[Previous] [Contents] [Next]

2.5 Buffalos and Bundweeds

Introduction

Weeds growing on the small embankments that surround and separate single rice fields so that farmers have a better access constitute another important farm resource. They serve as fodder for buffalos. Especially in the wet season when all rice fields are planted and the buffalos are kept in the homestead so that they cannot graze freely in the rice fields, the weeds that grow on the bunds of rice fields are harvested and fed to the buffalo(s). Weedings are also performed to reduce nutrient uptake and photosynthetic competition caused by shading (Dalsgaard, 1995b).

Water buffalos Bubalis bubalis are important enterprises for farmers and they are never missing on tropical farms. They serve as cheap labor for the farmers and are mostly used for the land preparation before rice transplantation. Usually they are kept as long as they are fit for work. Slaughtering on the farm and milking is uncommon. Sometimes the sale of calves can become an important income source for the farmers. As buffalos are kept long after they reach their mature live weight, no growth is assumed in the model. Reproduction is neglected as well, because buffalos need more than one season to reproduce.

Data and Methods

The weed growth model used here is also a statistical model based on the data collected by Dalsgaard and Oficial (personal notes)3. The dominant bundweed sampled was Imperata Cylindrica (cogon) during WS 94 and DS 95 on four different farms. Farmers often follow a schedule of bund weeding thrice per rice crop: before transplanting, half way through the growth period and shortly before harvest. This schedule provided a convenient frame to


Fig. 24: Bundweed harvest data show high variations even on the same farm or ricefield. The weeding pattern follows a fixed scheme of thrice per season.

monitor weed production on the farms (Fig. 24). On each sampling day and each rice field three samples were collected. Yields showed a very high deviation even on the same spot. Sometimes weed production on the bunds of one rice field varied between 1 and 5.5 kg/mô. For aggregated data see Tab.11. Since bundweeds are not explicitly fertilized, climate factors were considered to be the main contributors to weed growth and development. For the regression analysis farm 1 was not considered in order to have an independent data set for validation.

Climate data was obtained from CLICOM database (SARP, 1996) which is close to the place where the considered farms are located. The average climate data of the years 1990-1993 was taken into account.

Data on buffalo manure production was obtained from a buffalo-fish experiment in Thailand (AIT, 1986, p. 19). In this buffalo-fish experiment is was observed that buffalos produce an average of 2 kg dry matter of manure per day and animal.

Data on daily weed consumption was obtained by RESTORE and Dalsgaard (1997). Dalsgaard observed on one farm that 5.3 tons of weeds were cut and carried to two animals when housed for 3.5 months. This makes an average of 25 kg per animal and day.

Analysis of the RESTORE data revealed high variations of the daily feeding rates of buffalos. They ranged from about 100 kg per day per animal to only 1 kg per day per animal. This large variation is due to the different conditions on the farms according to season. Buffalos are only fed regularly in the WS when they are mostly housed, so they do not graze in the rice fields. For the WS average daily feeding amounted to 28 kg per day and animal.

Data on the relation of manure production to weed consumption could not be found.

Tab. 11: Aggregated bundweed data, taken from working notes and raw data from P. Dalsgaard and B. Oficial, collected on four different smallholder rice farms during WS 94 and DS 95.
Season/ Farm Harvest day growth period [d] av. yield [kg/m²] Rainsum [mm] Temp. Sum [°C]
Wet Season
Farm 1
224
81
1.61
830.67
2311.85
272
48
1.50
483.12
1321.70
312
60
2.40
428.12
1637.50
Farm 2
216
75
1.46
601.42
2130.30
272
56
1.28
548.37
1543.60
312
60
1.35
428.12
1637.50
211
96
1.71
605.47
2742.35
Farm 3
280
69
1.24
639.15
1903.67
308
28
2.40
215.22
756.75
355
47
1.04
430.17
1254.25
Farm 4
314
90
1.40
772.10
2459.34
12
63
0.74
482.77
1623.50
Dry Season
Farm 1
25
77
1.36
489.82
1977.00
85
60
1.38
59.50
1794.45
115
30
1.37
17.72
831.20
Farm 2
37
90
1.38
506.27
2305.30
75
38
1.02
32.45
975.90
115
40
0.70
27.35
1096.35

Since all samples were tripled, 'av. yields (kg/mô)' are average yields also omitting extreme values. Outliers are negelcted in four cases: farm 1, 28.9.94 and 7.11.94; farm 2, 24.4.95 and farm 3, 3.11.94. Data of farm 1 was not used for the regression analysis.

Results

Due to the tropical climate conditions rainfall and average daily temperature follow a well defined pattern throughout the year. It was observed that average growth performance in the wet season was slightly higher than in the dry season (Fig. 24). But this observation could not be related significantly to rainfall or daily average incident global radiation.

Since growth is directly related to photosynthetic activity, temperature is an important factor controlling growth. Temperature dependency of photosynthesis results from the fact that chemical reactions anticipating photosynthesis have a Q10-value >= 2. Therefore the attempt was made to correlate growth performance to daily average temperature or rather the temperature sum accumulated during the growth period. This approach is also widely used in many different models to simulate either growth (DAISY, Hansen et al. 1990) or developmental stage (ORYZA_0, ORYZA_1, ORYZA_N; Oryza Modules, SARP 1994; SUCROS, Penning de Vries et al. 1982; EPIC). As the function describing weed growth, the universal growth function of Bertalanffy is used here again. This function was fitted to the average yields on each sampling day omitting the data of farm 1 and one outlier of 2.4 kg/mô on farm 3.


Fig. 25: Despite the high variations of yields the relation of bundweed yields and accumulatd temperature was more promising than the relation to rain fall or incident radiation.

Using the Levenberg-Marquardt nonlinear regression algorithm as implemented in SPSS 7.0, the described approach resulted in the following formula:

Yield = 1.79 * ( 1- exp(-0.0008*Temperature Sum)) , n = 11, rô; = 0.254


Fig. 26: Average daily temperature follows a well defined pattern throughout the year.

In order to simplify the model, average daily temperature was fitted to an equation of the form

Temp = a* sin ((2*/365)*(b*(1-exp(-c*DOY))-d))+e,

using the same regression method as above. The following values for the parameters were found:

a = 1.9, b = 420.6, c = 0.0054, d = 146.77, e = 26.85; n = 4*365, rô; = 0.815 (Fig. 26)

Analysis of the data of Dalsgaard led to the assumption that the bundweed area covers about 10 % of the rice field area. Moreover, fixed weed harvest days were imposed according to the observed weeding schemes (thrice per season) and a hypothetical course of farming events which was figured to be rather typical. Harvest days were set to the 25th, 85th, 150th, 216th, 274th and 330th day of year.

Weeds(t) = Weeds(t - dt) + (weedgrowth - weedout) * dt

weedgrowth=

IF (DOH=1) THEN Bundweedgrowth*Hectar*10000*0.1 ELSE 0

Bundweedgrowth = 1.79*(1-exp(-0.0008*TempSum))

weedout = 28*numberofBuff

Buffalos

Daily weed consumption of buffalos was set to 28 kg per animal and day and daily manure production to 2 kg DM per animal and day. Nitrogen content of buffalo manure was also taken from Edwards (1994a), i.e. 1.4 % DM. Normally, if buffalos are housed in the WS, farmers feed the animals on a daily basis. In the model it was assumed that buffalos feed on the weeds that are harvested at certain times, like the whole weed yield was stored and given to the animals bit by bit.

Verification and long-term stability

The internal logic and the expected behavior of this submodel are not very complex and the implementation is easily verified. Weed growth approaches a maximum value and the harvest dates are fixed. This results in a periodical weed development which repeats itself each year. Buffalo manure production is constant and independent from all external factors. The fraction of manure which is not used is lost to the environment and does not accumulate. This submodel is stable in the long run.

Sensitivity analysis

All parameters used in the model are subdued to a rather high uncertainty, except the temperature curve whose calculated shape may vary by about 10 %. The growth period varies between 56 and 66 days, according to the imposed weeding scheme. Looking at a medium growth period of 58 days from 216 DOY until 274 DOY and varying the parameters 20 % from the standard value results in a deviation from the simulated yields of maximal 20 %.

Validation of the bundweed model

In Fig. 27 the model is run with the harvest data of farm 1. The simulated growth curves are plotted together with the measured yields, respectively.


Fig. 27: Validation of the bundweed regression model. The data of farm 1 was not used for regression analysis.

Discussion

As it can be seen from Fig. 27 the approach of relating weed growth to the temperature sum results in a good reflection of the general perception of bundweed yields. The simulated yields mostly lie within the range of measured yields. The last harvest of the year is underestimated. Since weed growth performs such high variations, it is clear that this model only gives an indication of the order of magnitude of the real yields. Nevertheless, it is assumed that the average picture suits the purpose of describing the general availability of bundweeds of a typical small-scale farm.

The most unsatisfying feature of the bundweed-buffalo complex is the missing link between weed consumption and manure production of buffalos. In the model every buffalo produces 2 kg DM manure per day, no matter if there is weed available or not. So it might happen that all stored weed is consumed way before the next harvest and there is still a lot of manure produced. This is the case if there are too many buffalos or, the other way round, not sufficient bundweed area. Normally in this situation the farmer would probably sell the appropriate number of animals in order to have enough food for them all. Since the model does not take care of this, it is up to the user to decide whether the number of buffalos on his virtual farm is too large or not. As a rule of thumb it can be said that if the stock of weeds from the last harvest is used half way before the next pile of weeds are harvested, the user should consider diminishing his imposed number of buffalos on the farm.



[Previous] [Contents] [Next]