Hand weeding is the most common weed management practice on small- to mid-scale diversified organic farms. While exceptionally effective, high labor costs make hand weeding an expensive input. Moreover, time required for high levels of weed control by hand increases with increasing weed density. Hand weeding may rely on pushed, wheeled tools, long-handled tools, short-handled tools, and/or hand pulling. We conducted eight field experiments, measuring working rate (i.e., row-feet weeded per minute) and efficacy (i.e., proportion of weeds controlled) in a standardized crop/surrogate weed system of corn and condiment mustard (Sinapis alba, ‘Idagold’). Wheeled tools generally had highest working rates, but occasionally lower efficacy than other tools or hand pulling. Importantly, working rates for wheeled tools were independent of weed density. Thus, even though efficacy may only average 60%, wheeled tools should be used before other hand methods because of their very high working rates. Long-handled tools may offer improved efficacy over wheeled tools, but generally with lower working rates. Short-handled tools and hand pulling offer potentially complete weed control, but with increasing time proportional to weed density.
Overall, wheeled tools should be the first step in a hand weeding program, followed by long- and then short-handled tools, with hand pulling a final step where very high efficacy is required. A comprehensive weed management plan focused on reducing the weed seedbank will result in both improved weeding outcomes with the use of hand tools, and lower hand weeding costs. Surprisingly, qualitative surveys of hand tools indicated a high level of variation in user preference. The Glaser® stirrup hoe was top-ranked in aggregate user scores for “feel,” “efficacy,” and “overall,” followed closely by the Glaser® wheel hoe. Contrary to expectations, tool rankings were, with a few minor exceptions, generally unaffected by gender, age, years of experience or scale of enterprise. It was difficult to get users to carefully evaluate many of the tools; some they would pass over based on observation or after only a brief test, moving quickly to a tool they were more interested in. Future qualitative tests should consider using focus groups or other “expert panels” to evaluate groups of five to ten tools that have a similar intended use (e.g., precision weeding; control of large weeds; wheeled tools).
Hand weeding is very common on small- to mid-scale diversified organic vegetable farms. The importance of hand weeding relative to other methods is evident in a “Wordle” (http://www.wordle.net/) of the 2010 Maine Organic Farmers and Gardeners organic certification data field for “How do you manage weeds?” (Figure 1). While simple and effective, hand weeding is costly and costs are proportional to weed seedling density (Figure 2; van der Schans and Bleeker, 2006). Furthermore, efficacy, i.e., the proportion of weeds killed by a weeding event, is independent of initial weed density. In other words, a proportion of weed seedlings are killed by a particular event; thus, a proportion of weeds also survive. Efficacy is typically measured by counting weed density in a known area before and again 24 hours after a weeding event (e.g., Figure 3). If the weed seedbank is high, it may take several cultivation events to reach an acceptable weed density. For example, efficacy of a typical row-crop cultivator averaged 0.78 over 64 locations in a 5-acre field (Figure 4; Gallandt, unpublished). Where the initial weed density was low, e.g., 10 seedlings per square meter, the density of surviving seedlings would be 2 m-2. However, with an initial density of 40 seedlings m-2 even this relatively high level of weed control, 78%, results in many surviving weeds (9 m-2).
There are essentially two solutions to this problem of density independent efficacy:
1) Reduce the weed seedbank (Gallandt 2006); or
2) Increase cultivation efficacy and/or working rate to make multiple events more affordable.
Weed seedbanks vary widely across Northern New England organic farms, ranging from a low of approximately 2,500 germinable seeds m-2, to over 25,000 m-2 (Figures 5 and 6; Gallandt, unpublished). Practical strategies to reduce the germinable weed seedbank begin with effective cultivation. Targeted fallowing, timed to coincide with peak emergence potential of problem species is also foundational for managing the seedbank. Also, in diverse enterprises, intensive sequences of very short-season crops can incorporate frequent soil disturbance, which encourages weed germination, while completely preempting weed seed rain. The combined approach of encouraging germination losses or “debits” to the seedbank, while preventing seed rain “credits,” is described in a recent webinar by the PI: http://www.extension.org/pages/62445/cultivation-and-seedbank-management-for-improved-weed-control-webinar
While large-scale and specialized growers have tractor-mounted options for precision-guidance systems to improve working rates, there have been few advances in technologies for small-scale growers who rely on hand tools (e.g., Figure 7). The objective of this project was to evaluate efficacy and working rates of hand weeding, simple hand hoes, and wheel hoes, and users’ preferences for these tools.
Our interest in innovative hand tools started with a NE SARE Partnership Project (ONE09-098), “Evaluation of Scale-Appropriate Weed Control Tools for the Small Farm:” https://projects.sare.org/sare_project/ONE09-098
In this project we compared hand weeding, short- and long-handled tools, a typical wheel hoe, and an innovative set of tools from Finland called the Weed Master (Figure 8). Unexpectedly, efficacy was similar with each of the tools, although there was greater variability with the wheeled tools (Figure 9). More remarkable was the greater working rates of the Weed Master compared to hand pulling, hand hoes and the Glaser wheel hoe (Figures 10 and 11). Although working rates with the Weed Master were highly variable (Figuere 11) mostly due to various operators in these experiments, the 4- to 10-fold greater working rates demonstrated that innovative tools could contribute to improved weed management for small-scale producers (e.g., Figure 12).
While results with the imported Weed Master were encouraging, we realized during this project that there is a great diversity of hand and wheeled tools available domestically and that it would be a great service to small-scale farmers and gardeners to have some objective information available when choosing tools. To this end, our current project aimed specifically to test the following hypotheses:
1) That innovation in hand tools could offer a level of efficacy similar to hand weeding;
2) That wheeled-tools could offer significantly greater working rates than hand tools;
3) That working rates of hand weeding and hand tools is density dependent whereas working rates with wheeled tools is density independent; and
4) That selected tools have a superior design, weight, finish and “feel” that make them more preferred than other tools.
To test these hypotheses we conducted a series of research station and on-farm experiments in 2010 through 2012.
- Figure 2: Weeding hours vs. weed density.
- Figure 4: Cultivation controls a proportion of weeds present. The density of surviving weeds thus increases with increasing initial densities.
- Figure 10: Measuring working rate of a short-handled tool, i.e., “row-feet weeded per minute.”
- Figure 12: On-farm evaluation of the Weed Master compared to colinear hoes.
- Figure 1: Wordle of MOFGA farmer “weed management methods.”
- Figure 3: Measuring “efficacy,” i.e., proportion of weeds controlled by a weeding event.
- Figure 5: Representative soil samples showing variability in germinable weed density and community; i.e., “weed seedbanks.”
- Figure 6: Weed seedbank densities on 21 northern New England organic farms (Gallandt, unpublished).
- Figure 7: One hundred years of minimal innovation in hand tools.
- Figure 9: Efficacy of hand weeding, and 7 hand tools, in two experiments.
- Figure 11: Working rates of hand weeding and seven hand weeding tools in two experiments.
- Figure 8: Hand weeding and hand tools, including the Finnish Weed Master, field tested in a previous NESARE-funded project.
- 1. Conduct qualitative evaluations to compare several familiar, traditional hand tools to newer innovative designs;
2. Conduct quantitative evaluations to determine how working rate and efficacy relate to qualitative factors; and
3. Determine through the interview process, perceptions that contribute to tool selection, and how this relates to the adoption of new technology.
Eight field experiments to measure working rate and efficacy of hand weeding tools (e.g., Figure 13) were conducted in 2010 through 2012 at the University of Maine Rogers Farm in Stillwater, ME (44 93’N, 68 69’W). The alluvial soils at this site are highly variable; experiments included fields subject to high resolution mapping as the following soil types: Lamoine silt loam, Boothbay silt loam, and Swanton very fine sandy loam.
We used standardized methods, with minor modifications, throughout. Field preparation consisted of moldboard or chisel plowing, tandem offset disking, and finishing with a light field cultivator. To test the hand weeding tools in the context of a row-crop, we used field corn (Zea mays L., ‘VK13’, Lakeview Organic Grain, Penn Yan, NY), planted in 81.3 cm rows at a population of approximately 50,000 plants per ha. Following corn emergence (V1 growth stage), condiment mustard (Sinapis alba, ‘Idagold’, Johnny’s Selected Seeds, Fairfield, ME) was broadcast at 5.6 kg ha-1, and tine-harrowed (Lely spring-tine harrow) to incorporate. Mustard is a useful surrogate weed when uniform populations of relatively even-aged individuals are desired (e.g., Kolb et al., 2010; 2012). We found it particularly useful for these studies of cultivation efficacy as it emerges quickly, uniformly, and target populations are generally attained. Resident weed populations are generally very patchy, adding undesirable within-plot variability that can reduce statistical power or require costly increases in sampling size and/or number. Idagold is similar in plant stature and architecture to wild mustard, Sinapis arvensis L., but with nearly twice the seed mass, and a predictable lack of dormancy (Kolb and Gallandt, 2013). The larger seed results in a larger, more robust seedling, and thus, our efficacy estimates are likely lower than would be expected for most dicotyledonous weed species.
We expected variability in cultivation efficacy to be explained by operators (an effect removed by blocking in our experimental design), tool (our treatments), possibly weed density, as well environmental conditions, including soil roughness, texture (i.e., proportions of sand, silt and clay), and soil moisture at the time of cultivating. Soil surface roughness and texture were measured within two days prior to cultivation experiments.
In 2010, soil surface roughness was measured using the chain method (Saleh, 1993) whereby a 1-meter roller chain (ANSI No. 35) was carefully laid along the micro-relief of the soil surface, following the direction of tillage, and the length subsequently measured. Soil surface roughness is estimated by calculating the proportional change in chain length (1-(L2/L1))*100, where L1 is the chain length and L2 is the chain length after being laid along the irregularities of the soil surface. Because the roller chain failed to accommodate the relatively minor variation in surface roughness in our fields, in 2011 we switched to a 2-mm stainless steel, beaded dog tag chain (91.44 cm long; Amazon.com).
Soil texture was measured using the Bouyoucos Hydrometer method (Day, 1965). Briefly, 50 g of dry soil was dispersed with 1 N sodium hexametaphosphate and temperature-corrected hydrometer measurements recorded at 40 sec. and 2 hours to estimate sand and clay, respectively; the proportion of silt was calculated by difference.
Soil moisture was measured using a hand-held ThetaProbe ML2X (Dynamax, Houston, TX). Three readings were taken just outside weed density census quadrats, immediately prior to cultivating; averages of the three readings were analyzed.
Cultivation experiments were performed using plots centered over a single row of corn, 30 to 60 m long, and 0.15 m wide; shorter plots were used if experiments featured short-handled tools; longer plots were used for experiments featuring long-handled and wheeled-tools. Each treatment was replicated 4 to 6 times based on resources available at the time of the experiment.
Experimental Design. Each experiment was established as a randomized complete block design; blocking accommodated possible human operator effects, not environmental variables as is more commonly the case. The contribution of environmental sources to variation in efficacy or working rate was tested by ANCOVA, correlation or regression. Homogeneity of variances were tested within experiments, and data were transformed to satisfy this assumption of ANOVA; actual means are presented. Levene’s test was used to compare variance between experiments where pooling and a combined analysis was desired.
Two experiments examined long-handled tools in 2010, the first conducted on July 27th, the second on August 11. Sown ‘Idagold’ mustard was the only species included in the pre- and post-cultivation censuses. Mustard was in the 1-2 leaf stage at the time of cultivation. Two widely used long-handled hoes, the Colinear Hoe and the Glaser Stirrup Hoe (5 in.), were compared to Hand Weeding, and several innovative alternative tools: the Hooke’n Crooke Hoe; Swoe; Loop Hoe; Ho-Mi Digger; and the Circle Hoe.
Working rate ranged from 0.05 to 0.57 m s-1 (Figure 14; mean = 0.30 m s-1). Treatment effects on working rate were similar over the two experiments (interaction P = 0.762). Surprisingly, only one tool, the Ho-Mi Digger, had a working rate greater than the Hand Weeded control (0.43 vs. 0.18 m s-1; P = 0.002). Increasing Working Rate resulted in decreasing Efficacy, i.e., mustard morality (P = 0.020), but this described a relatively small amount of the observed variation (r2 = 0.096). Mustard mortality was overall high, ranging from 63 to 97% (Figure 14; mean = 82%). However, efficacy was similar across the treatments (P = 0.172). Corn mortality was unaffected by treatment (P = 0.418), and was overall low (mean < 1% ).
Averaged over three field experiments conducted in 2011, the Glaser stirrup hoe had the fastest working rate (0.136 m s-1), similar to the Half-Moon, Three Tine, San Kaku, Diamond and Cobrahead hoes, but significantly greater than the Colinear Hoe and the Wire Weeder, both tools with relatively small heads that work on the pull stroke (Figure 15). As the stirrup hoe works on both the push and pull stroke, typically with a back-and-forth motion, we expected it to offer a greater working rate than each of the other hand tools, each of which work only on the pull stroke.
Efficacy ranged from 55% for the Three Tine weeder, to 76% for the Wire Weeder; Hand Weeding averaged 85% (Figure 15). Corn mortality was similar across all treatments, averaging 4.4% (P = 0.1366; data not shown). Not surprisingly, efficacy declined with increasing working rate (considering only the hand tools, not the control; P = 0.026), however, this explained only a relatively small amount of the variability observed (r2 =0.05) (Figure 16).
The high degree of variability in efficacy in our cultivation experiments was unexpected considering our use of a surrogate weed and thus a generally even-aged cohort of one species. Considering this, we speculated that soil conditions may explain much of the measured variability (e.g., Figure 17). However, we failed to detect a significant relationship between soil moisture at the time of cultivation and efficacy (Figure 18), nor between soil surface roughness and efficacy (Figure 19).
Wheeled tools were evaluated in separate experiments in 2010, which included both hand weeding and selected long-handled hoes for comparison. As expected, handing weeding was slowest, followed by long-handled tools, the single wheel hoe, the double wheel hoe, with then a large increase in working rate with the Weed Master implements (Figure 20). The Colinear hoe, Hooke’n Crooke, and Weed Master with disk hillers each provide a high level of efficacy (avg. = 0.77), similar to the hand-weeded control (Figure 21); the stirrup hoe and wheeled tools were comparatively less effective (avg. = 0.40; Figure 21).
In 2012 we expanded our experiment to include short-handled tools, finding relatively few differences in working rate or efficacy among tools, but, not surprisingly, differences between long- and short-handled tools (Figure 22). Working rates of long-handled tools were over twice that of short-handled tools, but with somewhat greater variability and a 12% reduction in efficacy (Figure 22).
Eighteen hand tools were featured in our qualitative assessments, which occurred at various field days and events (including the Maine Organic Farmers and Gardeners Common Ground Country Fair). At each event, tools were displayed in groups of three tools of similar design and attendees were provided brief instruction and access to a survey card (Figures 23-25). Tools were evaluated by a minimum of 11 users, up to a high of 29 (354 total surveys were recovered). Aggregate values for the qualitative ratings, a sum of scores for “feel,” “efficacy,” and “overall,” failed to identify a clear ranking of tools based on user’ opinion (Figure 26; Figure 27). We had hoped that qualitative scores would help others identify top-ranked tools, but our results indicate that there is a high level of individual preference that impacts users’ rating of various tools. This is further evidenced by selected quotes (Figure 28; see also umaine.edu/weedecology/).
Qualitative ratings were similar between gender (with a couple of exceptions: women preferred the Circle and the Homi hoes more than men; data not shown). There were no notable age effects on rankings, nor were there large differences attributed to years of experience of scale of enterprise.
Overall, the Glaser stirrup hoe was the top-ranked tool, followed by the Glaser wheel hoe, the Half-moon hoe and the Diamond hoe (Figure 26), but high level of variability indicates that for every person that loves a particular tool, there is likely another user that is not fond of it. In short, considering the relatively low cost of these tools (Figure 29), especially compared to labor costs, and the potential for improved working rates with effective and well-liked tools, it would be reasonable to field test and have on-hand several options for a diverse weeding crew.
Saleh, A. (1993) Soil roughness measurement: Chain method, Journal of Soil and Water Conservation 48, 527–529.
Day, P.R. 1965. Particle fractionation and particle-size analysis. Chap. 43 in Methods of Soil Analysis, Part 1. C.A. Black, ed. American Society of Agronomy, Madison. Pp. 545-567.
Kolb, L. N., Gallandt, E. R., and Molloy, T. (2010) Improving weed management in organic spring barley: physical weed control vs. interspecific competition, Weed Research 50, 597–605.
Kolb, L. N., Gallandt, E. R., and Mallory, E. B. (2012) Impact of Spring Wheat Planting Density, Row Spacing, and Mechanical Weed Control on Yield, Grain Protein, and Economic Return in Maine, Weed Science, BioOne 60, 244–253.
Kolb, L.N. and E.R. Gallandt (2013). Modeling weed dynamics in contrasting organic cereal production systems. Weed Research 53:201-213.
Van der Schans, D. and P. Bleeker (2006). Practical weed control in arable farming and outdoo vegetable cultivation without chemicals. Applied Plant Research, Wageningen UR (PPO 352; 77 pages).
Yadav, R. (2007) Development and Ergonomic Evaluation of Manual Weeder, Agricultural Engineering IX, 1–9.
- Figure 14: Working rate and efficacy of long-handled tools evaluated in 2010.
- Figure 17: Field conditions affecting cultivation efficacy.
- Figure 18: Lack of relationship detected between efficacy and soil moisture.
- Figure 19: Lack of relationship detected between efficacy and surface roughness.
- Figure 20: Working rates of long-handled and wheeled tools, 2010.
- Figure 26: Table listing rank order of hand tool scores, median, mean and standard deviations, and the number of surveys recovered (n).
- Figure 27: Box and whisker plots of combined qualitative scores for hand tools, demonstrating the high degree of variability in ratings.
- Figure 28: Selected quotes from assessment sheets.
- Figure 21: Efficacy of long-handled and wheeled tools, 2010.
- Figure 22: Efficacy and working rates of long- compared to short-handled tools.
- Figure 23: Qualitative evaluation of hand tools at a field day event.
- Figure 24: Blank and completed assessment cards attached to a tool stand.
- Figure 25: Assessment card filled out by participants of the qualitative evaluations of hand tools.
- Figure 15: Working rate and efficacy of long-handled tools, 2011.
- Figure 16: Efficacy vs. working rate in 2010 and 2011 field experiments.
During nine field day events, and presentations at winter meetings, this project has served to make small- to mid-scale growers aware of concepts of working rate and efficacy in the context of physical weed control. Analysis of areas to weed, and consideration weather and soil conditions favorable for weeding, i.e., “weeding days,” underscore the need to consider efficiency of weeding. Our datasets demonstrate clearly the high level of efficacy possible by hand weeding, but the working rate advantages realized in use of hoes, particularly long-handled and wheeled tools. Unexpectedly, tools with greater working rates were not always less effective, indicating that working rate should be a primary consideration, especially for a first weeding pass.
Education & Outreach Activities and Participation Summary
To demonstrate the dynamic action of each hand tool we created a brief (< 1 min.) video (36 in total) which are hosted on our YouTube Channel, “zeroseedrain:” http://www.youtube.com/user/zeroseedrain. As of May 20, 2013, these videos have been viewed 50 to over 500 times depending on the tool.
Cultivation and seedbank management for improved weed control. eOrganic webinar (7 February 2013). http://www.extension.org/pages/62445/cultivation-and-seedbank-management-for-improved-weed-control-webinar (516 views as of May 20, 2013).
Costs and Efficiency of Small-scale Weeding Equipment. New England Vegetable and Fruit Conference, Manchester, New Hampshire (15 December 2011; 200 attending).
Cultivation: Tools and Strategies for Improved Weed Control. Northeast Organic Farming Association of Vermont (NOFA-VT), Integrated Pest Management Series (16 February 2011; 30 attending).
Cultivation and seedbank management for improved weed control. Illinois Specialty Crop, Agrotourism and Organic Conference, Springfield, IL (7 January 2011; 60 attending).
Penobscot County Master Gardeners’ Demonstration Day. July 17th, 2012; 75 attending.
NOFA-VT Workshop: Low Energy Farming Systems: Replacing Fossil Fuels with Human and Animal Power – Cerridwen Farm, Rutland, Vermont. Included a lecture and workshop at Green Mountain College. June 15, 2012; 50 attending.
Ecological Weed Management. MOFGA Apprentice Training. Fisher Farm, Winterport, Maine. June 11, 2012; 30 attending.
Participant Evaluation of Short-handled Tools, Long-handled Tools, and Wheel-hoes. MOFGA Common Ground Country Fair. September 24, 2011; 50 attending.
Hand Tools for the Small Farm. MOFGA Apprentice Training. Crystal Spring Farm, Brunswick, Maine. August 1, 2011; 45 attending.
Qualitative and Quantitative Evaluation of Hand Tools. University of Maine Highmoor Farm Field Day. July 20, 2011; 60 attending.
Evaluating Hand Tools for the Small Farm. Small Farm Field Day. University of Maine Rogers Farm, Stillwater, Maine. June 29, 2011; 55 attending.
Demonstration of Experimental Design for Small Scale Weed Control Tools, Rogers Farm in Stillwater, Maine. University of Maine Cooperative Extension Organic Small Grains Field Day. July 1, 2010; 50 attending.
Scale Appropriate Weed Control Tools for the Small Farm. Peacemeal Farm, Dixmont, Maine. MOFGA Apprentice Training Program: Weed Ecology and Management. June 23, 2010; 75 attending.
Hand tools average $45; wheeled tools, excluding the Weed Master, average $255 (Figure 29). Considering these tool costs, labor at $10 per hour, and working rates (Figure 30), we conducted an example cost/benefit for the UMaine student CSA, the Black Bear Food Guild, who farm 0.4 acres, including 140 beds (14,000 total row feet) (Figure 31). Hand weeding in our experiments was, as expected, slowest, averaging 6 row feet per minute, compared to 152 row feet per minute for the Weed Master (Figure 32). In this scenario, even the most expensive tool, the Weed Master at $2,000, would be paid for in 16 weeding events due to improved working rate over typical long-handled tools (Figure 32).
- Figure 29: Cost of hand and wheeled tools.
- Figure 31: UMaine Black Bear Food Guild, student CSA.
- Figure 30: Working rate and tool cost.
- Figure 32: Cost/benefit of investing in weeding tools for the Black Bear Food Guild.
We did not conduct any post-project surveys to evaluate farmer adoption of any new hand tools based on this project. Our best evidence of impact has been strong interest of farmers, and others, evidenced by speaking invitations, and YouTube video traffic.
Areas needing additional study
Early in the project, in both qualitative surveys and our quantitative experiments, we recognized that tools varied greatly in their “feel” and many tools may be rated differently following an 8-hour day of use than the relatively brief field test used in our surveys. Moreover, tools likely vary in risk for repetitive motion injury, or in the efficiency of use (e.g., operator peak heart rate and rest/recovery time, see Yadav and Pund (2007)). We were not able to find a suitably trained cooperator to assist us with expanded measurements to address ergonomics of user-efficiency of the tools included in these experiments.