(1) A given spatial tree configuration is characterised by an
energy function Eold, which quantifies the difference
between the NNSS estimated from the plot core area and
the reconstructed pattern including the off-plot neighbours.
(2) To improve the reconstructed tree pattern, one of the following
actions is carried out: Either (a) a randomly selected tree
is moved to a new random candidate location or (b) a randomly
selected pair of trees with the same species is
selected and their stem diameters are swapped (Lewandowski
and Gadow, 1997) or (c) a new tree with corresponding
species and stem diameter randomly selected from the original
data is added to the point pattern or finally (d) a randomly
selected tree is deleted from the point pattern.
(3) If Enew < Eold, the action randomly selected from (a)–(d) in (2)
is accepted and finally implemented. Enew is then set as Eold.
Otherwise the action randomly selected from (a)–(d) in (2) is
rejected.
(4) Steps 2 and 3 are repeated until (a) a maximum number of
50z (with z – initial number of trees in the reconstruction
area) is exceeded or (b) the current energy function measure
Eold falls below a pre-set energy level of Eold = 106. Alternatively
(b) has a further option, i.e. a reconstruction simulation
is re-started if the number of current iterations
exceeds 10,000. This alternative helps to save computation
time.