First, a quick note for what happened since the last post. Autumn came, the Oregon visit ended. Then winter came. Now, winter is still around - at least here in Vienna. From my side activities for the iLand project were somewhat reduced during the last couple of month - this is on one side a pity, because I definitely enjoy working on iLand a lot. The second side, though, is that everything worked pretty well despite the heavy testing Rupert did. So much to the first part of bragging ;) Today, I found a cry for help from Rupert in my mail box. He described the issue he stumbled into while he was testing iLand (namely strange behavior of doug fir). He added - and that is the remarkable point - some additional information. The mail attachment consisted of:
Notably, the bug was detected by actually watching the visual representation of the forest in iLand. This points to the importance of proper visuals - it's all about the unrivaled power of the human visual system! And the video is IMHO an appropriate way to communicate visual impressions... Second, using the project-folder I was able to duplicate the exact simulation setup within minutes. This nicely confirms some design decisions like project specific XML files, the directory structure or the omitting of a central run-database. Well, that was the second part of bragging...
Lest I forget: the bug. As we introduced variable class widths for the storage of LIPs, I forgot to adapt the code for the highest dbh-classes. Eventually, the linking between trees and their respective LIP went bananas and super-large LIPs were assigned to small trees. Setting up 4000 small trees therefore resulted in an average light-influence-field value of about 10e-11. The scaling of LIF -> light response then increased small differences between the LIF levels of individual trees enormously which directly led to the observed behavior. Modesty is easily restored in the light of such stupid bugs...