Scouting for Weeds

Search PMN

Scouting for Weeds

Sharon Clay, South Dakota State University, Brookings, Gregg Johnson, University of Minnesota, Southern Research and Outreach Center, Waseca

Corresponding author: Sharon Clay. sharon_clay@sdstate.edu

Introduction

In the past, weed information has been collected in a casual way with little attention given to weed species, densities, or distributions. This is due to time and labor constraints for rigorous scouting, the complexity of the scouting information when sampled rigorously, the assumption that weeds are constant and uniform throughout a field, and the lack of equipment to easily manage weed variability even if it was noticed. Generally, a single recommendation (whether a single herbicide, formulation mix, tank-mix combination, or preemergence/postemergence split) for weed management was given for a field. The basis for the recommendation could have been based on past years' weed problems, scouting field edges, or driving a W or Z pattern across the field in the spring or fall. These approaches have been successful in managing weeds and improving profits. The questions are: (1) Do weeds vary enough in a field to manage them with precision techniques using different methods or herbicides? (2) Can we take advantage of technology (GPS, variable rate sprayers, direct injection controllers, etc.) to further improve weed management and profitability? This guide discusses different approaches and concepts for obtaining information about weed diversity in the field.

In most cases, weeds are highly aggregated in a field (3,4,8). For example in a 120-acre field, Canada thistle (Cirsium arvense) and foxtails (Setaria sp.) [mixtures of green and yellow foxtail (S. virdis and S. pumila, respectively)] were not uniformly distributed and occurred in different areas of the field (Fig. 1). Drainage, topography, soil type, microclimate, and other factors play important roles in where weeds will be located and how successful and competitive they will be at a specific site (11). The first step in developing effective site-specific weed management strategies is to obtain accurate and reliable data on the location of weed species and densities. The next step is to match the weed management solution to the problem in a site-specific manner. Spray equipment has been developed that allows for different chemical treatments and rates to be targeted to weed infested areas of the field (6). Using information about weed distribution and variability, equipment that matches the correct chemical with the weed(s) present has been shown to result in better weed control, lower herbicide costs, and increased net return (2,5,9).

With these concepts in mind, you can then formulate your own, detailed, step-by-step plan that works best for your scouting needs. Clearly, the most effective data collection strategy depends on how the data is going to be used and in what time frame.

Fig. 1 Examples of weed populations and locations in a 120-acre field. Scouting was done on an 100-×-100-ft grid and data were kriged to obtain estimated densities between sampling points. Note the differences between (A) Canada thistle (Cirsium arvense) distribution and densities (0 to 12 ft²) and (B) distribution and densities (0 to 50 ft²) of grass species [primarily green and yellow foxtail (Setaria viridis and S. pumila)].

Sampling Scale

It is impractical to determine the level of weed infestation on every square foot in a field. Therefore, cost-effective strategies are needed that provide fairly accurate and precise information at a manageable scale. This requires a great deal of thought about the sampling scale at which information is collected. Generally, the more data that are available and the more intensive the sampling method, the better the management decision should be. Correct interpretation of the data and the recognition of the errors, omissions, and pitfalls of different sampling methods are needed also. In order to optimize your time, both in scouting and application, you will want to consider equipment size, speed, and time it takes to change an operation on-the-go.

Equipment size. If the goal is site-specific herbicide application, then the width of the spray boom is a relevant sampling scale. For example, if the boom is 40 ft wide, then you may consider a 40-×-40-ft grid spacing. Grid sampling a 160-acre field at this scale results in about 4400 individual data points. Sampling at this density would be time- and cost-prohibitive, may be a data management nightmare, and the conversion of the information into useful decisions may be difficult. How can we reduce the amount of time involved to sample and process data and still get useful site specific information?

Speed and changing operations. In the above example, sprayer speed and the time it takes to change chemicals on-the-go can be included in the criteria for determining minimum grid size. If 10 seconds are needed to switch herbicides coming from the boom, at 5 mph ( 7.3 ft/sec) you travel 73 ft during the transition between herbicides. So perhaps a 40-×-100-ft grid would be a more realistic grid size. Now in the 160-acre field, 1750 samples would be taken. This is a 60% reduction in the number of samples taken in the previous example but this data set may still be difficult to work with. If you are traveling faster still, the results would be a larger grid size and still fewer samples.

Sampling on a larger grid and using statistical techniques to interpolate values to the size that fits your management area is one solution. However, as grid spacing increases, weed patches or infestations may be completely missed because the area of the patch or infestation is less than the grid spacing used (Fig 2). In addition, interpolation methods, such as kriging (4), require sampling be conducted on the proper scale. If not, overall weed distributions, patterns, and locations may be highly inaccurate. There is obviously a trade-off between obtaining accurate information and doing so in a cost-effective manner.

Fig. 2. Examples of weed patches that would not be sampled if a regular grid sampling technique is used.

Sampling Schemes

Scouts must rely on their own judgment in designing a sampling plan that meets their objectives with respect to accuracy, time, and labor constraints. The following are suggested methods that can be adopted and adapted to your operation.

Quick Assessment

Method:	Random assessment. Use an ATV to quickly identify the location of weed patches by driving in a grid pattern or pseudo grid pattern across the field, stopping only where weeds are present. Once a patch is identified, the area of the patch can be determined by driving around the perimeter of the patch. It is also not a bad idea to record the visual severity of each patch based on a predetermined set of criteria. For example, a patch can be recorded as severe, moderate, or light based on a quick density estimate or some other measure of severity. It is best if a GPS can be used to provide accurate position information, however, a rough estimate of location (either by counting rows or by using a measuring device) will do.
	Alternative assessment. This can be accomplished by visually splitting a large field into 4, 8, 16, etc., smaller fields or splitting the field into areas with similar topography (drainage area, hilltop, side slope etc). For each area, a herbicide recommendation is developed using techniques like those described above.
Outcome:	Data obtained from this procedure can be used to make a map showing the location of individual weed patches as well as relative severity. Some Geographic Information System (GIS) software will also allow you to determine percent field area occupied by all weed patches or individual weed patches. This information is important for monitoring patch size over time or deciding if a patch is large enough to warrant control.
Advantages:	Very quick and easy to implement. Works best for perennial weeds.
Disadvantages:	Annual weed patches may be difficult to identify, especially if sampling is done when weeds are in the seedling growth stage or are less than 2 inches in height. Information is limited to patch location but can include an estimate of visual severity. When using a visual severity assessment, you must be very clear as to what you mean when you classify as population as severe vs. moderately severe vs. light. Depending on this ambiguity, the value of information can be very limited. Visual estimates of patch severity can be misleading if care is not taken to standardize the estimates.
Regular (Uniform) Grid Sampling
Method:	Data is collected on a uniformly spaced grid coordinate system. At each grid node, weed density or presence/absence data is recorded in a quadrat. There are a number of ways to record weed density.
	Quantitative. One way is to record the actual weed density for each species. This method can be time consuming when weed density is high but does provide valuable information. Weed seedlings and small weeds may be difficult to identify and quantify.
	Semiquantitative by species. The second method involves counting weeds up to a certain number (e.g., 50) for each species. The idea here is that the value of knowing there is 100 verses 50 weeds per ft² is not relevant from a management threshold standpoint. As with Quick Assessment scheme (above), weed seedlings and small weed identification may be problematic.
	Semiquantitative by groups of species. You can also choose to count weeds by species groups (e.g., broadleaves vs. grasses, or small seeded broadleaves vs. large seeded broadleaves vs. grass weeds). This method is less time-consuming than the other methods, especially in heavily infested areas. However, this method does not provide detailed information on individual weed species and may not provide enough information for interpolation methods to predict weed infestations between sample points.
	Qualitative. The least time-consuming method is to simply assess the presence or absence of a given weed or weed complex at the given sampling points. The information obtained in presence/absence sampling provides a quick assessment of weed species diversity but does not provide information related to the severity of weed infestations. Identifying weed species is unimportant in this sampling scheme. Yes/no answers can be used to build probability maps of distributions in the field (7).
Outcome:	Interpolation techniques may be used to build a map showing the location, size, density, and probability of weed infestations or occurrence across the field. If a GIS data-handling technique is used, other information collected at the time of sampling (e.g., stand counts or disease presence) can be integrated onto this map.
Advantages:	Easy to implement and understand. Does not require any prior information about weed populations in the field and can therefore be done by less-experienced people.
Disadvantages:	Optimal grid spacing is unknown and depends on several factors. One can decide on a grid spacing that maximizes cost and labor efficiency (e.g., there may only be time to sample 100 points across a field). This has shown to be an effective method for soil sampling, where soil properties change gradually across a field and usually occur in larger blocks. Because weeds are highly aggregated, this method may result in entire weed patches being missed because the size of the weed patch is smaller than the grid spacing. One alternative would be to use a predefined grid system but allow the scout to deviate from this grid if weeds are noted but would otherwise be missed. The allowable degree of deviation off the grid is dependent on time and cost considerations.
Sampling at Harvest
Method:	Weed mapping at harvest is an easy method to generate maps of weed infestations in a field. Simple devices are connected to the GPS signal on the yield monitor to record weeds as a field is harvested. You simply push buttons on the device when you enter and exit a weed patch. If you have different “flags” for different weed species or types, you can determine the location and the weed problem.
Outcome:	A map is produced that is matched to the yield map.
Advantages:	This is a simple, easy method that is done with another operation so it does not require any extra time. The map can be used to see how well weed management operations worked for the past season and will give an indication of trouble spots for the next year. Scientists from Spain reported that defining patches of sterile wild oat from a combine during harvest was the cheapest method and fairly reliable for making weed maps for the following season (1). If the weeds are perennial, you may be able to treat problem areas after harvest.
Disadvantages:	This is an “after the fact” map and weeds have already caused harvest losses. This method requires close attention sustained over a long period, and moments of distraction can result in erroneous maps. Too many flags for different species can be confusing. One flag for annual broadleaf weeds, one flag for grasses, and one flag for perennials may be the easiest sampling scheme. New infestations of weeds usually occur at field edges and in end rows. Mapping these areas with care is recommended.
The above methods do not consider prior knowledge about the weed populations in the field in determining the design of sample number or location (grid spacing). Experience should not be overlooked, sin it has a lot to do with how accurately a weed population is characterized. Most farmers and scouts know what weed species are present and where they are located in the field. In fact, weeds tend to occur in the same general location over time (8,10), and it is possible to direct sampling efforts to areas where weeds are most likely to occur. Directed sampling allows you to accurately characterize weed populations in much less time because you are not sampling the entire field. Given this fact, it seems natural to use this information in designing an adaptive sampling strategy that stresses flexibility and recognizes experience. Adaptive Sampling Using Aerial Imagery In adaptive sampling, a grid system is used initially and then modified based on supporting information over time. This supporting information may come from a variety of sources and experiences. An example of supporting data is aerial imagery.
Method:	Multispectral remote sensing offers the potential to identify, categorize, and determine differences and similarities in crop and field conditions over a wide geographic area. Images can be taken several times during a season and over multiple years to give views of whole fields, farms, and watersheds.
Outcome:	Spatial variables relate to both pixel size and minimum object/feature size. Understanding what pixel resolution means will help you determine what would be useful for your scouting program. A pixel resolution of 1 m will integrate reflectance to give one value for an area about the size of a tabletop. A pixel resolution of 30 m integrates the reflectance in the area of six school buses parked together. Large, dense weed patches may show up in the 30-m resolution images, whereas the finer resolution would give more detailed information.
Timing of the flight also needs to be considered. Images taken just prior to post-emergence applications using a combination of visible and near-infrared bands can give information on the location of weed patches or other field anomalies. Images taken in the fall after a light frost along with adaptive scouting techniques are also an excellent source of weed information, since perennial weed patches may still be green even after the crop has senesced (Fig 3). Overlaying the visible and near-IR georegistered bands gives an enhanced false color view of the field which can also be used to determine areas related to other stress factors such as water, insects, diseases, or fertility problems.

Fig. 3. Aerial image of a 160-acre soybean field taken in early October with 1-m pixel resolution. Canada thistle patches are evident across the entire field as round, dark spots. Low areas of the field had high densities of common ragweed. Additional scouting using this image as a field map revealed the areas of quackgrass and annual grasses that had not been controlled. Since this image is georeferenced, it can be used as a spray guide for perennial weed control in the fall and, in the spring, as a scouting tool.

Advantages:

Aerial imagery allows you to modify a sampling strategy by having access to weed information on a field scale in a timely fashion.

Disadvantages:

This method often requires post-processing which means that images need to be analyzed by trained professionals. Therefore, there may be considerable time between when the aerial imagery data was collected and when it becomes available to the farmer or crop consultant. Timing is obviously critical for post-emergence herbicide application decisions. You must work closely with the company providing the aerial images to make sure it is delivered in a timely manner.

Summary

The idea of adaptive sampling is that you need not, and should not, be tied to a given sampling strategy, e.g., grid sampling with a 100-ft spacing. In fact, one should not consider grid sampling beyond the first 1 or 2 years. Instead, you should consider using flexible sampling systems that build on information and experience. Adaptive sampling requires access to information that has been collected over several seasons. Because the purpose of sampling is to provide accurate and timely information needed to make intelligent, cost effective decisions, weed management systems must be dynamic and flexible to meet changing demands. Using an adaptive approach to scouting for weeds naturally results in a more dynamic data gathering system that can be used to determine if the current weed management system is or is not meeting your goals. Often, a given weed species is increasing in density and area while others are decreasing. Being able to adjust sampling strategies based on these observations is critical and must be taken into account each year. Moreover, sampling is a key component in the design of effective weed management strategies that help manage risk by providing information needed to optimize the correct timing of herbicides and accurately monitor weed management successes and failures (12). Until more research is completed on how best to sample for weed spatial distribution, experience coupled with flexibility will be the key to obtaining reliable data needed to make informed site-specific weed management decisions. However, there is no single sampling strategy that is best in all situations and each sampling strategy has advantages and disadvantages.

Literature Cited

1. Barroso, J., Ruiz, D., Fernandez-Quintanillay, C., Rigueiro, A., Diaz, B. 2001. Comparison of various sampling methodologies for site-specific sterile wild oat (Avena sterilis spp. Ludoviciana Dur. (Nyman)) management. Conference Proceedings, 22 November, 2001, Leon, Spain. Spanish Weed Society, Madrid, Spain.

2. Biller, R. H., Hollstein, A., Sommer, C., Stafford, J. V. 1997. Precision application of herbicides by use of optoelectronic sensors. Pages 451-458 in: Precision Agriculture ’97. Volume 2. Technology, IT and Management. First European Conference on Precision Agriculture, Warwick University, U.K.

3. Cardina, J., Sparrow, D., and McCoy, E. L. 1996. Spatial relationships between seedbank and seedling populations of common lambsquarters (Chenopodium album) and annual grasses. Weed Sci. 44:298-308.

4. Clay, S. A., Lems, G. J., Clay, D. E., Forcella, F., Ellsbury, M. M., and Carlson, C. G. 1999. Sampling weed spatial variability on a field-wide scale. Weed Sci. 47:674-681.

5. Garibay, S. V., Richner, W., Stamp, P., Nakamoto, T., Yamagishi, J., Abivardi, C., and Edwards, P. J. 2001. Extent and implications of weed spatial variability in arable crop fields. Plant Prod. Sci. 4:259-269.

6. Humburg, D. 1999. Variable rate equipment: technology for weed control. Site-Specific Management Guideline #7. Online. Potash & Phosphate Institute Potash and Phosphate Institute of Canada.

7. Johnson, G. A., Mortensen, D. A., and Martin, A. 1995. A simulation of herbicide use based on weed spatial distribution. Weed Res. 35:197-205.

8. Johnson, G. A., Mortensen, D. A., Young, L. J., and Martin, A. 1995. The stability of weed seedling populations and parameters in Eastern Nebraska corn (Zea mays) and soybean (Glycine max) fields. Weed Sci. 43:604-611.

9. Lindquist, J. L., Dieleman, J. A., Mortensen, D. A., Johnson, G. A., and Wyse, D. Y. 1998. Economic importance of managing spatially heterogeneous weed populations. Weed Tech. 12:7-13.

10. Medlin, C. R., Shaw, D. R., Cox, M. S., Gerard, P. D., Abshire, M. J., and Wardlaw, M. C., III. 2001. Using soil parameters to predict weed infestations in soybean. Weed Sci. 49:367-374.

11. Radosevich, S., Holt, J., and Ghersa, C. 1997. Principles of weed ecology. Pages 43-65 in: Weed Ecology: Implications for Weed Management. John Wiley & Sons, New York.

12. Wallace, A. 1994. High-precision agriculture is an excellent tool for conservation of natural resources. Commun. Soil Sci. Plant Anal. 25:45-49.