La tipografía, reflejo de identidad

6.2.1 Simple regression model

Carbon dioxide emissions measured under laboratory conditions from simulated tillage (ST) and No-tillage (NT) in-situ soil cores (Chapter-4 section 4.3.1) were used to predict soil-C loss in field during the summer season. All replicates used to measure CO2

fluxes under laboratory conditions and in the field were used in the regression analysis. For field comparisons, each replicate used in the laboratory measurements represents the position of the static chamber installed in the field for both the RT and NT treatments.

In brief, to compare two tillage treatments: simulated-tillage (ST) and No-tillage (NT) and a non-disturbed (ND) control, in-situ soil cores (10 cm diameter, 10 cm depth) were collected at a soil depth of 0-10 cm from 4 different locations on each field site. Non- disturbed (ND) cores were collected before the start of field tillage operations and the NT treatment cores were collected over the slots immediately after the Cross Slot® No-tillage drilling. Each treatment comprised four replicates and each replicate was composed of three intact cores collected from one sampling location. Half of the total numbers of ND cores collected from each field site were broken up in the laboratory to simulate tillage treatment. The ST treatment was formulated from three ND soil cores by emptying the soil, breaking it into pieces, thoroughly mixing and packing it in a plastic container. Moisture contents in the cores and container as sampled from the field were maintained throughout the experiment by weighing the soil cores and spraying the required amount of deionised water onto the surface. Soil in-situ cores and plastic containers filled with soil were placed in closed base static chambers and CO2 measurements were taken on a daily basis at constant temperature

(230C) by placing a plastic petri dish containing 30 ml of 1M NaOH within each chamber for a 4 hour period. Carbon dioxide measurements continued for 92 days (Glen Oroua), 83 days (Tangimoana), 81 days (Kiwitea), 54 days (Feilding) and 99 days (Sanson) until the emissions subsided and the differences between the treatments became negligible.

Least squares’ fitting of the CO2-C evolved until days 2, 4, 6, 8, 10 and 12 with total

amounts of CO2-C evolved during the full incubation period were used to determine the

duration of incubation.

6.2.2 Two and three compartment decomposition models

Tillage induced CO2-C fluxes have been shown to be important over short durations

(Reicosky and Lindstrom 1993, Rochette and Angers 1999, La Scala et al. 2001, 2006) and are primarily related to the decay of the labile C fraction which has more rapid turnover than the total soil-C (La Scala Jr et al. 2009). Tillage also changes the soil conditions i.e. improved oxygen, temperature and moisture contents required for rapid decomposition (Six et al. 1998). Therefore, to build models that simulate CO2 emissions in field soils with

varying amounts of C in crop residues and more mature SOM, and to eventually accommodate the environmental effects of varying soil temperature and moisture, compartmented model structures based on the principle of the classical five compartment Roth-C model (Jenkinson 1990) were constructed.

6.2.2.1 Development of a temporally dynamic two compartment model using laboratory data

In the proposed two compartment dynamic model CO2 fluxes are expressed in terms

of C so they can be directly related to C present in the soil. In the two compartment model, the two compartments are an active crop or pasture residue pool and a more stable, soil C pool. It was assumed that the size of the active pool was negligible in comparison to the stable-C pool; therefore, the initial stable-C pool in the two compartment model was sized based on the measured total soil-C. Values for decay constants (for active and stable-C pools) were found by iteratively varying each value to maximize the coefficient of determination between the predicted (modelled) and observed 12 days daily flux values during laboratory incubations. After several iterations using Microsoft Excel 2010 for the five different soil types, standard size of active pool and decay constants for active and stable-C pools were fixed as described:

The two compartment model to describe the daily flux rate of CO2 from tilled soil:

Fn = (Cn x kc) + (An x ka)………...………..(Eq. 6.1)

Where:

Stable-C pool (Mg C ha-1) is represented as C (pool-C), the initial stable-C pool value was determined by multiplying total-C concentration by soil bulk density and its subsequent values i.e. Cn on any day n was determined using the following formula:

Cn= [C (n-1)–(C (n-1) x kc)]

The most labile or active-C pool (crop or pasture residues Mg C ha-1) is represented as A (pool-A). An on any day n was determined using the following formula:

An= [A (n-1) - (A (n-1) x ka)]

The decay constants kc and ka were estimated experimentally based on the best fit of the

model to measured CO2-C fluxes during the 12 day laboratory incubation of the simulated

tillage treatment of each of the five soils used in the laboratory incubation study (Chapter-4). The model was then used to predict emissions for the full term incubations.

The initial assumptions were that the pool-C decay constant (kc) was 0.0005 per cent of the

initial size of pool-C per day and was assumed to be constant for all the soils irrespective of soil type. For example, if the initial size of the stable-C pool for a particular soil is 44.4 Mg ha-1, it will decay at 0.00022 d-1.

The size of pool-A was based on the amount of C lost as CO2 during 12 days incubation

from disturbed/simulated tilled soils under controlled laboratory conditions. For pool-A, the decay constant (ka) was fixed as 0.10 d-1 and was assumed to be constant for all the soils

irrespective of the soil type. For the No-tillage treatment the decay constant for pool-A was allowed to vary in the two and three compartment models.

The daily decay constant values equate to < 0.2 per cent of the stable-C pool which is similar to the values used in the models calculating annual C change (Bayer et al. 2006; Leite et al. 2009). The fast pool decomposes very quickly and cannot be equated with annual models.

The two compartment model does not allow the transfer of C from the stable pool to the labile pool for the period of decomposition. To predict the CO2 emissions from the

decomposition of crop residues over a short period of time it may not be necessary to include the C transfer between pools.

Variations to a three compartment model with and without temperature and moisture are discussed in the Results and Discussion section 6.3 of this chapter.

6.2.2.2 Model efficiency (ME) also known as one of the expressions of R2 (coefficient of

determination) in non-linear fitting evaluations, was calculated by the following formula as stated by La Scala Jr et al. (2009):

ܯܧ ൌ ͳ െσ ሺܨ௧௢௕௦ െ ܨ௧ ௣௥௘ௗ_ሻଶ ௡ ௧ୀଵ σ ሺܨ_௧௢௕௦_{െ ܨത} ௧௢௕௦ሻଶ ௡ ௧ୀଵ Where: ܨ_௧௢௕௦ _{is the observed CO} 2-C flux

ܨത௧௢௕௦ is the mean of observed CO2-C fluxes throughout the measurement period

ܨ_௧௣௥௘ௗis the predicted CO2-C flux

Model efficiency/R2 will vary between minus infinity and 1 with higher values (closer to 1) indicative of superior performance.