Water research commission

Carolina

A case-study farm was selected in Middelburg, Mpumalanga to model the impact of climate change on a typical summer rainfall dry land farming system. Due to the TOR of the project and budget constraints, no survey was done in the surrounding area of the farm. The participating case-study farm has a high level of record keeping and provided most of the information needed to do the modelling.

Agriculture in the Middelburg region is generally dominated by extensive grain production and the grazing of beef cattle and sheep. Mainline grain production includes maize, sugar beans, soybeans and sunflowers.

5.4.1 The existing sources of livelihoods

Main crops produced in the area include maize, sugar beans, soybeans and potatoes. Livestock production consists mainly out of cattle (weaner production), sheep (mutton and wool production) and dairy production.

The case study farm has typical Highveld mixed farming activities consisting of grain and livestock production. Activities include weaner calf, lamb and wool production.

5.4.2 Current and projected future crop yields and carrying capacities

5.4.2.1 Current yields

Error! Reference source not found. displays the current average crop yield for different crops.

Table 45: Average crop yields – Carolina case study

Source: Own calculations, with inputs from case study farmer

5.4.2.3 Future yields

ACRU model

As a model of intermediate complexity in respect also of its crop yield modules, the ACRU maize yield model is phenology and daily soil water budget based.

At the outset, two points should be noted from fig 32, viz.

• 
  For all maize yield and associated analyses in this report the baseline used for comparisons was not derived from the 50 year (1950 - 1999) historical record, as in other analyses, but rather the mean of values derived for the present period (i.e. 1971 - 1990) from the multiple GCMs used in this study.
  
• 
  In all analyses on maize the option available in the ACRU model to account for the so-called CO₂ “fertilization effect”, effectively through transpiration suppression by this C4 plant, was not invoked because of uncertainties remaining on the effect on the long term and under large field conditions.
  

The summer rainfall Blyde catchment is clearly climatically suitable for maize production (at a 1 November plant date), with the lowest means of seasonal yields in excess of 3 t/ha/season, and parts of the southwest showing that under sound management mean yields of the order of ~ 7.5 t/ha/season could be attained.

Maize yield projections into the IF in the Blyde catchment from the climates of the multiple GCMs used with the ACRU model, on the other hand, show increases of 10 - 100 %, with the highest projected increases in the west where yields could possibly change from ~ 3 to ~ 5 - 6 t/ha/season (Error! Reference source not found.2 bottom left). These projected increases are dampened into the MDF, with the presently high yielding southeast showing maize yield reductions of up to 10 %.

Figure 32: Mean seasonal dryland maize yields in the Blyde catchment

(Mean seasonal dryland maize yields in the Blyde catchment, estimated by the ACRU model under present climatic conditions (top) and projected changes into the intermediate future (bottom left) and the more distant future (bottom right))
For further descriptions on the modelling results, see Appendix.
APSIM modelSimulating future maize yields in Middleburg with APSIM

The data available to set up the model were sufficient to run the model and to present the following results. However these data are not representative of the fine scale APSIM can deal with, and translate generic soil conditions and generic crop managements. Hence we advise any user of this data not to extrapolate information from a resolution higher than the data resolution inputs.

In order to present a descriptive interpretation of various future climate projections, we present the following results for the A2 and B1 CO₂ emission scenarios (SRES), and for 15 GCMs (9 with A2 and 6 with B1).

Figure 33 to 38 show the simulated rainfed maize yields on the y-axis, against its ranked occurrence (percentile) on the x-axis. Hence the reader can appreciate the response to multiple GCMs and multiple years, the worst possible output under percentile 0, the best possible output under x-axis percentile 1, and the evolution from the former to the latter, particularly taking not of the median case for percentile 0.5. We expect these statistical plots to provide a general sense as well as a sense of variability of the biophysical response under future climate.

Error! Reference source not found.33 displays the control yield vs. percentiles, simulated with 9 GCMs for 1961-2000 and with historical data for 1979-1999.

Figure 33: Simulated yield under observed (1979-1999) and control (1961-2000) climatesControl yield vs. percentiles

(Control yield vs. percentiles, simulated with 9 GCMs for 1961-2000 and with historical data for 1979-1999)
The control simulations are consistent with the simulated yield computed with observed weather data, except for worst case scenario (< or = to 10^th percentile). This is likely due to the mathematical nature of the crop model, where a crop can dies die simply by simply reaching a mathematical threshold, which in the field hardly result in a complete/total loss.could not be the case in the field. The control simulations seem to increase the variability of the simulated yields, as we can see low er worst casesyields being mostly underestimated (< 10^th percentile) and high yields mostly overestimateder best cases (>90^th percentile) compared to the simulations ran with observed data.
Further explanation could be reached by further validation, i.e. (much) more data.

Figure 34: Simulated yield under future climate (2046-2065) for 9 GCMs driven by SRES A2 emission scenarios.

Future (2046-2065) yields vs. percentiles

(Future (2046-2065) yields vs. percentiles simulated for 9 GCMs driven by A2 emission scenarios)

Figure 35: Changes in yield (future minus control) under SRES A2 emission scenario

Changes (future minus control) driven by A2 emission scenario

The CCCMA GCM projections (SRES A2) stand apart from the 8 other GCMs. It shows a decline of yields simulated straight from the lower yields, while the 8 other GCMs project a noticeable increase for low yields , and this anomalies anomaly diminishes toward the higher yields, at which point there is no evidence of a change.

Figure 36: Simulated yield under future climate (2046-2065) for 9 GCMs driven by SRES B1 emission scenarios.

Future (2046-2065) yields vs. percentiles

(Future (2046-2065) yields vs. percentiles simulated for 6 GCMs driven by B1 emission scenarios)

Figure 37: Changes in yield (future minus control) under SRES A2 emission scenario

Changes (future minus control) driven by A2 emission scenario

As well as for the SRES A2 emission scenario, and though not in the same extend, the CCCMA GCM projections (SRES B1) stand apart from the 5 other GCMs (SRES B1). It shows a decline of yields simulated straight from the lower yields, while the 5 other GCMs project a noticeable increase for low yields, and this anomalies anomaly diminishes toward the higher yields, at which point there is no evidence of a change.

Summary

As an attempt to sumarise the former results we show here the production and changes comparing observed, control, and futures driven by SRES A2 and B1.

Figure 38: On the left, simulated low to high rainfed maize yields in Middleburg under observed (1979-1999), control (1961-2000) and future (2046-2065) for 9 GCMs driven by A2 scenario and 6 GCMs driven by B1 scenario. On the right, minimum, median and maximum changes simulated (future minus control).

Production and changes comparing observed, control, and futures driven by SRES A2 and B1

As an attempt to summarize the former results, we show in Figure 38 the average simulated yield for observed, control and futures under SRES A2 and B1, as well as a detailed minimum-median-maximum changes from control to futures (A2 and B1).

The production variability from low to high yields shown on the left confirms that the climate baseline is less variable that the GCMs controls. This is especially true for high yields independently of the SRES scenarios. (On the left, simulated yields vs. percentiles for crop simulation of the rainfed maize in the middleburg area under observed (1979-1999), control (1961-2000) and future (2046-2065) for 9 GCMs driven by A2 scenario and 6 GCMs driven by B1 scenario. On the right, minimum, median and maxim GCMs changes (future minus control)).

We draw the following observation by considering the percentiles > 10^th, as for explained under the control simulations.

The results from the right on Figure 38, above suggest a low yield increase – high yield no change linear trend. This seems to translate both SRES A2 and B1 in the same proportion, and even the various GCMs. the following change in rainfed maize production at this location. ItResults shows a median increase of 38% for low yields (20^th percentile), a median 10% increase for median yields (50^th percentile) and a median insignificant 1.9% decrease for higher yields (80^th percentiles). These observations are consistent for all but one GCMs (CCCMA), and for both A2 and B1 scenarios.

5.4.3 Projected shifts in optimum cropping areas

Some preliminary model projections have been made for maize, wheat and soya by Schulze (2011) and Estes et al. and these show changes in crop suitable areas, where the largest areas of lost suitability is found in the Free State, North-West and Limpopo Provinces, while modest gains in suitable areas are to be found in the Eastern Cape and Mpumalanga. (see fig 39)


In the Blyde and Olifants irrigated catchment areas, the suitability is merely a function of temperature and as such can be measured by changes to yield and not spatial suitability. The modelling results are covered in the above report.

Figure 39: Shifts in optimum growing area for Soya (Schulze, 2011) – case study area in block.

Figure 40: Gain/Loss of suitability for growing maize (Estes et al., 2011) - case study area in block

5.4.4 Current and future farming management practices (e.g. fertiliser/manure application, irrigation, tillage practices)

5.4.4.1 Soil characteristics

Error! Reference source not found.46 illustrates the soil characteristics in the Carolina area.

Table 46: Soil characteristics – Carolina

Source: School of agricultural, earth and environmental sciences, UKZN (2012)

The soils characteristics are area weighted from the land type information in the ISCW soils database (ISCW, 2005) for the Quinary Catchment in which the location of interest is sited. The 4-digit number (location) is the Quinary number in the SA Quinary Catchments Database (Schulze et al., 2010). The methods by which these characteristics for a 2-horizon soil have been derived are described in Schulze and Horan (2008) using the AUTOSOILS decision support system developed by Schulze and Pike (1995 and updates). Values of wilting points, field capacities and porosities (i.e. at saturation) imply the soil water content (in meter of water per meter thickness of soil) at those thresholds. Saturated drainage implies the fraction of soil water above field capacity that drains into the next horizon (i.e. from the topsoil to the subsoil or from the subsoil out of the active rooting zone) per day. From the characteristics in the table the soils tend to have a sandy loam texture at Carolina.

5.4.4.2 Adapted crops for the region

Main crops produced in the area include maize, sugar beans and soybeans. Livestock production consists mainly of cattle (weaner production), sheep (mutton and wool production) and dairy production.

Error! Reference source not found. reflects the physiological lifecycle of maize, sugar beans and soybeans.

Table 47: Physiological lifecycle of maize, sugar beans and soybeans

Source: Carolina workshop and expert group discussions (2012)

5.4.4.3 Current cultivation practices

Error! Reference source not found.48 summarises the current cultivation practices for maize, soybeans and sugar beans in the Carolina area.

Table 48: Current cultivation practices



Source: Carolina workshop and expert group discussions (2012)

The case study farm has typical Highveld mixed farming activities consisting of grain and livestock production. Activities include weaner calf, lamb and wool production.

Error! Reference source not found.49 reflects the carrying capacity for the farm.

Table 49: Carrying capacity – Carolina case study

Source: Case study farmer (2012)

5.4.4.4 Crop rotation practices

Crop rotation includes maize and soybeans/sugar beans.

5.4.5 Appropriate household and whole farming systems modelling

5.4.5.1 Case study farm

Table 50 below reflects the composition of the summer rainfall dryland case study farm.

Table 50: Description of case study farm: Carolina



Source: Case study farmer’s records (2012)

5.4.5.2 Crop Enterprise budgets

Error! Reference source not found.51 to Error! Reference source not found. summarise the crop enterprise budgets for the Carolina case study.

Table 51: Crop enterprise budget summary: maize, sugar beans and soybeans

Source: Own calculations, with inputs from case study farmer

Table 52: Crop enterprise budget summary: beef and mutton production



Source: Own calculations, with inputs from case study farmer

5.4.6 Organisation of farmers in formal and informal groups. Existing support service.

5.4.6.1 Organisation of farmers in formal and informal groups

The reader is referred to Error! Reference source not found. for details regarding the discussions with farmers and cooperative members. Farmers belong to study groups.

5.4.6.2 Existing support services

Farmers study groups are organised by AFGRI, the local silo/cooperative which offers assistance in respect of seed, pesticide, herbicide and fertiliser requirements as well as informal study groups.

6.