Changes from 991214

What I did was basically take the full 9-species parameter suite from the Sortie papers and try to resolve the discrepancies in results. I was shooting for 3 points of congruence:

  1. Basal area of between 30 and 70 m2 per hectare. (Sort of successful.)
  2. Population density of 1000-3000 stems per hectare. (Can arrange that.)
  3. Correct species succession behavior. Not Even Close. Beech and Hemlock behavior are especially problematic.
Anyway, this unearthed some bugs, and recommended some changes, though the effort did not reproduce Sortie's results. The primary accomplishment was to get the forest population density and basal area more-or-less consistent with the published data the Sortie papers compared their results to. The results of the basic "Half Hectare" scenario are now only a bit higher than published values for basal area per hectare, and low on population. I know a way to correct both of those, at the expense of speed and correct forest succession. So, this base scenario is a trade-off.

In the interest of load speeds, the Sortie resolution suite du jour is elsewhere:

Sortie-like 9-species version.

The changes to the student (3-species) Woods are:

  1. Basal area calculation reported square centimeters. Now reports square meters.
  2. Trunk aversion used centimeters as meters. Fixed, and added new thinning factor (settable in the "All Options" scenario) to tune the multiplier on trunk radius to use for killing saplings. The default is 10 x tree radius. Setting this value lower results in higher population density. Setting it to zero turns off trunk aversion.
  3. Shadow-casting is now done every 5th year instead of every year, for speed. This seems to have no adverse effect on results. Crown-cropping, though slow, works significantly worse if done less often, and results in higher population, thus no speedup. So crown cropping remains a yearly event.
  4. The AllTrees Fragments statistic now uses a 10x10 array of fragments.
  5. Several changes were made to the "Half Hectare" scenario, with analogous changes to the others:

    1. This scenario was "One Hectare", but having (correctly) increased the population, the scenario was painfully slow. The half hectare gets results proportional to one hectare. Initial seedings were adjusted to match the decrease in area and the initial seedings of 225 trees per hectare used by Sortie.
    2. Changed the Hemlock mortRandom parameter from .0025 to .005. This makes the simulation take longer to achieve Hemlock succession. However, the deathrate (mortRandom in this case) appears to directly control the forest basal area. (Makes sense....) Note that a mortRandom here of .01 (matching Birch and Maple) would generate exactly the right basal area for this patch of forest. But Hemlock would never win the forest succession. So, this value is a compromise. The basal area is high, and Hemlock succeeds Birch.
    3. The maxCrownRadius parameter value changed slightly on Sugar Maple and Yellow Birch to exactly match the values I used for the Sortie-alike scenario. There's no noticeable effect.
I also experimentally added a second, indirect light scheme to the shadow scheme for calculating light indices. This slows things down enormously, without any visible improvement. So though the code remains and can easily be turned back on programmatically, the option is unreachable from the user interface.


Ginger Booth, July 13, 2000