I’ve done a complete run using the WA basecase using the Windows-modified script that I archived a couple of days ago. From the output of this run, I extracted the results of the AD1400 step and made a stepwise splice, which I’ve saved here together with the MBH98 stepwise reconstruction. This may be of interest to some people. Here is a smoothed version of the MBH98 stepwise reconstruction and the WA stepwise reconstruction. While it substantially emulates MBH98 results (as did our emulation), you’ll notice the divergence in the problematic early 15th century, which I also had a problem with. There’s no reason why exact replication should not be possible. Obviously this replication does not deal with issues like bristlecone pines; this is just benchmarking.
Figure 1. Stepwise NH Temperature Reconstructions from MBH98 Data. Black – MBH98; red -WA05. Both smoothed with 25-year filter.
The largest difference between the WA-version and the MBH-version exceeded 1 sigma. WA do not venture into the calculation of confidence intervals. However, it is a little disquieting that a supposed replication using the same data and methods uses up over 1 of the 2-sigma confidence interval merely in arithmetic. At a minimum, the error bar should be expanded to be at least the outer envelope of both calculations. I pointed out the particular problem in emulating MBH98 early 15th century results and it’s reassuring to see that WA have exactly the same problem (although their advertising doesn’t disclose this.)