Temporal variation in weed occurrence and biomass under conservation agriculture and conventional farming practices

Conservation agriculture (CA) is advocated as a management system for sustainable productivity, while preserving the environment simultaneously. CA has many advantages, but weed management is regarded as one of its biggest challenges. This study reports on the temporal variation in weed occurrence and biomass under conservation and conventional farming practices. The treatment design was a split-split plot, with a randomised complete block design with three blocks as replicates. Tillage was the main plot factor (reduced tillage [RT] and conventional tillage [CT]), and treatments (a combination of cropping systems and fertilizer levels) were treated as the sub-plot factor. Only cultivation year (F(2.48) = 9.12, p < 0.001) and the cultivation year and tillage interaction (F(2.48) = 22.41, p < 0.001) significantly affected weed biomass. Weed biomass and species diversity increased under RT from cultivation year 3 to 5. Under CT weed biomass had a slight downward trend and species composition was similar across the three years with two dominant weeds representing between 87.2 and 75.1% of total weed biomass. The results suggest that tillage practices can affect both the biomass and diversity of weeds. It is therefore important that practitioners understand such variation and apply weed management practices accordingly.


INTRODUCTION
Conservation agriculture (CA) is advocated to improve soil health, optimise crop yields and reduce input costs when effectively applying three principles: (i) Minimal soil disturbance (including practices such as no-till or reduced till); (ii) Permanent soil cover (including crop residues or cover crops); and (iii) Crop diversification (inclusion of various crops, especially legumes, and introduction of rotation or intercropping system) (Dumanski et al., 2006;Hobbs et al., 2008;Wall, 2008).In recent years, CA has been promoted under small-scale farmers in sub-*Corresponding author.E-mail: swanepoelc@arc.agric.zaAuthor(s) agree that this article remain permanently open access under the terms of the Creative Commons Attribution License 4.0 International License Saharan Africa to improve food security (Mazvimavi and Twomlow, 2009), as conventional agricultural practices are difficult and costly to implement for many of these farmers.For example, it is estimated that up 35% of the world's population do not have access to high input technologies (such as intensive soil preparation and use of fertilizers and chemicals) associated with conventional farming practice (Pretty, 1999).Africa has been slower in adopting CA compared to other developing countries, and one of the major limiting factors is the number of necessary weeding events (Mashingaidze et al., 2012;Andersson and D'Souza, 2014).Most weeding is done manually, and labour is reported to be a constraint on many small-scale farms.Therefore, despite many successes achieved in the adoption of CA practices, weed management remains one of the biggest challenges (Rockstrom and Steiner, 2003;Hobbs et al., 2008;Giller et al., 2009;Andersson and D'Souza, 2014;Nichols et al., 2015).
Weed biomass, density, composition and temporal variation are closely associated with management practices, especially tillage (Teasdale et al., 1991;Garcia de Leon et al., 2014;Nichols et al., 2015).For example, conventional tillage practices may effectively control weeds by burial (Froud-Williams et al., 1984;Hobbs et al., 2008;Wall, 2008), or stimulate weed germination by raising soil temperature (Froud-Williams et al., 1984;Teasdale et al., 1991;Murphy et al., 2006).Alternatively, minimal or reduced tillage can shift weed composition from broadleaf to grass species (Swanton et al., 1999) or perennial weeds (Vogel, 1995), or increase weed species diversity when specific habitats for certain weeds are created (Murphy et al. 2006).Mulch or soil cover may reduce or inhibit weed germination through the release of allelopathic compounds (Christoffoleti et al., 2007) or smothering of weeds (Teasdale et al., 1991;Thierfelder and Wall, 2010).Furthermore, weeds can be influenced by location, time, nitrogen management (Swanton et al., 1999), timing of cultivation, rainfall (Teasdale et al., 1991), crop residue management, crop rotations, harvest procedures and other aspects of the production system (Wall, 2008).
Conventional tillage practices have many advantages, such as preparation of the seedbed, uniform placement of seeds, temporary relief from compaction, and effective removal of all weeds prior to planting (Hobbs et al., 2008), all of which level and clean the soil and simplify subsequent farming operations.On the other hand, reduced tillage lessens the use of fossil fuels, decreases runoff and erosion, thereby conserving soil organic matter (SOM) and preventing soil physical degradation (Hobbs et al., 2008;Wall, 2008).All three of the CA principles are intended to increase SOM, which, in turn, results in improved physical, chemical and biological properties of the soil that are associated with practising CA (Bot and Benites, 2005).High input cost, decrease in soil productivity and depletion of SOM could motivate a farmer to change from conventional to conservation agriculture.However, this change could also present new challenges regarding weed management (Rockstrom and Steiner, 2003;Murphy et al., 2006).
It is therefore important to understand not only the effect of agricultural practices on weed dynamics, but also the temporal variation in weed species expected under different agricultural practices.This study aimed to understand (a) the effects of conservation and conventional agricultural practices (tillage, crop rotation, soil cover) on weed abundance and biomass, and (b) temporal variation in weed community composition in a medium-term field trial at Zeekoegat, Gauteng, South Africa.The effect of tillage on weeds was the most evident, and therefore discussed in more detail compared to other CA aspects, such as permanent soil cover and multi-cropping.

Experimental site
To evaluate the effect of CA on soil and plant properties, the Agricultural Research Council (ARC) initiated the Zeekoegat onstation, dryland field trial.The trial started in October 2007 and ended, after 6 growing seasons, in May 2013.The trial was conducted north of Pretoria at Zeekoegat, Roodeplaat (25°36'55"S, 28°18'56"E), in Gauteng Province, South Africa.The soil is moderately fine to medium structured with a clayey texture (45% clay).The long-term, annual mean daily minimum and maximum temperatures are 10.8 and 27.1°C, respectively, and the mean long-term rainfall for the area is 704 mm year -1 (ARC- ISCW, 2006).An automatic rain gauge (Texas 525 TE) was installed adjacent to the field trial, providing site specific rainfall measurements for the duration of the trial (Table 1).

Trial layout
The trial was designed to compare conventional farming practices (ploughing and monoculture) with various CA aspects (reduced tillage, permanent soil cover and multi-cropping systems).The treatment design was a split-split plot, with experimental design, a randomised complete block, with three replicates as blocks.Each replicate was split into two tillage systems (main plots) with each main plot (reduced tillage [RT] and conventional tillage [CT]) further subdivided into six treatments (three cropping systems x two fertilizer levels (subplots)), giving a total of 36 plots.Repeated measurements were taken over years and regarded as sub-subplot factors (Little and Hills, 1978).
The three cropping systems were: Maize (Zea mays) monoculture (MM)), maize/soybean (Glycine max) rotation (MS) and maize/cowpea (Vigna unguiculata) intercropping (MC)).The maize in the MC treatment was planted in 1.8 m tramline rows to accommodate the intercropping of cowpea between maize rows.Cowpea and soybean were planted in 30 cm rows.Plot dimensions were 7.2 m × 8 m with 0.9 m planting rows for maize.
Fertilizer was applied at two levels: an optimal level to represent ideal nutrient supply, and a low level (50% of the optimal) to represent situations of reduced inputs from small-scale farms.The optimal level was calculated according to the fertilizer application guidelines and soil analysis, using a target yield of 4 ton ha -1 .Fertilizer was band applied during planting (60% of total) and top-up fertilizer (40% of total) was surface-applied 6 to 8 weeks later.
Limestone ammonium nitrate and superphosphate were applied, but no potassium fertilizer was needed, as the natural K content of the soil was very high (average of 475 mg kg -1 ).For legumes, only superphosphate was used (assuming a target yield of 1 ton ha -1 seed).

Land preparation and weed management
At the onset of each season (October), all crop residues from the previous season were flattened and slashed.The CT plots were ploughed with a mouldboard plough and then disked with a disk harrow.Furrows for planting were drawn with a four tine cultivator frame.The RT plots were undisturbed, except for furrows, created similar to those in the CT plots.Crops were manually planted with a hand-held planter.
Weed control was applied consistent across the trial and comprised a combination of chemical treatment (before and after planting) and manual weeding (hand hoed 2 to 3 times after planting).Every year, during November, just before planting, a mixture of Roundup® (glyphosate) and DualGold® (S-metholachlor) was applied equally across the trial by using a tractor and sprayer.This was repeated the day after planting, before crop emergence.Crops were planted after significant rainfall, usually by the end of November or early December.After crop emergence, weeds between the crops were manually removed instead of chemically, to prevent negative interaction of chemicals with microbial populations which were sampled during January each year.

Data collection
The Zeekoegat trial was designed to quantify effects of CA on soil and plant properties, and since weed evaluation was not originally included in the aims of the trial, weed data was initially opportunistically collected.However, in the third trial season (2009/10) the obvious difference between weed composition under reduced and conventional tillage systems served as a motivation to start thorough weed data collection.Weed samples were thus collected for three consecutive years, in January 2010 (third cultivation year (Yr 3)), January 2011 (Yr 4) and January 2012 (Yr 5).Weeds were sampled before the first manual weeding, which only occurred in January of each year (Table 1).The first weeding was delayed due to the nature of the sampling (chemical weed control interfered with biological samples for microbial research), and due to the planting time (logistic constraints during December holidays).This situation is representative of most small-scale farming practices, where either labour for manual weeding or funding for chemical treatments is limiting and weed control remains a challenge.
Two 1 m 2 quadrants were destructively sampled for weed biomass from each plot.Samples were taken in a fixed grid to eliminate bias sampling.Individual weeds were removed at ground level (above-ground biomass) after which they were oven-dried at 40ºC and weighed.The weights of two dominant weeds, large thorn apple (Datura ferox L.) and purple nutsedge (Cyperus rotundus L.), were measured separately, to calculate their relative abundance.Other weeds, which included flax-leaf fleabane (Conyza bonariensis (L.) Cronquist), khaki weed (Tagetes minuta L.), khaki bur weed (Alternanthera pungens Kunth), common blackjack (Bidens pilosa L.), narrow-leaved ribwort (Plantago lanceolata L.), and devil's thorn (Tribulus terrestris L.) were grouped in a separate class 'Other'.The focus was initially only on large thorn apple and purple nutsedge as they were the obviously dominant weeds.By the fifth cultivation year, other weeds were also becoming dominant, albeit in the RT plots (such as flax-leaf fleabane and khaki weed), but since these species were not included from the start, they were not specified at the end.
Maize yield was determined by hand harvesting two 5 m rows from the centre of the plots.Plants and cobs were counted.Cobs were removed and grain was stripped and weighed.Values were adjusted by taking into account the moisture content of the grain.

Statistical analyses
Weed biomass data was subjected to analysis of variance (ANOVA), using a split-split-plot model, with tillage as main-plots, treatments as sub-plots, and year as sub-sub-plots (sub-samples) (Little and Hills, 1978).ANOVA was used to test differences between effects.The data was acceptably normal distributed and separated using Fishers' protected t-test least significant difference (LSD) at the 5% level of significance (Snedecor and Cochran, 1967).This was done using Genstat Statistical package (Genstat, 2011).
Cluster analysis (vertical hierarchical tree plot) was used to statistically analyse data on weed biomass using STATISTICA 6.1 (StatSoft Inc.Tulsa, OK, USA).A dendrogram was constructed with Ward's clustering algorithm, and the Euclidean distance measured, that is, the geometric distance between variables in a multidimensional space.Since our study was not designed and set up to investigate the effect of weeds on yield performance, we could not formally test these effects.However, we used a linear regression model between weed biomass and grain yield as heuristic test to evaluate effect of weed biomass on grain yield.Finally we used a Shannon-Wiener diversity index and Evenness index to investigate effect of tillage on species richness and abundance (Magurran, 1988).

RESULTS AND DISCUSSION
During the three sampling years, rainfall was poorly distributed and resulted in low grain yields and low biomass production with subsequent low, ineffective soil cover (Figure 1).Soil moisture has a significant effect in delaying or restricting crop and weed emergence, which can affect the outcome of interference between crops and weeds (Roberts, 1984).The low soil water could delay weed seedlings, which could result in possible insignificant responses of weeds to cropping systems (Roberts, 1984).However, the comparative results from weeds under different tillage systems were more enhanced, as well as comparisons across three consecutive years.For this reason, the focus of the results is on the tillage aspects of the two farming systems (CA vs. conventional agriculture), and how this changed over the three sampling years.
No other main effects or interactions were statistically significant.Weed biomass under RT systems was low in the third cultivation year (weed biomass for cultivation years 1 and 2 was not determined), but increased significantly by the end of the fifth cultivation year (Figure 2).Under CT, however, weed biomass was initially high, but it stayed more constant, possibly showing a tendency to decrease (Figure 2).The results suggest that a temporal variation can be expected; with an increase in weed biomass under RT practices, while under CT practices weed biomass was more stable over cultivation time.Similarly, we detected a temporal trend in weed species diversity, where species diversity increased under RT but decreased under CT (Figure 3).Following this trend we would expect a time effect on species composition, with higher diversity the longer the trial continues.Indeed species diversity for the last year ( 5th year of treatment) suggests that RT had a higher species diversity than CT (Table 2).CT also had a low Evenness index (E), which suggests that CT is dominated by a few weed species, but that these species occurs in high abundance, while RT had a lower E value and hence higher diversity, but at lower abundances (Table 2).
The interaction between tillage and cultivation year is illustrated in the cluster analysis (Figure 4).Cluster analysis assigns treatments into groups, thus clustering similar treatments together.Three distinctive clusters were identified, with a RT-only cluster on the left (exclusively from cultivation years 3 and 4), a CT-only cluster on the right (mainly from year 3 and 5), and thirdly, a mixed cluster (consisting of both CT and RT) closely linked to the CT-only cluster.It is interesting to note that the RT treatments included in the mixed cluster are all from cultivation year 5, where the biomass was, contrary to previous cultivation years, exceptionally high.The dendrogram successfully illustrated the shift in weed biomass dynamics under CT and RT from the third to the fifth cultivation year.
The results concur with other authors (  1991 ;Vogel, 1995;Mashingaidze et al., 2012) who also recorded increased weed biomass with cultivation time under RT practices.The relatively high weed biomass (14.69 to 79.51 g m -2 ) corresponds with results from Vogel (1995), who reported values ranging between 20 and 220 g m -2 on small-scale farms, as well as Teasdale et al. (1991) with values between 48 and 623 g m -2 on a maize-based system.Furthermore, the temporal variation in weed biomass under different tillage practices concurs with Swanton et al. (1999) who reported that weed biomass varied between tillage practice and cultivation year.In conventionally tilled soils, this can be explained by the increase in environmental variables, such as temperature and moisture, as a result of tillage.Increased environmental variables could lead to more favourable conditions in certain years, and this in turn could lead to a large year-to-year variation in weed density (Clements et al., 1996).Additionally, tillage could increase the soil temperature and stimulate weed germination (Froud-Williams et al., 1984).Weed biomass in tilled soils could thus be a function of burial depth, periodicity of emergence, climatic conditions such as rainfall and temperature (Froud-Willliams et al., 1984;Teasdale et al., 1991;Mashingaidze et al., 2012) or  density dependence due to change in spatial distribution (Garcia de Leon et al., 2014).In contrast, under RT soils, weed seeds are often buried shallower, which might lead to more weed germination.Weed germination under RT systems would more likely be linked to rainfall than temperature, as is the case in CT (Froud-Williams et al., 1984).Thus, weed biomass, emergence or diversity might be increased or decreased in both tillage systems, but for different reasons.It is therefore possible to see large year-to-year fluctuations between RT and CT as the specific weed seed bank changes, or certain climatic factors favour one or the other system (Menalled et al., 2001;Murphy et al., 2006).

Weed diversity
The CT system was annually dominated by the two pioneer weed species, that is, large thorn apple and purple nutsedge, which when combined, contributed to 87.2% (yr 3), 75.1% (yr 4) and 78.5% (yr 5) of the total weed biomass (Figure 5).Ploughing seems to be conducive for these species, and one possible reason therefore might be the succession of a plant community that must start anew each year after ploughing, resulting in these dominant pioneer plants germinating first.However, under RT systems, there was a temporal increase in the relative weed diversity (Figure 5) largely due to the successional trajectory of the weed community (Murphy et al., 2006).Weed communities are indicative of ecological succession, and since a reduction in soil tillage leads to a reduction in agro-ecosystem disturbance, the associated change in weeds can be measured accordingly (Clements et al. 1996).Under RT the dominant weeds (large thorn apple and purple nutsedge) actually declined in biomass (34.8% [yr 3], 28.3% [yr 4], 13.4% [yr 5]; Figure 5).The decrease in pioneer weed species under RT coincided with an increase in other weeds, such as flax-leaf fleabane, khaki weed, khaki bur weed, common blackjack and devil's thorn (Figures 3 and 5).

Crop performance
Since this study was not set up to investigate the impact of weeds on crop performance, our results on crop performance here should be interpreted cautiously.For example, we did not have control sites, and as such we cannot investigate the effects of weeds on grain yields.Nonetheless, using a linear regression model heuristically suggested that in this trial weed biomass had a negligible effect on grain yield (Figure 6).We provide two, not necessary mutually exclusive, hypotheses for this observation.First, rainfall during the trial was poorly distributed, especially during key grain growing periods.Such varying rainfall could have caused low grain performance, irrespectively of weed biomass (Roberts, 1984).Secondly, weeding was done timeously before interference with critical developmental stages.According to Hall et al. (1992), maize yields will not be reduced if weeding is done between the 3 and 14-leaf stage, depending on climatic and site conditions.Overall our observation concurs with Murphy et al. (2006), where increased weed species composition under commercial no-till systems did not result in significant crop losses.Nonetheless, in smallholder CA systems increased weed biomass and diversity could lead to an increase in the amount of labour required for weeding, or alternatively requires increased use of herbicides (Giller et al., 2009; CT Andersson and D'Souza, 2014), which can be a major limitation in implementation of CA on small-scale farms.

CONCLUSION AND MANAGEMENT IMPLICATIONS
After five cultivation years of CA practices, a measurable shift in weed biomass and species composition was observed.Low weed diversity under conventional farming practices (ploughed soils), changed into a diverse weed population under CA practices in a relatively short time period.Cover crops and crop rotation did not have an effect on weed biomass, due to low and poorly distributed rainfall that resulted in low biomass production from previous years, in turn leading to ineffective soil cover and weed suppression.Tillage and cultivation year, however, did have an effect, and while weed biomass was initially low under RT in cultivation year 3, it had increased considerably by cultivation year 5. Species composition also changed; the two main pioneer weeds (large thorn apple and purple nutsedge) made up a decreasing fraction of the weeds under RT, while under CT, more than three quarters of the weed biomass consisted of these two weeds.Adaptable weed management (where continued monitoring of a system should influence the decision-making process and can be changed as needed, depending on available resources) or integrated weed management (combination of biological, chemical and mechanical weed management) should be practised for effective weed control.Such weed control programmes can include crop rotation and application of mulch (Teasdale et al., 1991;Swanton and Murphy, 1996).Cover crops or mulch only suppress weeds effectively when the cover is adequate, and should cover at least 30% of the soil surface (CTIC, 1999).

Figure 1 .
Figure 1.Crop biomass under reduced (RT) and conventional tillage (CT) for three consecutive years during weed sampling.

Figure 2 .Figure 3 .
Figure 2. Effect of tillage practice (CT = conventional tillage; RT = reduced tillage) and cultivation year on weed biomass (g m -2 ) at the Zeekoegat field trial, 2009-2012.* Significant differences are indicated by different letters (a,b, c and d).

Figure 4 .
Figure 4. Dendrogram illustrating the clustering of weed biomass under CT and RT from the third to the fifth cultivation year.

Figure 5 .
Figure 5. Temporal variation in weed species composition expressed in percentage of biomass, under conventional tillage (left) and reduced tillage (right) practices, for a long-term trial at Zeekoegat.

Figure 6 .
Figure 6.Effect of weed biomass on maize grain yield under conventional tillage and reduced tillage practices, for a long-term trial at Zeekoegat.

Table 1 .
Site-specific rainfall data for Zeekoegat experimental farm (South Africa) for the period 2009 to 2012 indicating rainfall as corresponding with crop planting dates and weed sampling dates.
*Season started in June and ended in May.