Linking simulation modeling and farmer participatory research to develop fertility management technologies for smallholder farmers in Malawi and Zimbabwe.

Background

CARMASAT ( Collaboration on Agricultural Resource Modeling and Applications in Semi- Arid Tropics) had established collaborative links with 3 existing research teams in southern Africa whose focus was fertility management in smallholder farming systems utilizing farmer participatory research methods. CARMASAT's role in these collaborations is an assessment of best-bet technologies (using APSIM) for testing on-farm, simulation of on-farm researcher and farmer managed experiments, and assistance in communication of research results to the wider extension and agribusiness sectors. Expected output from the collaborations is an assessment of how a simulation capability enhances the effectiveness of on-farm fertility management research done in a farmer participatory mode.

Based on 2 seasons of mainly on-station experiments, scientists at a workshop in September 1999 concluded that APSIM was reliable for simulating maize response to N fertilizer inputs but there were some uncertainties about parameterization of local sorghum cultivars and the new APSIM-Manure module. It was suggested that any collaborative effort on linking simulation and farmer participatory research (FPR) should initially concentrate on fertilizer use in maize-based systems.

Three workshops have been conducted in Zimbabwe and Malawi since Sept 1999. At each of these, APSIM was used to conduct scenario analyses to explore how this tool can contribute useful analysis and insights for fertility management research in southern Africa. It also provided participants, especially national extension agency scientist, their first exposure to and participation in system analysis using simulation.

Workshop Scenario Analyses – an example.

At the September 1999 workshop, a scenario analysis was conducted to explore how APSIM could be used to answer the type of resource allocation question faced by resource poor farmers. In this instance, trade-offs in allocating limited capital resources between fertilizer purchases and hiring labour for extra weeding were examined. The analysis was possible using APSIM because of its flexible management capabilities and its inter-cropping module to simulate weed competition.

One scenario involved a farm household of 2 adults and 4 children in the Bulawayo region of Zimbabwe. Farm size is 5 ha with 4 ha of cropping. The soil is shallow infertile sand, and there is a moderate weed pressure on the croplands. The farmer plants the 4 ha in stages of 1 ha, firstly as a strategy to avoid risk of crop failure and also because of labour constraints.
Household labour is sufficient to control weeds on the first 2ha sown (perhaps because of education demands for children). A remittance from a cousin in the city is available and is sufficient to purchase 2 bags of ammonium nitrate or to hire labour to weed the 2 nd 2 ha or hire a draft animal to prepare an additional 1 ha of land for earlier sowing. Given the unreliable rainfall for this region and its influence on the outcome of such decision-making, the question relevant to a resource poor farmer might be, ‘on average, which allocation of the resource offers the best prospects of an acceptable return?'

Investment scenarios simulated included; none (baseline), purchase of 2 bags of fertilizer, splitting the investment between fertilizer (1 bag) and labour hire to weed field 3, and investment in labour only to weed fields 3 and 4. Where the fertilizer option was pursued, strategies for targeting the fertilizer were also simulated. Eleven seasons were simulated, with simulated soil moisture, nitrogen and crop residue status at harvest carried over from one year to the next. The 4 x1 ha fields were sown within 4 different planting windows (Nov 15-30, Dec 1-15, Dec 15-31, Jan 1-15) using a sowing rainfall criteria. Crop surface residues were removed from the system after harvest to represent the stubble grazing by animals.

The baseline maize yields for the 4 fields were low and variable (data not shown) and for the conditions specified, participants agreed that they were well in line with expectation. Figure 1 shows how maize yield would respond if labour were hired for weeding fields 3 and 4.

Figure 1. Simulated maize yield (kg grain /ha) for weeded and un-weeded fields sown in late Dec and early Jan on infertile sand at Bulawayo, Zimbabwe.

Figure 2. Simulated whole farm (4 ha) maize production (kg grain/ha) for different inputs and distribution of N fertilizer (35 kg N/ha on earliest sown maize field or 17.5 kgN/ha on first 2 sown maize fields) on infertile sand at Bulawayo.

Figure 2 shows the response for whole farm maize production when 2 bags of AN is purchased and either concentrated on the earliest sown maize at 35 kgN/ha or split between the first 2 sown fields, both of which are able to be weeded using available household labour resources. The simulated data show that in most seasons there is a large yield benefit for the fertilizer investment (Z$1600) and at Z$4000/tonne of maize, the average yield increase of 1500kg would give a good return. However, in 4 out of the 11 seasons there would be little if any return, and 3 of these years are in succession.

Simulated data in Figure 2 indicates a marginal but relatively constant advantage to splitting the small fertilizer input over a larger area. This result generated discussion about the concept of agronomic efficiency, which tends to be higher at lower rates of N fertilizer. (Agronomic efficiency is defined as extra kg of grain per unit of nutrient applied. It is derived from the response curve and is typically maximal at the lower N rates where the response curve has the steepest slope.). When coupled to the increased production area resultant from spreading a fixed amount of nutrient at a lower rate, agronomic efficiency can attain some significance in terms of total grain production, especially in simulated yields where other field factors that may mask this increment are absent.

The ON treatment data in Figure 2 represents the baseline household maize production with no investment in fertility. It shows that in only 2 seasons is the household food requirement estimate of 1200 kg/year exceeded, and that simulated output is in-line with the type of subsistence living faced by households in these farming situations.

Overall, the analysis suggested that the highest expected return (2616kg/ha, Table 1) on the small investment available was to purchase fertilizer and distribute it on the 2 early sown, weeded fields. However, splitting the investment between fertilizer and labour hire to weed field 3, provided an almost equally high expected return, but with a much lower risk expectation (as measured by the standard deviation and minimum yield data in Table 1). This might be a very important and useful insight for risk-averse farmers typical of semi-arid agriculture.

Table 1. Simulated whole farm production statistics (kg grain/ha, 11 years of data) for 3 investment scenarios

Investment	Mean	Stdev	Minimum
None	1192	505	413
Buy and distribute N (17kgN/ha on F1&F2)	2616	883	923
Buy labour and N (Weed F3 & Dist. N -9kgN/ha on F1&F2)	2408	500	1585

Conclusions

The scenario analyses have proven to be very effective in showing how simulation can contribute to researcher learning about fertility management technologies in small-holder farming systems. For example, following the workshop, one collaborating project included extra weeding as an experimental treatment in its best-bet N technologies for testing on-farm with farmers. Similarly, another project included low rates of N as part of its on-farm experimentation for sorghum. The key ingredient missing to date has been input from farmers. It is hoped to share on-farm experimental data and to use simulation with farmers in the latter half of 2000, to obtain their input on formulating scenarios and feedback on simulated, as well as experimental, results.

Collaborators

Drs Snapp and Rohrbach (ICRISAT), Drs Vaughan and Shamudzarira (CIMMYT), Dr Muza (DRSS), Drs Carberry, Keating and Robertson (APSRU).

For more information please contact:

Dr JP Dimes
Scientist (Farming - Systems Modeling),
ICRISAT-Patancheru 502 324.
Fax # +91 40 3296182.