In natural ecosystems, soil organic carbon (C) is derived almost exclu
sively from the residues of plants growing in situ. In agroecosystems,
it has at least two origins: one is the remains from the previous nat
ive vegetation, and the other is the remains of the crop and the decom
position of its residues. Where vegetation has changed from plants wit
h the C3 photosynthetic pathway to C4 pathway or vice versa, changes i
n the natural abundance of C-13 in soil organic matter (SOM) over time
can be used to identify sources of organic C in soil and to determine
the turnover rate of SOM. For example, large areas of C3 tropical for
est have been replaced with C4 pasture or cropland. Changes in the del
ta(13)C values of soil organic C in these areas reflect soil organic m
atter turnover rate, and provide insight regarding the functional role
of tropical ecosystems in the global C cycle. This paper illustrates
how the stable isotope C-13 can be used to estimate SOM turnover rates
and the sensitivity of different models and different model parameter
s, using a chronosequence of forest and pastures of different ages fro
m the Brazilian Amazon. A single-compartment exponential decay model a
nd a two-compartment model in which SOM was divided into stable and la
bile components yielded similar estimates of soil C turnover time at t
he surface but divergent estimates at depth. The one-compartment model
gave the least variable estimates of model parameters and turnover ti
mes and was also relatively insensitive to individual C stocks in sing
le pastures of a particular age. Estimates of soil stable and labile C
pools obtained using changes in forest soil delta(13)C With depth dif
fered from estimates obtained using the chronosequence. This suggests
that upon burning and pasture creation, a portion of the previously st
able soil C pool is rendered less stable. Model r(2) was a poor criter
ion for selecting an appropriate soil C turnover model to apply to chr
onosequence data. In the absence of substantial justification for segr
egating SOM into different compartments based on lability, modeling sh
ould be done with the simplest models possible. (C) 1998 Elsevier Scie
nce B.V.