterized scale of 0.416 degrees (~4.6km) including mean annual temperature, average annual max/min temperatures, mean annual precipitation, and the mean precipitation of the three wettest and driest months. To test the hypothesis, climate tolerance profiles were defined as the range limits of these variables that any one species was found in. To evaluate CRACLE’s performance, Harper and Nixon used the good old Pearson’s R to compare CRACLE climate estimates to the actual climate data from Worldclim. They also ran smaller scale CRACLE analyses on The Harvard Forest and Barro Colorado Island, to test for the effects of sample taxa quantity on accuracy.

Harper & Nixon found that CRACLE demonstrates more accuracy than previous ECV models, both physiognomic and taxonomic. CRACLE had an average deviation of 1.5C from the Worldclim data, whereas other models deviated up to 5C. Overall, CRACLE showed more accurate mean error and narrower estimates for mean annual temperature. In addition, the small scale trials demonstrated that CRACLE’s accuracy is not dependent on a large number of species, converging on the actual mean near 40 species (“Figure 5”).

So, CRACLE is an excellent method for estimating climate through vegetation, perhaps the most accurate to date, with the bonuses of requiring minimal field sampling and an achievable number of input taxa. There are implications here for Iowan flora and climate which should be discussed. Despite Harper and Nixon’s use of CRACLE on a global scale, nothing in their study puts a spatial limit on a geographical area of Iowa’s size. The authors stress the potential use of CRACLE in paleobotanical studies, although there may be issues regarding the climate tolerance profiles of extinct species - there do not seem to be precise ways of measuring a species’ historic climate niche change, and any assumptions or associations to the nearest living relative may be useless in timeframes beyond the Pleistocene. Moreover, while a species’ presence in the past may be confirmed or denied through fossil or pollen records, the lack of exactness of georeferenced individuals may reduce CRACLE’s accuracy.

Yet CRACLE is clearly a powerful ECV model and should not be discounted for paleobotanical studies, considering other ECV methods face many of the same hurdles. In this author’s opinion, it is, at the time of writing, across-the-board the best method for paleobotanical ECV studies. Although field sampling and sample dating are required for paleobotanical studies, CRACLE is still less labor intensive and more accurate than other ECV models, with the bonus of requiring a relatively low number of input taxa. In Iowa, it would be possible to use the pollen record to reconstruct potential climates for whenever fossil pollen represents enough to taxa to perform CRACLE analysis, as long as the timeframe is not substantial enough to possibly feature altered climate niches among relative species. It seems that, using CRACLE, we may get one step closer to cracking the case of Iowa’s climate history.

Works Cited:

Harbert, R.S., and Nixon, K.C. 2015. Climate reconstruction analysis using coexistence likelihood estimation (CRACLE): A method for the estimation of climate using vegetation. American Journal of Botany 102(8): 1277-1289.