Upland rice growth and yield response to weed management practices under rainfed conditions in Morogoro , Tanzania

Field experiments were conducted in two seasons at the farm of Sokoine University of Agriculture in Morogoro, Tanzania (6.85°S; 37.64°E and 568 m.a.s.l.) during the short rain (November 2014 to January 2015) and the long rain (March to June 2015). The experiment was a split plot in randomized complete block design (RCBD) with 4 replicates. Weed management practices (herbicides, hoe weeding (3x) and weedy) were the main plot treatments; four rice genotypes (NERICA-1, NERICA-4, NERICA-7 and Mwangaza) were the subplots. Significant differences (P<0.05) were recorded on weed counts. Dominant weed groups as determined by Summed Dominance Ratio (SDR) in both experiments were broadleaf species (50.8%), sedges (25.2%) and grasses (24.0%). Post-emergence (8.6%) and hoe weeding (12.3%) significantly reduced weed dry biomass as compared to pre-emergence (17.8%) and weedy (61.3%) treatments in both experiments, respectively. Significant differences (P<0.05) were recorded among the rice variables. Data showed that Mwangaza and NERICA-1 had the tallest and shortest plant height (129.8 and 39.1 cm), respectively in both experiments. The highest and lowest tiller (35.3 and 7.5 m 2 ) count was recorded for both these genotypes, respectively. The lowest and highest LAI (2.5 and 4.5) were recorded on Mwangaza and NERICA-7 respectively; and NERICA-7 had the highest and lowest straw biomass (1603 and 305.1 g/m 2 ) in both experiments. The highest rice grain yield were recorded for NERICA-1 on hoe weeded plots and plots applied with post-emergence herbicide (2187.5 and 1562.5; 4176.1 and 4630.6 kg/ha) as compared to plots applied pre-emergence herbicide and weedy plots (965.9 and 0.0; 3323.8 and 0.78 kg/ha) in 2014/2015 and 2015 experiments, respectively. The highest return on investment, 3 352 846 Tanzanian shillings (Tshs) was obtained on NERICA-1 in post-emergence herbicide plots, and this was also similar (P<0.05) to hoe weeded plots. Post-emergence herbicide was also effective in weed control and had significant effect on profit analysis. This treatment/practice should be used in combination with hoe weeding under integrated weed management for better weed control.


INTRODUCTION
Rice (Oryza sativa L.) is the most important cereal crop in agriculture and the economy of the world (MOAC, 2007).According to FAO (2008), one third of the world's population depends on rice for 50% of their daily caloric intake.Tanzania is the second largest producer of rice in Southern Africa after Madagascar with a production level of 1.1 million tons.In Tanzania, rice is the second most important food and commercial crop after maize and among the major sources of employment, income and food security for farming households (FAOSTAT, 2010).The rice cultivated area by 2012 was 720 000 hectares with a very low 10-year (2003-2012) average yield estimated at 1.8 tons per ha as compared to Madagascar where production is 2.5 tons per hectare (FAO, 2014).However, imported rice is considered inferior in quality as compared to local rice by consumers; therefore, imported rice is sold at lower prices as compared to domestic rice (Minot, 2010).Agriculture is the backbone of the Tanzania's economy through employment, food production and export (MAFC, 2011).Upland rice is an important cash crop in many areas of eastern and southern Tanzania; including Morogoro (Kinyau et al., 2013).Morogoro is one of the major rice producing regions in Tanzania and rice production in Morogoro accounts for 45.6% of the total rice produced in Tanzania (RLDC, 2009).
There are many environmental effects and practices to increase the plant products such as irrigation (Yazici and Babalik, 2016), weeds and fertilization.Weeds are a major constraint in rice production and cause huge yield reductions in rice, millet, sorghum, maize and cowpea (Maiti and Singh, 2004).Globally, 9 to 32% yield losses in rice is attributed to weeds (Oerke and Dehne, 2004).Weeds are among the greatest yield-limiting constraints to rice production in Africa (WARDA, 1996) including Tanzania (Anwar et al., 2011).In irrigated production systems where rice is directly seeded, weeds are the major yield constraints (Becker et al., 2003).Ramzan (2003) reported yield reduction due to weed infestation up to 48 and 53% in transplanted and direct seededflooded rice, respectively.Sunil et al. (2010) also reported that season-long weed competition in direct seeded rice may cause yield reductions of up to 80%.
Weed infestation has been mentioned as a major cause of the yield gap under rain fed agriculture in the tropics, thus contributing to about 25% yield losses in cereal crops according to Affholder et al. (2013).Inappropriate weed management practices have been highlighted as constraints in rice producing regions in Tanzania including Bagamoyo and Morogoro where hoe weeding was found to be the most preferred management option for small holder farmers due to cost and the lack of basic knowledge on the use of modern agricultural technologies (Mkanthama, 2012).Weed control is important to prevent losses in yield and production costs and to preserve good grain quality (Zhang, 2001).It is important to develop effective weed management strategies to control the damage of weeds in rice fields.Effective control and management of weeds in upland rice farming will enable farmers to maximize and enhance sustainable rice production.This study seeks to identify appropriate and effective weed management strategies to help reduce losses caused by weeds and thereby optimize yield and profitability.

MATERIALS AND METHODS
Field experiments were conducted over two seasons at the farm of Sokoine University of Agriculture in Morogoro, Tanzania (6.85°S; 37.64°E and 568 m.a.s.l.) during the short rain (November 2014 to January 2015) and the long rain (March to June 2015).The experiment was a split plot in a randomized complete block design (RCBD) with 4 replicates.Weed management practices (preemergence 2, 4-D 720 EC, post-emergence Hansunil 600 EC, hoe weeding (3x) and weedy) were the main plot treatments and four rice genotypes (NERICA-1, NERICA-4, NERICA-7 and Mwangaza) were the subplots.
Fertilizer applications were done according to current agronomic practices.Nitrogen was applied at the recommended rate of 100 kg/ha, namely 50% (50 kg N/ha) of the recommended N was applied 21 DAS as a basal application and the remaining 50 kg N/ha was applied as a topdressing 35 DAS using urea (46% N).All fertilizer application was done using the broadcasting method (Kanyeka et al., 2007).
Weed counts were done 20 days after sowing using a 0.5 x 0.5 m quadrat placed randomly in the net harvest area (3.52 m 2 ) in each subplot.Two counts were made in each subplot and the calculated average was recorded.Weed counts from each quadrat were summed to find a total number of weeds by plant group (broad leaves, grasses and sedges species).This was done before and after application of treatments.Six weeks after sowing or 42 days after sowing, weed counts was also done by using 0.5 x 0.5 m quadrat.This was done before the second hoe weeding.Sampled weeds were classified according to species.Weed dry biomass was determined at 63 DAS by throwing 0.5 m quadrat at either ends of each subplot.The weeds inside the two measured 0.5 m quadrat areas were uprooted and arranged by weed group and weed species.This was oven dried for 72 h at 70°C.The 0.5 x 0.5 m quadrat was preferred in order to get the full representation of weeds density within a given measured area considering 1 m 2 quadrat as the measurement for the determination of the size of weed populations.The contribution of individual weed species to the weed community was determined by the summed dominance *Corresponding author.E-mail: davidkolleh29@gmail.com or kollehd@ymail.com.Tel: +231 777 061 345 or +231 886 232 The exchange rate for Tanzanian shillings (Tshs) to US$ during the experiment: Tanzanian shillings 2000 to US$ 1, Qty = quantity.ratio (SDR) calculated using relative density (RD) and relative dry mass (RDM) (Janiya and Moody, 1989) as follows: Where Five plants from the net harvest area (3.52 m 2 ) of each subplot were randomly selected and tagged.The height of each tagged plant was taken at three intervals: 39, 62 and 83 days after sowing using a 200 cm ruler.Plant height was determined by placing a meter ruler at the soil surface to the tip of the flag leaf of each tagged plant and the mean calculated and recorded.The five plants selected were used to record all other rice variable data such as tillers, panicle, and leaf area index and spikelet fertility.Grain straw was determined as the ratio of dry grain yield to dry straw mass and this was measured by the given ratio: (4) Rice yield was harvested from 3.52 m 2 in the middle of each subplot sun dried, weighed and recorded as grain yield.The profit margin was determined by subtracting the cost of production (which includes cost of seed, land preparation, herbicide application and labour for planting, hoe weeding, bird scaring, harvesting and processing) from the revenue derived from the sale of paddy rice.Net profit was calculated by subtracting the cost of production from the gross return.Return on investment was calculated as described by Jolly and Clonts (1993) as stated below: (5) The inputs or variable costs are shown in Table 1.

Statistical analysis
Data obtained from the experiments were subjected to statistical analysis of variance (ANOVA) using the computer programme GENTSAT statistical package 14 th edition (Payne et al., 2009).The treatments mean separation were done using Tukey's honestly significant test (Abdi et al., 2009).Daily data for rainfall (mm), minimum and maximum temperature (°C) and % relative humidity (RH %) were collected from Tanzania Meteorological Agency (TMA), at SUA station in Morogoro.

Rainfall (mm)
The total rainfall during the growing seasons is indicated in Figure 1.The highest rainfall during the first experiment was 155.5 and 84.6 mm in the month of December 2014 and January 2015, respectively.The repeated experiment started with a high rainfall of 144.3 mm in the month of March 2015 (Figure 1).

Temperature (°C) and relative humidity (%)
The recorded mean maximum temperature during the growing season was 30.7°C while the mean minimum temperature during the period was 22.1°C, respectively.Relative humidity ranged from 74.4 to 84.6% for December 2014 to January 2015 as shown in Figure 2. The mean RH during the growing season was 90.4% Figure 3.
In the study conducted during the short rain of 2014/15 and long rain of 2015, weeds observed in the experimental plots were composed of broadleaf, grasses and sedges as listed.Significant differences (P<0.05) were recorded for weed counts among weed management practices.Cyperus rotundus, Echinochloa colona and Cyperus eculentus were the most prevalent weed species in the 2014/15 experiment (39.7, 33.9 and 26.4%) respectively and Amaranthus retroflexus, Panicum maximum and Cyperus eculentus were recorded as the most prevalent weed species in 2015 experiment in species (37.6, 34.7 and 27.7%), respectively (Tables 4 and 5).
Broad leave weeds were the most dominant group and grass species, the least dominant in both experiments.In the 2014/15 experiment, sedges were recorded in preemergence plots as the second most dominant group of weeds and grasses were recorded in post-emergence treatments; hoe weeding and weedy plots as well.In the 2015 experiment, plots unto which pre-emergence, post-  emergence and hoe weeded were applied, showed sedges as the second most dominant weed group recorded, while grasses were the second highest in weedy plots (Figures 4 and 5).
The weed free treatment produced maximum rice yield in both years but 2015 was greater as compared to 2014/15 (4630.6 vs. 2272.7 kg/ha).This might be attributed to better growth of plants on account of reduced weed competition at critical crop growth stages resulting in increased availability of nutrients, water and light.All the weed control treatments in 2014/15 experiment significantly (P≤0.05)increased tiller numbers, LAI, straw dry biomass, panicles, spikelets and filled grains and ultimately the yield over weedy plots; but in 2015 experiment, tiller count, LAI, straw dry biomass, panicle count, spikelet count and filled grains were nonsignificant.In both years, the weed management practices significantly (P≤0.05)produced maximum tiller number (35.  6 and 7).Singh et al. (2005) reported similar results with the use of pre-emergence herbicide in rain fed direct seeded rice.
The highest grain yield (2187.5 kg/ha in 2014/15 and 4630.6 kg/ha in 2015) was recorded for NERICA-1 in both years, whereas, NERICA-4and Mwangaza were also statistically similar (Figure 3).This was perhaps due to high weed control efficiency of the treatments except weedy plots with low yield (Figure 3).Hansunil 600 EC had a significant influence on weed management practices and was closely followed by hoe weeding.In each case, the use of Hansunil 600 EC followed by (3x) hoe weeding or other herbicides indicates that Hansunil seems to be an effective post-emergence herbicide for weed control in upland rice.The high efficacy of Hansunil 600 EC as post-emergence herbicide was reported by several authors (Moody, 1991;Valverde et al., 2001).In combination with hand weeding, it was reported to be effective in controlling weeds in rain fed direct seeded rice (Ramamoorthy et al., 1998 andSingh et al., 2005).Rainfall pattern in the second experiment was well distributed during the crop growth period and resulted to better crop performance.The experimental plots were heavily infested with Cyperus species which subsequently reduced yield in 2014/15 experiment (Tables 2 and 3).

Economic analysis
Different weed control practices involved different costs which affected total production costs.Hansunil 600 EC post-emergence herbicide was effective in weeds control with the cost of Tshs (291 249 ha -1 ). Hoe weeding was laborious and more expensive; however, hoe weeding three times gave maximum weed control cost of Tshs 640 174 ha -1 (Table 8).All the herbicide treatments gave lower cost of weed control; but pre-emergence of 2,4-D herbicide gave lowest weed control cost of Tshs 205,809 ha -1 for the experiments but recorded the lowest profit (Table 8).The benefit of post-emergence (37.7%) and hoe weeding treatments (33.9%) increased grain yield, reflecting a good level of control over weeds growing in the experimental plots.The pre-emergence herbicide performed less well, achieving (28.3%) greater yield over the weed control.These results are in line with findings by Mirza et al. (2007) who reported that hand weeding is laborious and gave higher weed control cost while the use of herbicide gave the lower cost of weed control.The highest revenue (2 505 580 Tsh and 6 482 840Tsh) for both experiments was obtained with post emergence plots due to higher grain yields (1789.7 and 4630.6 kg/ha -1 ), respectively and lesser cost of production (291 249 Tshs) as compared to three times hoe weeding with highest grain yield (2187.5 and 4431.8 kg/ha -1 ) and cost (640 174 Tshs) of production for the experiments, respectively.
Pre-emergence 2,4-D herbicide was applied once given low weed control cost (205 809), but this was not profitable as the grain yield (1306.8 and 3465.9 kg/ha) The yield was yield obtained (kg/ha), Pre x N-1=pre-emergence and NERICA, MWG = Mwangaza, post x NERICA=post-emergence and NERICA, hoe weeding x NERICA= hoe weeding and NERICA and cont x NERICA=control and NERICA and Mwangaza.
was lower than post emergence herbicide and hoe weeding for both experiments (Table 8).These results are supported by works of Upadhyay and Chaudhary (1979) who reported that hand weeding and hoe weeding was three times more economical than applying herbicide only.Average return on investment for both experiments ranged from 7.5 to 12.9% with the highest benefit cost ratio observed in pre emergence herbicide, (2,4-D).This finding is in line with reports by Chakraborty and Majumdar (1973) who obtained best economic return with 2,4-D.Sabio and Pastories, (1981) also reported that application of herbicides was more economical than manual or hand weeding alone.

Conclusion
Weeds are a major constraint of the yield of upland rice.The present study revealed that Hansunil 600 EC postemergence herbicide, hoe weeding and pre-emergence 2,4-D 720 EC herbicide treatments provided a level of control as compared to the weedy plots (37.8, 33.9 and 28.3%), respectively.Although, hoe weeding was an effective means of control, Hansunil post-emergence herbicide was more economical, delivering a return on investment of 37.8%.Weeds can be effectively and economically controlled in upland rice using Hansunil 600 EC post-emergence herbicide.The most effective weed management practices were hoe weeding and postemergence herbicide treatments which resulted in the attainment of high grain yield and subsequent high returns on investment.These treatments offer alternatives to both resource-poor and large-scale producers, respectively.

Figure 1 .
Figure 1.The mean monthly values for maximum and minimum temperature, and rainfall for the growing season of 2014/2015 in Morogoro, Tanzanian.

Figure 2 .
Figure 2. The mean monthly values for relative humidity and radiation for the growing season of 2014/2015 in Morogoro, Tanzanian.

Figure 3 .
Figure 3. Weed counts before and after the application of treatments during 2014/15 experiment.Bars denote standard errors.

Figure 4 .
Figure 4. Weed counts before and after the application of treatments during 2015 experiment.Bars denote standard errors.

Table 2 .
Weed species recorded, total number, and summed dominant ratio during the 2014/15 experiments in Morogoro, Tanzanian.

Table 3 .
Weed species recorded, total number, and summed dominant ratio during the 2015 experiments in Morogoro, Tanzanian.

Table 4 .
The mean of broadleaf, grass and sedges number weeds/m 2 during 2014/15 and 2015 experiments in Morogoro, Tanzanian.

Table 5 .
Contd.Figures followed by the same letter (s) in the marginal and interaction means are not significantly different at P< 0.05 according to Turkey's test.

Table 7 .
The mean of panicle number, spikelet number and percent filled grain/m 2 during 2014/15 and 2015 experiments in Morogoro, Tanzanian.Figures followed by the same letter(s) in the marginal and interaction means are not significantly different at P< 0.05 according to Turkey's test.

Table 8 .
Grain yield and net return as influenced by different weed management practices at SUA, Morogoro, Tanzania.