[Previous] [Contents] [Next]

2.6 The Integration of the Submodels

After the identification and definition of the submodels and their quantitative response to potential inputs, the quantity of those inputs themselves and the different connections between the submodels have to be modeled. This task was realized as follows. All secondary outputs were 'collected' in the STELLA-model in a compartment called 'Farm by-products' (Fig. 31).

Farm by-products can be looked at as a pool where all by-products from the farm enterprises are collected and then redistributed. The amount of by-products collected and recycled in the farm system is a measure of integration. This compartment plays an important role for developing and testing sustainability indicators. The intuitive idea is that the more feedback cycles are established within the farming system, the less resources, i.e. nutrients contained in by-products, are wasted or lost. This diminishes the required inputs in the form of fertilizers in order to sustain the productivity of the farm.

By-products from the rice compartment are a) rice-straw, which can be transformed into fertilizers by composting, b) rice bran, which is usually fed to the pigs but can also serve as input for the fish pond, c) broken rice, which is used as chicken food and d) waste.

After threshing, all harvested dry matter is distributed between certain yield components (Yoshida, 1981).

straw 60 %
grains 30%
hulls 10%
------------------------------
total 100%

The ratio economical yield (grains)/biological yield (straw + grains + hulls) is called Harvest Index. The Harvest Index (HI) is about 0.3 for traditional tall varieties and 0.5 for improved, short varieties (Yoshida, 1981). On the farms considered the HI is about 0.3 (Oficial, private conversations). After threshing, the grains and hulls are dried and milled which produces other by-products, such as rice bran and broken rice. After milling the components of the grains are

rice 50%
rice bran 36%
broken rice 2%
hulls 4%
leftovers 8%
----------------------------
total 100%

These data are taken from field trials conducted by Dalsgaard and Oficial (1995).

In the model rice bran can be fed to pigs (pigfooddemand, kg/d) and/or fish (Ricebranto_Fish, kg/d), whereas broken rice can be fed to chicken.

If the feeding rate (chickfeedrate or pigfeedrate, % ad Libitum food intake) of pigs or chicken is adjusted in a way that their food demand (chickfooddemand or pigfooddemand, kg/d) cannot be satisfied by rice bran or broken rice, then additional food has to be purchased (purchasedChickfood or purchasedPigfood, kg). The pig - or chickfooddemand is maximal the ad Libitum food intake as modelled by Eqn. 2.4.

In the model the food demand is set according to the feeding rate which is arranged by the user and not by available food resources.

purchasedChickfood(t)=

purchasedChickfood(t-dt)+(buyChickfood)*dt

buyChickfood = chickfooddemand-tochicken

chickfooddemand = chickfeedrate*NumberofChicks

Here tochicken (kg/d) is the rate by which broken rice is fed to the chicken. It is maximal the same as chickfooddemand, but when resources become less than what is demanded tochicken becomes less in the same way. That means when all, e.g. broken rice is fed to the poultry, tochicken becomes zero and commerial food has to be purchased at the same rate as chickfooddemand. For pigs this is modelled in the same way.

Modelling pig manure production

The starting point was the regression formula for the manure production of pigs by Hopkins and Cruz (1982, p. 9):

Y = 23.55-4.2*ln(X), n = 60, r² = 0.8382 (2.14)

for pigs weighing between 20 kg and 70 kg and

Y = 8.452 - 0.0495*X n = 60, r² = 0.7450

for pigs weighing between 70 kg and 100 kg, where X denotes the liveweight of a pig in kg and Y the fresh manure output per day in percent liveweight. Hopkins and Cruz used two formulas because the different food fed to pigs of different weight classes also resulted, of course, in a different manure output. Although these formulas were obtained for pigs that were fed 7 % - 3.5 % liveweight per day as described above, this formula is also valid for other feeding regimes. Since the feeding regime finds no entry in the formulas and pig weight is always attained according to feeding rate, this approach also seemed to be useful for different feeding rates. Applying either formula to the whole range of pig weights possible in the model yields negative manure outputs above 262 kg and 170 kg, respectively. So for modeling the manure output of pigs another rather simple approach was tried:

It was assumed in the pig model that the liveweight equivalent of the food intake q(t) was AB*q(t). That means that only a certain ration of the food intake can actually be used for growth and further catabolic processes. Now it was assumed that the rest is lost through faeces. This portion is therefore (1-AB)*q(t). The advantage of this approach is that it can be applied to all kinds of feeding regimes and pig growth performances. This simple formula was validated by comparing the results of the theoretical approach to the regression equation of Hopkins and Cruz.

The regression equations applied in their extent of validity resulted in a daily manure output of 3.5 - 11.5 % total live weight. Converting this into dry matter yields 1.05 - 3.45 % daily manure output of total live weight (Hopkins and Cruz found an average total solids content of pig manure of 29.9 % (Hopkins and Cruz, 1982, p. 12)). Moisture content of the food applied was 13 %. Thus, the theoretical manure output in terms of dry matter is (1-AB)(1-0.13)*q(t). Running the pig model with the same feeding regime as the pigs of Hopkins and Cruz (linear decreasing feeding rate from 7 - 3.5 % live weight (see above)) and the parameters derived earlier results in a daily manure output of 2 - 4 % liveweight. This deviation of 1 % of the two models in daily manure production resulted in a different total amount of produced manure in one season of more than 50 %. Therefore the simple theoretical approach was rejected in favor of Hopkins' regressions formula. For simplicity the first formula was applied to the whole range of possible pig weight with one restriction: since maximal mature food intake was set to 8.7 kg/d and using the arguments from above then maximal (1 - AB)*8.7 kg FW/d can be lost through faeces. This corresponds to 1.5 % body weight. Therefore daily manure production can only decrease up to 1.5 % live weight. This value is attained at a weight of 190.6 kg using formula 2.14.

Pigmanure can serve as input for the fishpond (fert, kg FW/d).

Furthermore, buffalo manure can also be applied to the fishpond (NtoFish, kg N/d). All inputs are converted to nitrogen and flow into a stock called Ninput (kg/ha). In order to get the average nitrogen input (NperHectarperDay, kgN/(ha*d)) which was used to derive the statistical relationship the value of Ninput is divided by the simulation time. This is a buffer for daily fluctuations in the nitrogen input.

Ninput(t) = Ninput(t - dt) + (dailyNAppl) * dt

dailyNAppl= ((RiceBranTo_Fish*0.011)/Pondsize)+

((fert*0.299*0.019)/Pondsize)+

NtoFish/Pondsize

NperHectarperDay = IF (TIME=0) THEN 0

ELSE

IF(Ninput/TIME<4.5) THEN Ninput/TIME ELSE 4.5

Nitrogen content of all rice products was set to 1 % DM. This is an average value from different sources which stated different values (Dalsgaard, 1997; FNRI, 1990). Nitrogen content of pig manure was taken from Hopkins and Cruz (1982). For a listing of all considered nitrogen contents see Tab. 12.

It has already been described how bundweeds serve as input or rather fodder for buffalos (see chapter 2.5). Buffalo manure can either be used as input for the fishpond (NtoFish, kg N/d) or as fertilizer for the rice field (NtoSoil, kg N/d).

Drawing a diagram of the considered main compartments and their connecting flows as they are modelled in FARMSIM we receive Fig. 28:


Fig. 28: The diagram of all modelled compartments and resource flows of the final concept.

Verification and long term stability

All submodels were already verified, also concerning their potential inputs and outputs. Therefore the internal structure of the whole integrated model only had to be verified with respect to the 'cooperation' of the submodels. Many different scenarios were run by the model over a long time period and the outcomes were compared to the internal logic of the model. The model behaved as expected.

Long term analysis does not reveal new results. The model is stable in the long run. The only variables that do not have a maximal or minimal value, but grow ad infinitum the longer the model runs are total chicken production, total fish production (if these compartments are simulated at all) as well as the purchased chicken and pig food. But this is not very astonishing and does mirror the real situation as well. All other stock values are either periodical (chicken production, fish production, bundweed growth) or approach an equilibrium (pig production, rice yield, pig and buffalo manure production).



[Previous] [Contents] [Next]