Download PDF version here
Project Title:  Detection Of Ipomoea Species In Soybean For Harvest Aid Decisions
Principal Investigators:  Daniel H. Poston and Clifford (Trey) Koger

Technical Summary: 

Annual morningglories are consistently ranked as some of the most troublesome weeds in Mississippi soybean fields. Many morningglory plants survive herbicide application or emerge following herbicide application and create problems at harvest. Harvest aids are available to desiccate morningglory vines, but economic thresholds for morningglory vine densities have not been established relative to harvest aids. 

Field studies will be conducted to determine the level of morningglory infestation required in soybean at harvest to economically justify a harvest aid and to determine if critical levels of morningglory infestation can be remotely detected with multispectral imagery. Studies will be conducted in a weed-free location. Six levels of morningglory infestation will be established by transplanting cotyledon-size morningglory seedlings into plots shortly after final postemergence herbicide applications made at the V4 soybean growth stage. This will simulate morningglory plants that emerge as problems late in the growing season and that cause problems at harvest. Pitted morningglory biomass will be determined from a designated area within each plot prior to harvest aid application and immediately after remotely sensed images have been acquired. Percent canopy cover occupied by pitted morningglory and soybean will also be visually estimated for each plot. Remote sensing will be incorporated into the study to determine if multispectral imagery can be used to assess pitted morningglory densities and canopy composition present in soybean fields near harvest and in turn be used to determined if harvest aide applications are economically feasible. Images will be acquired when soybean plants are in the R7 growth stage (beginning maturity, plants senescing and dropping leaves) and morningglory vines are still green. Efforts will be made to correlate actual plant densities and canopy cover estimates collected on the ground with remotely derived data. Main plots will contain a specified morningglory density. Main plots will be subdivided into subplots that receive a harvest aid and those that do not. Data from this study should provide critical information that can be used by producers to determine when harvest aids are economically feasible and if remotely sensed imagery can be used to detect these levels of weed infestations. 


Objective 1. Establish economic thresholds for harvest aids relative to morningglory infestation levels present in soybean fields at harvest.  

Objective 2. Determine if multispectral imagery can be used to assess levels of morningglory infestations present in soybean fields at harvest. 


Studies will be initiated in 2002 at the Delta Research and Extension Center in Stoneville, MS. A location will be selected that has been maintained weed free for several years. Metolachlor will be applied preemergence to remove annual grasses from the test area and postemergence graminicides will be used if necessary to remove grasses that escape soil applied herbicides. Roundup Ready MG III or IV (depending on planting date) soybeans will be planted using 15-inch row spacings. Plots will be tilled prior to planting to remove existing vegetation. Glyphosate applications will be made at V2 and V4 soybean growth stages, but may vary depending on weed emergence patterns and soybean development. Following the final glyphosate application, cotyledon-stage morningglory plants will be transplanted into row middles to simulate morningglory plants that emerge following glyphosate applications and become a problem at harvest. Plant densities of 0, 0.5, 1, 2, 4, and 8 plants m-2 will be established. Plants that do not survive will be replaced with new ones. Other broadleaf weeds that emerge in plots will be removed by hand weeding or hoeing. Plots will be 30 x 40 feet arranged in a randomized complete block design with 4 replications and a split plot arrangement of treatments. Main plots will contain a specified morningglory density. Main plots will be sub-divided into subplots that receive a harvest aid and those that do not. Above ground pitted morningglory biomass will be determined by harvesting vines from one square meter of each plot just prior to harvest aid application, image collection, and soybean harvest. Percent canopy composition occupied by pitted morningglory and soybean will be visually estimated for each plot during the same time frame. Harvest aids will be applied to appropriate plots following collection of vine and canopy composition data and multispectral images. Soybean will be harvested 7 to 10 days after harvest aid applications. Seed samples will be screened and percent foreign matter and soybean yield determined.  

Image collection will be coordinated through the RSTC using a Geovantage multispectral system. In addition, several other vendors are available as backups if flights scheduled flights by RSTC are interrupted or delayed for various reasons. GPS coordinates for the corners of the test area will be provided to RSTC personnel to ensure the collection of images for the study area. Plots are located in established RSTC flight lines. Multispectral imagery will be collected immediately prior to biomass collection and in the same timeframe as visual ratings. This imagery will correspond with R7 to R8 soybean growth stages when soybeans are senescing and have lost most or all of their green foliage. MG III soybeans planted in early-May will mature in late-August to early-September depending on environmental conditions. Multispectral flights coordinated by the RSTC are scheduled every 2 weeks for Stoneville through September and possibly into November. This should allow ample opportunities to collect imagery of plots during the desired time frame. Spatial resolution will be 0.5m. Therefore, plot sizes of 30 x 40 feet should be more than adequate for this study. For each image, reflectance data will be collected in the visible blue, green, red, and near-infrared (NIR) spectrums. Imagery will be geo-referenced at time of data acquisition. This will allow images to be processed and received within approximately 24 hours of time of data acquisition. Reflectance data for each spectral band will be extracted from a 4.0- by 4.0-m sampling area for each plot center. To account for 1.0-m accuracy and minute errors associated with image geo-referencing, reflectance data for each 1.0-m2 pixel within each sampling area for each spectral band will be averaged. Reflectance data for the red and NIR spectral bands will be used to derive a Normalized Difference Vegetation Index (NDVI) for each plot. The spectral band and NDVI data will be as classification features for discriminating pitted morningglory infested and weed-free soybean plots.   

Potential technology products from this project may include the development of software packages that use data from images to estimate weed densities and generate prescription spray maps. Any intellectual property issues will be handled according to Mississippi State University intellectual property guidelines.   

Justification/Literature Review:   

Pitted morningglory (Ipomoea lacunosa L.) and entireleaf morningglory (Ipomoea hederacea var. integriuscula L.) are the 2nd and 3rd most troublesome weeds in Mississippi soybeans, respectively (Byrd, 2001). Morningglories often emerge as a problem as early-maturing (maturity groups III and IV) soybeans are nearing harvest thereby reducing yields and harvest efficiency. Harvest aids applied 7 to 14 days prior to harvest can be used to desiccate vines and improve harvestability, but threshold levels for morningglory infestation have not been established relative to harvest aids. Based on research conducted in 2000 and 2001, harvest aids were generally economical when morningglory control was 70% or less (Poston et al. 2001) With actively growing morningglories in soon to be harvested fields, moisture can be elevated causing loss in seed quality and harvesting speed (Shaw, 1998).  Bennet and Shaw (2000) found that 0.25 lb ai/A paraquat + 3 lb ai/A sodium chlorate applied as a pre-harvest desiccant provided the most economical and efficacious control of morningglory species and other weeds. Unfortunately, no threshold levels for weed infestation at harvest have been established. Therefore, efficacious harvest aids are available but producers have no method of assessing when weed infestations are at high enough levels to make harvest aids profitable. Additionally, farm size has increased in recent years as profit margins for farm commodities have narrowed and scouting larger land areas manually is likely to be cost prohibitive.  

Interest has been expressed in using remote sensing techniques to detect weed infestations in fields and developing customized spray maps from the data (Christensen et al. 1999, Thornton et al. 1990). Several problems have been encountered with attempts to discriminate weeds and crops early in the growing season. Medlin et al. (2000) were able to successfully detect pitted morningglory and sicklepod early in the growing season with 90% accuracy provided weeds were 5 to 10 cm tall and at populations of 10 plants m-2. Medlin et al. (2000) also noted that reflectance of background soil and vegetation interfered with the ability to discriminate between weed-free and weed-infested plots.  

Multispectral imagery has proven more effective for detecting late-season weed infestations in soybean. Multispectral imagery has been used with at least 90% accuracy to discriminate weed-free soybean from soybean infested with barnyardgrass, browntop millet, and large crabgrass after soybean canopy closure when soybean was in the vegetative to late-senescence growth stages (Koger et al. 2001). Richardson et al. (1985) as well as Menges et al. (1985) used multispectral remote sensing to differentiate weed-free cotton from various monocot and dicot weed species late in the growing season. Multispectral imaging is also useful for discriminating weed infestations using discriminant functions developed from other images. Koger et al. (2001) used multispectral imagery to detect late-season weed infestations for one image with 81 to 90% accuracy using discriminant function developed for another image. Based on this information, multispectral imagery may have the potential for detecting late-season pitted morningglory infestations at or below levels that require pre-harvest desiccation and in turn help determine when harvest aid applications are cost effective. In addition, detection of late-season weed infestations can provide growers useful information as to where weed problems occur in fields thereby assisting in the development of future herbicide programs.   

Late in the growing season interference from background soil should also be reduced because of crop canopy, especially in narrow-row soybeans. Additionally, green weed should be easily distinguishable from yellowish brown senescing soybeans.  

Literature Cited: 

Bennet, A.C. and D.R. Shaw. 2000. Effect of preharvest desiccants on weed seed
production and viability. Weed Technol. 14:530-538. 

Byrd, J. 2001. The Southern States Most Common and Troublesome weeds in Soybean. Proc. South. Weed Sci. Soc 54:253. 

Christensen, S., A. M. Walter, and T. Heisel. 1999. The patch treatment of weeds in
cereals. Pages 591-600 in Proc. Brighton Crop Protection Conf. Weeds. Farnham, 
UK: British Crop Protection Council. 

Koger, C.H., D.R. Shaw, C.E. Watson, and K.N. Reddy. 2001. Potential of remote
sensing for detecting late-season weed infestations in soybean (Glycine max).
Weed Sci. (In Review).  

Medlin, C. R., D. R. Shaw, P. D. Gerard, and F. E. Lemastus. 2000. Using remote sensing to detect weed infestations in Glycine max

Menges, R.M., P.R. Nixon, and A.J. Richardson. 1985. Light reflectance and remote
sensing of weeds in agronomic and horticulture crops. Weed Sci. 33:569-581. 

Poston, D. H., D. R. Shaw, C. Smith, and R. M. Griffin. 2001. Weed control alternatives for maturity group III soybeans in Mississippi. Proc. South. Weed Sci. Soc. 54:48. 

Richardson, A.J., R.M. Menges, and P.R. Nixon. 1985. Distinguishing weed from crop
plants using video remote sensing. Photogramm. Eng. Remote Sens. 51:1785-1790. 

 Shaw, D. R. 1998. In Getting the green out of soybean harvest. E. A. Dorris. Mississippi
Farmer. August 1998. 

Thornton, p. K., R. H. Fawcett, J. B. Dent, and T. J. Perkins. 1990. Spatial weed
distribution and economic thresholds for weed control. Crop Prot. 9:337-342.



Back to Top

Back to Menu

Visit: MSU ||  || MAFES Intranet || Remote Sensing Technologies Center (RSTC)
For information about this web site, contact Wade Givens
Last Modified: 01/06/2004