The optimal design of nature reserves has been a hotly debated topic f
or some years. One contentious issue has concerned the ideal shape of
reserves, with several researchers suggesting that reserves should be
as nearly circular as possible to minimize edge-related degredation an
d potential 'peninsularity' effects. This paper demonstrates an advant
age to non-circular reserves: given the spatial autocorrelation of env
ironmental conditions and species ranges, elongated samples should ten
d to capture a greater diversity of conditions and, consequently, a gr
eater number of species. To test this idea, samples of 16 cells, arran
ged as 4x4 squares, 2x8 rectangles and 1x16 lines were considered in m
apped plant distributions at spatial scales ranging from cells of I m(
2) to 2500 km(2). Overall, elongated samples tended to capture signifi
cantly larger numbers of species than did their square counterparts, a
s predicted. The degree of elongated sample advantage was very similar
at all but the smallest of the scales considered, suggesting that the
pattern of species turnover (beta-diversity) is nearly scale-independ
ent or fractal in nature. As the advantages of elongation are largely
scale-independent and the only documented disadvantages (edge effects)
are very scale-dependent, the optimal shape of a reserve for species
sampling should shift as a function of reserve size, with large reserv
es becoming increasingly elongated. (C) 1997 Published by Elsevier Sci
ence Ltd.