Training Livestock to Eat Weeds in the Tropical Pacific and Evaluating the Effects on Meat Quality for Stronger Ranch Profits

Final Report for OW11-309

Project Type: Professional + Producer
Funds awarded in 2011: $49,610.00
Projected End Date: 12/31/2014
Region: Western
State: Hawaii
Principal Investigator:
Matthew Stevenson
University of Hawaii
Expand All

Project Information


Given exceptionally high transportation costs, Hawaii ranchers and farmers have asked for more research and education on reducing or eliminating off-farm inputs. The cost of pasture weed control can greatly reduce ranch profitability, and inadequate weed control leads to unacceptable production losses. Multi-species stocking has greatly improved pasture use efficiency through differences in grazing behavior, though some plants remain problematic. A U.S. mainland developed method to condition grazing preferences shows good potential for expanding livestock diets but has been untested on tropical plants. The purpose of this project was to evaluate this training method for application on priority tropical weeds in a demonstration proof-of-concept approach. We attempted to train multiple classes of cattle, sheep, and goats to eat nine species of tropical plants previously avoided. These livestock successfully ate seven of the targeted species in test troughs and paddocks: Amaranthus spinosus (L.), Digitaria insularis (L.) Mez ex Ekman, Elephantopus mollis Kunth, Leonotis nepetifolia (L.) R. Br., Persea americana Mill., Schizachyrium condensatum (Kunth) Nees, and Sphagneticola trilobata (L.) Pruski. Senna occidentalis (L.) Link was rejected for testing owing to several reports in the literature of acute toxicity, and despite conditioned preference training, goats did not eat the base parts of Asclepias physocarpa (E. Mey.) Schltr., though sheep ate leaves in later training. Blood tests did not reveal any undue stress during the weeklong training period. Weed populations did not appear to change during the study period. One collaborating ranch was successful in adapting the initial training method as well as providing an energy supplement to increase cattle intake of a pasture weed. The U.S. mainland developed behavior modification method appears to increase cattle, sheep, and goat intake of various weeds of concern in Hawaii. Longer term impacts on established weed populations and long-term effects on livestock health should be priority areas of study to further refine this approach for application in practice. At least one collaborating ranch has adopted this training method for their operations and post-workshop evaluations indicated other rancher interest in this method.


Attendees of the 2007 Western SARE sub-regional conference in Hawaii identified reduction of imported farm inputs as the leading type of research and education necessary to help economically sustain farming and the environment over the next 10 years. Costs of farm inputs, such as herbicides, in Hawaii and the Pacific Basin are high owing to great transportation distances on top of application costs. Incorporating different types of livestock in a grazing system to exploit inherent differences in foraging behavior has had dramatic success in replacing herbicides for managing certain weeds. Despite these successes, some plants remain problematic as livestock avoid them altogether or do not eat enough of them to check their increase in the pasture. The next step in grazing management for weed control in the tropics is to realize the idea that grazing animals can eat certain weeds if they learn how. The purpose of this university-ranch collaborative proof-of-concept demonstration project was to evaluate a North American developed approach to train livestock to eat pasture weeds in a tropical environment.

Project Objectives:

We sought to achieve four objectives:

1) Trained cattle, sheep, and goats will consume unfamiliar foods.

Training cattle, sheep, and goats to eat a number of previously avoided pasture weeds, at least briefly in a test pen, has by and large been successful at four ranches. Detailed results for each site and species are reported below.

2) Livestock offered a free choice diet supplement will be able to eat weeds with high levels of secondary compounds.

This approach, offering a supplement such as a powder form of condensed tannins that may attenuate plant secondary compounds, was considered for increasing intake of Senna occidentalis at Kapapala Ranch. However, literature reviews revealed this Senna species to be highly toxic to cattle and goats. Given the success of goats controlling other target weeds, we decided to drop this plant from consideration. Energy supplements in the form of cracked corn and molasses were used at Haleakala Ranch for training cattle and sheep onto Digitaria insularis and Leonotis nepetifolia. Livestock readily ate both target species after overcoming neophobia via conditioned preference training.

3) Targeted weed populations will decrease with increasing grazing pressure from trained livestock.

We monitored plots established at three ranches and set up short-term photo-monitoring sites at a fouth. Most data showed modest changes in weed and dominant species cover or density over the project period. Several paddocks have low stocking densities and weed populations are patchy. Under these conditions, it has been difficult to tell if the lack of weed patch expansion has been due to grazing pressure, competition with dominant pasture species, or an environmental response. Regardless, ranch managers were glad to see no noticeable increase in weeds of interest compared to previous years. Photo-monitoring of more intensely grazed sites showed noticeable impacts on target weeds, though long-term grazing management strategies for control and monitoring were outside the scope of this observation and proof-of-concept project.

4) Effects of consuming weeds on meat quality will be detectable.

This objective became untenable for reasons described in the following sections.


Click linked name(s) to expand/collapse or show everyone's info
  • Randall Cremer
  • Greg Friel
  • Dr. Yong Soo Kim
  • Lani Petrie
  • Dr. Fred Provenza
  • Tom Runyan
  • Dr. Ashley Stokes
  • Dr. Mark Thorne
  • Dr. Juan Villalba
  • Duke Wellington


Materials and methods:

Target Weed Identification: The first step in this project was to identify which plants would be targeted for preference conditioning and control attempts. After consulting with participating ranchers to identify priority weeds livestock currently do not use, we reviewed literature to identify acutely toxic plants. Through this process we eliminated Senna occidentalis (L.) Link and targeted the following (all plant nomenclature per USDA PLANTS database):

Amaranthus spinosus (L.) – Haleakala Ranch, Maui
Asclepias physocarpa (E. Mey.) Schltr. – Haleakala Ranch, Maui
Digitaria insularis (L.) Mez ex Ekman – Haleakala Ranch, Maui
Elephantopus mollis Kunth – Jurassic Kahili Ranch, Kauai
Leonotis nepetifolia (L.) R. Br. – Haleakala Ranch, Maui
Persea americana Mill. – Haleakala Ranch, Maui
Schizachyrium condensatum (Kunth) Nees – Jurassic Kahili Ranch, Kauai
Sphagneticola trilobata (L.) Pruski – Rocking W Ranch, Tom’s Goats, Jurassic Kahili Ranch; Kauai

Conditioned Preference Training: We followed methods described by Voth (2010) that were in turn based on principles of animal behavior and learning developed by Dr. Fred Provenza and colleagues participating in the Utah State University BEHAVE Network. The underlying principle in the training process is to establish in the animal a positive association with unfamiliar feeds. This is achieved by offering small amounts of a nutritious but unfamiliar feed in the morning followed by a different one in the evening. All animals used in this project had been entirely pasture-raised and, thus, commercial feeds such as cracked corn, timothy hay, alfalfa pellets, etc. were unfamiliar. In each case, livestock generally avoided feed and training tubs in the first morning feeding. The process of changing feeds each morning and afternoon is repeated for five days, rotating new feeds and feed forms (e.g. pellets vs. hay) at each offering. Feeds were placed in the same tubs or troughs at each offering with only enough tubs to allow for four to five animals each. By limiting tub space, the competitive nature of the animals is supposed to draw interest in the feeds. On Day 6, the morning offering is skipped with the target weed introduced in a mix with a nutritious supplement. The final day, Day 7, the morning offering is again skipped and only the weed is offered in tubs or troughs (Figure 1).

Pasture Monitoring: To observe any changes in weed populations, we established vegetation monitoring plots initially at all five participating ranches. Owing to changes in land tenure and management priorities, plots at two ranches were discontinued early into the sampling period. At Rocking W Ranch, two plots were set up in a S. trilobata dominated paddock, and two more plots were set up in a separate paddock with an incipient population. At Jurassic Kahili Ranch, two plots were set up in sites with moderate populations of E. mollis and two more plots in paddocks with small stands of S. condensatum. Kapapala Ranch had a slightly different objective where two plots were established in brush and tree stands browsed by goats, and two more were set up in similar sites where goats were absent. The center-posts of all monitoring sites were randomly placed within paddocks meeting our weed criteria of interest. Three 50-m long transects were established radiating at equidistant compass points from the center post. We determined herbaceous vegetation cover with the line-point sampling method recording each species at one meter intervals the length of the transects. Woody species and S. condensatum tuft density were estimated by counting the number of individuals per species within 1 meter of each side of a transect. Woody species were further stratified into size classes: A – individuals below 0.5-m tall, B – individuals between 0.5 – 1.0-m tall, C – individuals above 1.0-m tall. Density of dead brush species was also determined with this belt-transect method. Sampling began every quarter, but owing to logistical issues with management priorities of each ranch, sites on three islands, and rates of vegetation change that were slower than anticipated, the sampling period was changed to a twice annual basis starting February 2013.

Research results and discussion:

Objective 1: Trained cattle, sheep, and goats will consume unfamiliar foods.

Haleakala Ranch – Initial training efforts in October 2011 resulted in a mix of 80 steers and heifers successfully eating A. spinosus in a roughly 10 hectare infested paddock as part of a grazing rotation (Figure 2). However, the weed areas dramatically increased in cover by June 2012. Ranch managers were thus forced to take more aggressive weed management actions on that particular site. At the same time, 20 goats were trained targeting the base of A. phsyocarpa as literature indicated this part of the plant has the least amount of secondary compounds. Goats did not increase intake of A. physocarpa during or after training. Ranch managers decided to focus on L. nepetifolia instead. In June 2012, 18 weaned calves and 50 weaned lambs were trained with the same methodology as reported above. Both species readily ate L. nepetifolia by the end of the training period, both out of the tubs and in the test pasture (Figure 3). An interesting side note is that on the third day of training, three lambs completely defoliated a lone A. physocarpa bush in the training pasture (Figure 4). This pasture weed had not been introduced to the animals in the tubs. Ranch manager Greg Friel said he had never seen sheep or goats eat all of the plant (goats will eat the fruit only), and despite the 2011 trainin,g goats did not successfully adopt eating A. physocarpa beyond fruits. Due to severe drought by fall 2012, the sheep were removed from Haleakala Ranch for several months and sent to another ranch in a different region of Maui. Monitoring plots thus became futile lacking any treatment.

Kapapala Ranch – Between drafting the project proposal and visiting the ranch to conduct the project in October 2011, ranch managers were able to get grazing pressure on the initial weeds of concern without specialized training. However, all classes of animals on the ranch avoided S. occidentalis, and managers were interested in targeting this weed for control with livestock. On researching this genus, we felt reported toxicity to be a concern and did not conduct the training onto this plant. We considered testing this plant for Objective 2, offering supplements to attenuate plant toxins; however, discussion between Co-PIs and ranch managers led to the conclusion that this plant posed too much risk to the welfare of the animals. Therefore, vegetation monitoring sites were established to observe goat browsing impacts on Psidium guajava L. and Morella faya (Aiton) Wilbur, which are management priorities of the ranch.

Rocking W Ranch – The principal investigator conducted the training program with nine heifers in November 2011 (Figure 5). The target weed was S. trilobata, and animals ate S. trilobata during the training and in the training pasture.

Jurassic Kahili Ranch – In December 2011, the principal investigator conducted the training program with 10 heifers to train them onto S. condensatum. They successfully ate the weed in the trial pen. Cattle were kept in a dry lot for the duration of the seven-day training with the majority of their daily diet coming from commercial feed. Unfamiliar feeds and target weeds were fed as a small supplement as described in methods. Pasture intake of S. condensatum by these cattle was difficult to observe as the ranch regularly mows this weed to delay or prevent flowering. Informal and monitoring observations did not indicate a noticeable decline in S. condensatum over the study period.

We also trained 18 adult ewes and rams onto E. mollis at the same time as the heifers but in a separate temporary training paddock set up with electric net fence. The sheep were observed to eat the flowers only before the training period. However, with the training, sheep readily ate all parts of the plant in the training paddock. Both in informal observation and in the vegetation monitoring data reported below, E. mollis appeared to decline from pastures. While this could be from an environmental or other reason, given the degree of intake in test pens there is a good possibility the decline could be from grazing pressure. Sheep stocking density is low on the ranch, so if the decline is from grazing pressure this would indicate high selectivity for E. mollis by these sheep (Figure 6). Future studies should examine this further.

Tom’s Goats, Kauai – The principal investigator trained 15 does onto S. trilobata in a pen in December 2011 (Figure 7). Goats were given a daily ration of P. maximum from the neighboring paddock where their herd mates were grazing. After the training period, the goats readily ate S. trilobata in the pen despite avoiding it in normal rotation. Unfortunately, owing to land tenure issues, these goats were removed from this site and dispersed among other herds before in-pasture use of S. trilobata could be observed.

Animal Health – Blood was collected from representative animals on three ranches before the training and within three days of eating target weeds. Blood was sent to Antech Diagnostics (Irvine, CA) for general analysis primarily for blood urea nitrogen (BUN) as an indicator of kidney or liver distress due to toxicity. None of the ten heifers at Jurassic Kahili Ranch, nine heifers at Rocking W Ranch, or ten goats from Tom’s Goats showed dramatic increased or decreased levels of concern between sampling periods.

Objective 2: Livestock offered a free choice diet supplement will be able to consume weeds with high levels of secondary compounds.

Kapapala Ranch – As noted above, this additional step in the conditioned preference training method was considered for use on S. occidentalis. However, this plant was dropped from consideration out of an abundance of caution arising from reports of acute toxicity in multiple livestock species.

Haleakala Ranch – Based on previous experience from this project, ranch managers modified the training method described above to shorten the period to three days and used grain as both an energy supplement and positive experience for the animal in eating something new. In September 2013, three heifers, 10 sheep (mixed sex), and 10 goats (mixed sex) were penned and fed grain together as part of a separate bonding project. P. americana leaves were also introduced when the animals were fed (Figure 8). The Merck Veterinary Manual lists P. americana as toxic to livestock; however, these animals had free access to pasture as well as the energy supplement and could mix their diet possibly minimizing adverse effects. Informal observation revealed that all species preferred to eat leaves when first let into the pasture containing the tree before switching to grass. All species ate old leaves, fresh leaves, and fruits with no apparent toxicity issues.

In another variant of the livestock training, in December 2013 we used straight molasses as an energy supplement and enticement for this group to eat L. nepetifolia. The above mixed group of livestock was put on a small three-acre paddock. Each afternoon for three days they were offered a supplement of L. nepetifolia drizzled with approximately a pint of molasses (Figure 9). Livestock ate most of the plants offered in the trough, and initial photo monitoring and inspection showed they ate it in the pasture without any molasses added.

The molasses approach was also used to train cattle to increase intake of D. insularis. This grass is usually left untouched by cattle in a rotational grazing system. Four hundred cows and heifers were let into a three-acre paddock heavily infested with D. insularis that had been partially sprayed with a diluted molasses mixture. After 12 hours the cattle were moved out with significant use on this grass in both sprayed and unsprayed areas (Figure 10). Increasing stock density while offering an energy supplement appears to increase intake of D. insularis in cattle. This method also increased intake of another group of 100 head in a 20-acre paddock with small patches of this grass. Cattle ate patches of both sprayed and unsprayed D. insularis where previously none would be eaten. The increased intake of unsprayed D. insularis may result from simply overcoming neophobia or this in combination with increased energy status to deal with potential plant secondary compounds.

As all other conditioned preference training sessions besides A. phsyocarpa showed marked increased intake, additional supplementation was not deemed necessary. A. physocarpa was abandoned as a management priority by ranch managers at this time as goats will eat the fruits and other weeds showed more promise for this method. A. physocarpa may be identified in future projects for increasing intake with energy or toxin attenuating supplements.

Objective 3: Targeted weed populations will decrease with increasing grazing pressure from trained livestock.

Kapapala Ranch – Figures 11, 12, and 13 show very little change in P. guajava and M. faya densities at the Kapapala sites. Both sites were only browsed twice during the study period owing to changes in grazing management due to unprecedented drought. Stocking density was approximately 500 adult goats on two-acres. While only grazed twice in three years, individual trees showed significant damage to bark and branches within reach of goats for duration of project period. Several M. faya bushes showed signs of limited or light browsing. Goats were moved to new paddocks every other day, so prolonging the grazing period at this density may lead to heavier use of M. faya (Figure 14). Dead individuals were also counted but, as numbers were trace, they are not reported here.

Rocking W Ranch – Ranch managers selected S. trilobata for control with conditioned preference training of cattle. S. trilobata, a strongly stoloniferous and rhizomatous forb, can form dense stands between grass tufts after establishment suppressing desirable forages. While Figure 15 shows low relative cover of S. trilobata according to line point data, the weed was present in much of the interspaces. Ranch managers were more concerned about its increase rather than its current extent (Figure 16).

Jurassic Kahili Ranch – Ranch managers identified S. condensatum to be controlled with cattle and E. mollis to be controlled with sheep. S. condensatum is a perennial bunchgrass of very poor forage quality and palatability to cattle and sheep that can form monotypic stands. E. mollis tends to form dense patches in sod-forming forage types. Like S. trilobata sites at Rocking W, the monitoring sites were chosen in areas with currently low cover to note any increase. Figure 17 shows both species’ cover remained low during the study period.

Tom’s Goats – Photo-monitoring plots were set up at the initial training site in December 2011. However, owing to shifting ranch priorities and difficulties with the land lease, the trained group of goats were moved and dispersed among other herds. Furthermore, ranch ownership changed in 2013, ultimately resulting in herd liquidation. Therefore, the PI could not follow long-term impacts of this conditioned training program at this ranch.

Objective 4: Effects of consuming weeds on meat quality will be detectable.

Several factors combined to make pursuit of this objective untenable. Chiefly, periodic in-field observations of trained animals by ranch managers and the PI as well as the vegetation monitoring plots showed only incidental intake of target weeds over the project period. While increasing stocking density increased weed intake, these were short and infrequent periods over the life of the animal. Therefore, the lack of treatment effect could not justify resources spent on meat chemical and flavor analysis. Furthermore, slaughterhouse practices made identification of carcasses from trained animals uncertain. In some cases, heifers originally intended for market were retained in the breeding herd.

Meat analysis and travel costs were the largest budget items for this project. The elimination of meat analysis as an objective, a less frequent vegetation monitoring protocol, and lodging provided by ranches resulted in significant cost savings to the project unanticipated at project start. These cost savings resulted in the project coming in well under budget without a loss of data in three of the four main objectives.

Participation Summary

Research Outcomes

No research outcomes

Education and Outreach

Participation Summary:

Education and outreach methods and analyses:

Professional and Industry Group Meetings

The PI presented a 15-minute talk on the project background and initial findings to an audience of 15 ranchers at the Maui Cattlemen’s Association annual meeting on December 15, 2012. A ranch manager from Haleakala Ranch was present and led much of the question and answer session following the presentation.

On September 26, 2013, the PI gave a 20-minute talk on the concept, training method, and preliminary results of the project to an Extension professional workshop held at the Poamoho Experiment Station of the University of Hawaii with about 40 professionals from across the state in attendance. Owing to different groups going to multiple field visit sites immediately after the talk, most participants left before any formal evaluation data could be collected. However, the questions and comments during the talk indicated interest in promoting this approach for weed control when meeting with clients.

On November 8, 2013, the PI presented a 45-minute talk accompanied with slides to the Hawaii Sheep and Goat Association annual meeting attended by 28 people. Post-presentation evaluation data using the SARE form were collected via online survey and are summarized in Figure 18.

From November 15 – 16, 2013, the PI and one collaborating rancher were present at the College of Tropical Agriculture and Human Resources (CTAHR) booth with a looped computer presentation on the project. In this informal setting, it was difficult to collect formal evaluations, but of the approximately 200 attendees, many visited the booth and asked questions about the project.

On May 3, 2014, the PI presented a 45-minute talk on the project background and results to date as part of a Goat Basics Workshop held at the CTAHR Kauai Agricultural Research and Extension Center reaching 21 small-scale producers.

While outside the project period, an abstract detailing project background and results has been accepted for a 20-minute oral presentation at the Society for Range Management Annual Meeting to be held February 2015 in California. The PI will be traveling with other project funds to make this and another presentation thus incurring no cost to the current project which has already terminated.


An introductory article as well as a brief follow-up article was written for the Kauai Livestock News newsletter circulated among 220+ e-mail recipients in the Fall/Winter 2011 and Spring/Summer 2012 editions respectively.

A draft of an Extension Report to be published by the University of Hawaii CTAHR is currently being finalized for collaborator and peer review to make a permanent reference of project findings. The report acknowledges USDA Western SARE as the funding authority and will be available online in perpetuity for no charge.

A brief video is also in development by CTAHR faculty using footage shot for the project to present project background, methodology, and results. Once completed, this video will be available at no charge through the Kauai Livestock Extension YouTube channel. A limited number, approximately 100, hard copy DVDs will be made with materials purchased during the project period. There are no other publication costs to Western SARE for this video and above Extension Report as these are University of Hawaii publications done on faculty time, and due acknowledgement of USDA Western SARE as the funding agency is given.

Education and Outreach Outcomes

Recommendations for education and outreach:

Areas needing additional study

The methods we employed in this study convincingly showed an ability to overcome livestock neophobia for several plant species. Cattle, sheep, and goats of various classes were willing to eat target species to at least some degree in test pens or paddocks. The next step for improving the application of this method is determining the relationship between stocking density and weed impacts. Once animals are trained to eat weeds to at least a small degree, how long can a manager leave animals on a paddock to achieve significant weed defoliation without overuse of desirable forages? Does the return period based on desirable forages allow for weed recovery as well? Will repeated intense grazing periods followed by long rest reduce weed populations without overusing key forages? Lastly, will long-term exposure to these weeds lead to chronic health problems such as liver lesions or reproductive issues? These types of grazing management questions need to be addressed in future studies.

While part of the original design, an evaluation of potential impacts on product quality should be a study in its own right. This should be achieved by a combined experimental and observational approach. By experiment, a known amount of weeds and/or specific plant secondary compounds could be fed at various concentrations to livestock and would reveal with greater confidence the degree of feed effects on meat flavor, if any. An observational study could evaluate the flavor of meat from animals known to eat a large amount of weeds, such as those used in intense high stocking density targeted grazing, compared to meat from livestock coming off of similar or even same paddocks if they did not receive training. In the present study, we did not anticipate such an apparently low intake of weeds despite positive results in the training pens. In other words, while weed consumption was apparent in the paddocks following training, given the amount of alternative feeds available owing to low stocking density, it was difficult to determine any treatment effect. Along these lines, should off-flavors be detected in meat from animals eating weeds, 1) Can this off flavor be attenuated by grazing weed-free pasture, and 2) If so, how long do they need to be on weed-free pasture before slaughter?

Finally, while we assumed that replacing herbicide application with grazing management via behavior modification would reduce costs, a formal economic analysis should quantify or negate the validity of this assumption. Several important parameters should be set in this analysis to look beyond herbicide replacement alone. For example, a test paddock containing roughly equal amounts of a particular weed could be split in half. One side would receive grazing management only by animals trained to eat more of this weed. The other would receive herbicide treatment and grazed by untrained animals. Labor and material costs would be recorded for both up until slaughter. Economic considerations beyond herbicide and labor would be average daily gain and carrying capacity. In other words, herbicide cost savings could be lost by slower rates of gain from potential anti-nutritional qualities of grazed weeds. Or improved forage quality or carrying capacity might justify herbicide application at current costs. Considering these factors in their totality would lead to a more informed outlook on this approach’s economic feasibility.

Any opinions, findings, conclusions, or recommendations expressed in this publication are those of the author(s) and do not necessarily reflect the view of the U.S. Department of Agriculture or SARE.