2.4. Change point analysis and soil accretion rates
Change point (CP) refers to a point along a distribution of values of a
variable where values before and after the change point are significantly
different. The method for identification of a CP uses deviance reduction
along a distribution of points such that the sum of deviances on either
side of the CP is lowest compared to the deviance of the overall dataset.
The percent of error reduction associated with splitting the data is calculated
by an iterative process, moving along the data systematically and
identifying a point (change point) that minimizes the deviance (Qian
et al., 2003). A schematic depiction of CP as an indicator of the boundary
between RAS and native soils in treatment wetlands is presented in
Fig. 2.
We used a freely available program on the internet — SegReg, to determine
CP within 90% confidence interval using various soil variables
along depth profiles (Oosterbaan et al., 1990) (http://www.waterlog.
info/segreg.htm) This program uses the principle of deviance reduction
within a particular dataset by introducing one or more CPs whereupon
separate linear regressions are made. The intersection or meeting of
these linear segments represented the CP, which corresponded to the
‘boundary’ between RAS and pre-STA (native) soil. The depth along
soil profile where CP exists was equivalent to the thickness of RAS
layer. Four variables — soil bulk density (g cm−3
), TP (mg P kg−1
),
δ15N (‰) and δ13C (‰) were used for identification of CP using SegReg
program. An example of the output from SegReg program using the distribution
of four variables for one soil core is presented in Fig. 3. Other
physico-chemical properties such as TN (g N kg−1
), TC (g C kg−1
),
and nutrient ratios (N:P, C:P and C:N) did not vary sufficiently enough
along depth profile to indicate presence of a significant CP.