Guest Post by Willis Eschenbach
“Out of passions grow opinions; mental sloth lets these rigidify into convictions”
Friedrich Nietzsche, 1844-1900
Anthony discussed the press release about the paper Past and ongoing shifts in Joshua tree distribution support future modeled range contraction. I didn’t have a copy so I wrote to the lead author, Kenneth Cole, to request one. He responded immediately and sent me a copy. My thanks to him, that’s science at its best. Other than the ritual obeisance to the climate models, the paper tells an interesting story about Joshua trees and the extinct Shasta ground sloths.
The Joshua “tree” is
a cactus not a tree, it’s one one of the famous Yucca tribe of cacti spiny spiky things that aren’t cacti according to folks who know better than I do. Calling it a tree is merely the other cactus’s Yucca’s polite way to try to make it feel better about its funny appearance and desolate condition. It grows where almost nobody else can grow, in a very restricted climate range in the American Southwest. Not too hot, not too cold, not too much rain or too little rain, just right.
The fruit of the Joshua tree is a seed pod that was a favorite food of the Shasta ground sloth. The sloth appears to have been the only major seed dispersal mechanism for the Joshua tree. Which is hardly surprising, since other than pack rats and ground sloths, there’ve never been many herbivores hanging out where the Joshua tree grows. It’s way dry in that corner of the US.
The black stars in Figure 2 (from their paper) show the current location of Joshua trees, in and around the Mojave Desert in the hot Southwest.
Figure 2. The authors used rainfall and temperature maps to determine where Joshua trees might possibly grow, with green showing the most favorable climates. Map shows the lower parts of California (left) and Nevada (upper middle), along with a section of western Arizona.ORIGINAL CAPTION: … (A) Suitable climate model for Joshua tree created with mid 20th century (AD 1930 1969) PRISM mean precipitation variables and extreme mean monthly temperature events.
Now, to start with their map is interesting. I mean, the Joshua trees are certainly not growing extensively in what is the best part of the their range according to the authors. Unfortunately, much of the area they say is best for Joshua trees is on Nellis Air Force Base, so there’s no information for large areas. On the other hand, some areas with orange or even red (low probability) have a number of Joshua trees. However, nature is never as neat as we’d like it to be, and they’ve done well to generally outline the range by climate variables. And if it becomes desirable to plant Joshua trees, we know where they’ll likely grow.
Fifteen thousand years ago, in the days of the now-extinct Shasta ground sloth, the geological evidence from pack rat middens and sloth dung shows that the Joshua tree was much more widespread. However, humans happened upon the North American scene around that time, and converted all of the ground sloths into ground slothburgers and barbecued them. Which was bad news for the Joshua trees (not to mention the sloths), because no one else had much taste for Joshua tree seeds. As a result, the range of the Joshua tree is much reduced from its former glory.
The authors also show that the current range of the Joshua trees is further north, and at a higher elevation, than in the times of the ground sloth. During the last ice age the range of the Joshua tree extended across other areas that are now too hot or otherwise unsuitable to support them. As it warmed at the end of the ice age, the Joshua trees retreated (and advanced) to their current locations.
The problem is that without the ground sloths, the Joshua tree’s only seed dispersal mechanism is pack rats. The authors estimated the rate of spread from packrats at two metres per year. This seemed low to me, but is explained in the paper. The missing parts of the puzzle were a) unlike sloths, the packrats only carry the seeds a maximum of about forty metres or so to their homes, and b) the trees take twenty years to reach maturity and produce seeds. Result … 2 metres per year rate of spread. Always more to learn.
So far, so good. And if they’d quit there, it would have been an interesting paper. But no, they had to bring in the climate models. Now, climate models are notoriously bad at predicting precipitation (rain and snowfall). So they figured they’d pick the best of the bunch, viz:
Future downscaled GCM projections
To assess potential future changes in Joshua tree’s suitable climate space we compared future projections from several GCM’s for the late 21st century (2070 2099; ~2X CO2). Five individual models and one ensemble of 48 runs of 22 models based upon the A1B carbon emission scenario were obtained from the Program for Climate Model Diagnosis and Intercomparison (PCMDI; AR4) archive (available online). The five individual models used were: Hadley Center for Climate predic tion (Hadgem1), Max Planck Institute for Meteor ology (Mpi_echam5), CSIRO Atmospheric Research (Csiro_mk3), National Center for Atmospheric Research (Ncar_ccsm3), and Centre National de Recherches Météorologiques (Cnrm_cm3). They were selected because they represent a wide range of future moisture availability conditions for southwestern North America (Seager et al. 2007), and they all were ranked within the top half (of 22 models tested) for their ability to hindcast 20th century precipitation seasonality within the southwestern U.S. deserts (Garfin et al. 2010). These models, especially the Hadgem1 and Mpi_echam5, outperformed most models in replicating the 1950 to 1999 AD geographic distribution of average seasonal precipitation (Garfin et al. 2010).
Then, once they had what they figured were the best five models giving a “wide range” of rainfall results, they ran them and tried to figure where the Joshua trees might live in the future. The models gave differing results, so the authors defined a threshold for suitability (18%). If three of the five models said a particular gridcell would be above the 18% suitablity threshold for Joshua trees, they called it an “area of agreement for future suitable climate (AAFSC)”.
Then they show what those 5 models (and the 22 model ensemble) said would be suitable areas for Joshua trees in the years 2070-2099. Which all sounds vaguely reasonable until we look at their results in Figure 3 …
Figure 3. Results for five models (B-F) and 22 model ensemble (G). Pink areas with thin black outline show current range of the Joshua trees. ORIGINAL CAPTION: … (B G) The Joshua tree future suitable climate model runs for late 21st century (AD 2070 2099 AD): (B) Hadgem1, (C) Mpi_echam5, (D) Csiro_mk3, (E) Ncar_ccsm3, (F) Cnrm_cm3, and (G) Ensemble (44 runs of 22 GCMs).
See, if those results were mine, I’d throw up my hands and say “Not ready for prime time”. Those models are all over the map. They said they picked the models to “represent a wide range of future moisture availability conditions”, but I wasn’t expecting that huge a range. One model is “everything’s fine”, another is “they’re all gonna die!” That’s so wide as to be useless, and these are among the best models.
I don’t see any way that an average of those, or an “AAFSC” of those (area of agreement for future suitable climate), has any meaning at all. One of the models shows a wide area of hundreds of thousands of hectares where the Joshua tree could live, and another shows no suitable area at all.
Finally, once again we have the problem I have called “Models all the way down“. While the model results are interesting, they’ve skipped a big step. I want to see a map, just like the maps above showing possible future distributions of Joshua trees. But I want one showing how well the individual models (and the 22 model ensemble) did at hindcasting the current distribution of the Joshua trees.
I mean seriously – before showing us the model forecasts for 2070-2099 Joshua tree distributions, shouldn’t they show us the model hindcasts for the 1970-1999 Joshua tree distributions? It wouldn’t rescue the paper, but at least it might give some reason to think the selected models might be better than throwing darts at a map of the Mojave region.
Because without that, it’s just models disagreeing and quarreling with each other, all the way down …