It can be seen in Tabs. 1 and 2 and Fig. 26 that the final model is not quite of the complexity as outlined or desired in chapter 2.1. The identified compartments (fruit)trees, rice-fish and vegetables which also represent important enterprises on Philippine small-scale farms are not integrated into the whole farm model. Unfortunately, due to the lack of data, appropriate models and time, those features of a farm were not considered within the scope of this thesis. So the whole farm model has to remain incomplete, yet, and certainly this also applies for its applicability. Another weakness of the final model concept concerning sustainability and system analysis is the lack of feedback cycles (Fig. 26). In fact, the model is an input-output model with various intermediate stages of converting external inputs, like fertilizer, food and climate factors in valuable and marketable primary outputs. Therefore efficiency analysis is a feature which can be well performed. Showing how outputs and waste can be used more effectively through reuse (rather than recycling) and different management strategies is one of the main effects which can be studied with this model. Still, comparing the real situation of Integrated Resource Management, as, e.g. illustrated in Fig. 1, with the concept established here (Fig. 26), it becomes obvious that the main features of Philippine farming systems are indeed included in the FARMSIM concept.
Integration of the submodels
Taking a look at the methods of integrating the submodels it can be noticed that all inputs for a specific submodel are treated in the same way. They all have the same effect, no matter what source they come from. For instance, it make no difference in the growth performance of pigs or chickens if they are fed rice waste or commercial food. On the other hand, commercial food contributes more than twice as much nitrogen to the nitrogen balance than rice bran or broken rice. It was already stated that the protein content of food does have a great effect on the growth performance of livestock. So it would normally be expected that feeding commercial food comes along with a better growth performance of livestock than inferior rice waste. To what extend growth performance improves is not known. Still, it is likely that this would also effect nitrogen efficiency. In the present implementation, feeding commercial foodstuff is 'punished' in terms of nitrogen efficiency because the relation 'gain in liveweight - nitrogen consumption' is lower than when rice by-products are fed.
The same applies to modeling pond fertilization. On Philippine farms is was observed, for instance, that fish feed directly on the rice bran which is thrown into the pond. This will surely have a different effect on fish growth than food consumption through the food chain, as assumed when manure is applied.
All these special effects concerning feeding and growth of animals are not quantified. Considering all other simplifying assumptions that have been made to assess the weight of animals, e.g. constant growth parameters, it must be presumed that the effects of varying protein contents of the food are rather minor.
Scenarios and Sensitivity
The three scenarios modeled give an impression of how certain sustainability indices can be analyzed with respect to different farm management strategies. Other scenarios can be modeled as well studying ecosystem attributes, like nutrient throughput, primary production over biomass (P/B ratio), biomass/throughput (B/E ratio), productivity, etc. (Christensen et al., 1993; Dalsgaard, 1996). Efficiency here is calculated by dividing primary output by external inputs. External inputs are defined as all inputs and sources which drive the model in terms of nitrogen. This also includes initial values and soil nitrogen supply. In an economic efficiency analysis soil supply would surely not have been considered, but it can be seen in Tab. 13 that in this case soil nitrogen supply is the largest quantity contributing to external inputs. Not counting this contribution in the scenario would results in efficiency values above 100 %, which does not make much sense from the system analysis point of view. Analyzing economic efficiency in future studies, one can think about the economic weight of soil nitrogen supply but in a mass balance calculation it certainly has to be included. When talking about sustainability this value is even of major importance. What is meant by saying that 'the natural resource base cannot be sustained'? It means decreasing soil nutrient supply due to exploitative farming methods.
The sensitivity analysis, however, points out some characteristics of the model which are caused by the inherent assumptions. At first it becomes obvious in the first scenario (Tab. 13) that efficiency is highest with no use of fertilizer. Rice yields completely depend on the natural nutrient supply. This result is also expected by our intuitive perception of efficiency. Certainly we would not have the problems described in the introduction if rice yields were sufficient without commercial fertilizers. But here lies the actual problem: that yields have to be increased with the help of fertilizers in order to satisfy the need for rice. The result that the whole system becomes more and more inefficient the more fertilizers are applied is the starting point of each sustainability debate. So the question is how can yields be increased without decreasing efficiency or sustainability? A step in the right direction is implied by scenarios 2 and 3 and the fertilizer optimization analysis. Making more efficient use of the applied fertilizers (simulated with the help of mathematical models) is one step. We cannot prevent a certain ratio of the fertilizer from being lost, this is natural, but we can surely minimize this ratio. Further, valuable nutrients that we put into the system by fertilization are lost again by not using crop or animal by-products. Recycling buffalo manure increases yields by 50 kg (Tab. 15) in this case. Whether this is economic or not has to be further investigated. However, it shows a way out of the fatal positive feedback cycle of increasing yields by fertilization, which decreases soil nutrient supply in the long run, this in turn decreases yield, which results in higher fertilizer application rates. Decreasing soil fertility would be an interesting feature to incorporate into the farm model as soon as reliable data is available. The adverse effects of intensive use of pesticides also play an important role in assessing ecosystem health. It is, however, difficult to determine quantitative numbers concerning those influences. This task must be left to future studies.
Increasing the number of chickens and their feeding rate also increases efficiency in small steps. It is astonishing that efficiency increases with more chickens and more purchased food beyond any reasonable number of raised chicken. Normally it would be expected that efficiency decreases with more and more imported food, since efficiency of keeping livestock decreases to zero with the life period of the animal (Fig. 31).
Fig. 31: Livestock raising efficiency, i.e. the ratio outputs/inputs decreases as the animal (here pigs) approaches maturity. It drops from about a hundred per cent to zero with diminishing rate.
TThe reason for this lies in the introduction of chicken life cycles. As pointed out above, it was assumed that chickens are slaughtered when they reach a weight of approximately half a kilo. Analyzing the nitrogen efficiency of chickens after each life cycle, it is found that the minimum average efficiency (without feeding rice by-products) lies at around 21.5 %. This efficiency of the submodel is significantly higher than the efficiency of the rest of the farm in scenario 2, which lies around 13.5 %. Therefore the chicken section efficiency is biased by this lower efficiency. The more chicken are raised, the less the submodel efficiency is influenced by the efficiency of the rest of the farm. As the total numbers of the chicken section become more dominant, the whole farm efficiency approaches the efficiency of the chicken compartment. The efficiency of 21.5 % is approached above a number of 100 000 chicken, which certainly makes no sense. Still, the effect is astonishing at first sight and it becomes 'mathematically clear' that there is a time beyond which keeping and feeding livestock is not efficient and possibly not economic anymore.
Such an effect can be seen when looking at efficiency-sensitivity of the second scenario. Here an optimal number of pigs can be observed. The larger the livestock grows or the older it gets (this applies to pigs and chickens alike), the less efficient it is, too. Output/input - efficiency drops from one hundred per cent to zero per cent with diminishing rate. This is obvious because when an animal reaches mature weight it does not grow any more, but still consumes food which decreases efficiency. So the optimum between efficiency and liveweight is of high interest for profit-oriented production (see Fig. 31).
In the second scenario the efficiency of pigs first increases because of the rice bran fed to those animals. The rice bran works like a buffer, which holds the negative effect of increasing the number of pigs for a while. The optimal value here is also of great interest to the farmer and depends on the amount of secondary farm by-products available. Efficiency of the pig submodel and the whole farm as well could be increased if pigs are not raised until the ratio output/input drops below the value of the rest of the farm.
In scenario 3 the integration of a fishpond always has positive effect on the overall efficiency because in all cases the efficiency of this submodel is higher than that of the rest of the farm. In fact is always lies above 100 %. This is because fish production always exceeds the initial stocking biomass by far. Of course, this results from the models' assumption that minimal final fish length is 11.1 cm, which is always approached after a certain time. Furthermore, no external inputs are used. This assumption is not all too far-fetched. In earthen ponds there is always a certain amount of phytoplankton and detritus produced by natural primary productivity. In the Hopkins - Cruz experiment no. 7 (see Tab. 5) fish grew 3 cm in 30 days with no fertilizer input. Unfortunately, this experiment was not finished because of a typhoon. Nevertheless, it is possible to produce fish without any input at all. Not at least this is the reason why aquaculture is propagated on Philippine farms. Thus, the question of how big a fishpond should be in order to increase primary production and farm efficiency must be answered with 'as big as possible'. It is clear that everybody would shake his head to this conclusion . Other constraints, like labor availability or simply farm size would soon limit the increase in efficiency.
Such results show the limits of the applicability of this farm model. It becomes clear that when sustainability indices and management strategies are to be developed and tested, many more features and levels have to be integrated. The biophysical level is the starting point and the basis, but many other aspects of sustainability belong to different levels. One important 'limiting factors' to the adoption of new technologies is labor. Small-scale farms in the tropics are managed on the family level and labor supply is always short. Integrating new technologies always means additional labor. When workers have to be hired the profits from the new farm enterprises could vanish. Dalsgaard (1997) reported that one farmer renounced recycling farm by-products and making compost because he considered it too time and labor intensive. Rice and vegetable cultivation is labor intensive enough. Yet, these aspects give support to the notion that such farm models are badly needed. Maybe the introduction of sustainable farming methods can be forced when it could, for instance, be forecasted by a farm model that composting and investing time and labor into recycling would be rewarded in the end. Indeed it is possible that compost reduces the need of commercial fertilizer to an extent where it becomes profitable to invest additional labor.
The model, of course, has to be further validated, but once the uncertainty of the input-output relations of the single compartments are reduced, no more time intensive measurements of the specific flows have to be carried out in order to calculate sustainability indicators for a given farm. The most important variables, like farm size, fertilizer input, etc. (see Tabs. 13-15) which drive the model are readily available and farms that are studied can be quickly simulated and analyzed.