Progress 09/15/00 to 09/14/05
Outputs
A 4-year study was initiated in the fall of 2004 with the following objectives as goals: to evaluate methods for determining nutrient management zones;to combine information from different zone sampling methods to improve delineation; and to determine methods to automate zone boundaries through the development of computer-driven algorithms. Five field sites were located in North Dakota, 1 site (2 field locations on the same farm) in Montana and 2 sites in Minnesota. The sites were diverse in cropping systems. The site in Montana was a wheat-fallow rotation. The site in Williston was in continuous wheat. The site in Mandan was in a spring wheat, winter wheat, sunflower rotation, the site in Valley City was in a spring wheat, barley, sunflower rotation, Oakes was in continuous corn. Crookston was in a grain and sugarbeet rotation. Renville was in a corn soybean sugarbeet rotation, and the Minot location was in canola and fallow/prevented planting. Despite differences in
climate and cropping, there were similarities in the delineation tools that were effective. Three different computer-driven algorithms were explored to develop zones, with supplemental methods for improving data layers collected using different tools. The three methods were clustering using the k-means procedure, clustering using a neural network procedure, and a weighted, classified method, with topography weighted more than other methods included in the analysis and five classes. Topography developed zone boundaries well relative to other methods at all locations. In the west, soil EC sensors developed better boundaries than in the east. Yield maps were useful sometimes individually, but use of a yield frequency map that combined different years using a normalization of data placed into three classes was even more useful. Satellite imagery was generally better than aerial photography at most locations. Most fields were divided into large subplots and variable rate N applications
were imposed based on three treatments; uniform based on a field composite analysis, variable based on the researchers favorite method, and variable based on a combination of methods, usually developed by one of the three automated methods. Results showed a general yield and/or quality advantage to the zone-based approach. Economic analysis of the results showed an advantage in sugarbeets to a zone-based approach, but not for grains. Ending nitrate levels were generally not related to treatment. The studies showed that topography, satellite imagery, yield frequency maps and soil EC sensors would be useful tools in delineating nutrient management zones. The data are easily classified or clustered using computer algorithms, which eliminates the tedium related to zone delineation in the past. Use of two or more delineation tools enhances the usefulness of the zones. Clustering methods are viable alternative methods, however, the weighted, classified method may be more intuitive for
ag-professionals to use and might be easier to explain its operation to farmers.
Impacts
The study showed that the tools needed to delineate N management zones are similar across the Northern Plains. Researchers across the region are now telling the same message and the number of growers and consultants commonly using a zone approach to N management has greatly increased since the beginning of the project.
Publications
- Gautam, Ramesh. 2005. Development of neural network to aid in nitrogen management. M.S. Thesis. North Dakota State University.
- Nanna, Tania. 2005. Management zone delineation in precision agriculture using a weighted, classified algorithm method. M.S. Thesis. North Dakota State University.
- Franzen, D., Nanna, T. Gautam, R., Casey, F., Derby, N., Staricka, J., Panigrahi, S., Long, D., Sims, A. and Lamb. J. 2005. Evaluation and Effectiveness of Nitrogen Zone Delineation Methods. In Annual Meetings Abstracts [CD-ROM]. ASA, CSSA, and SSSA, Madison, WI.
- Casey, F.X.M., Derby, N.E., Franzen, D.W. and Ralston, D.V.P. 2004. Lessons Learned from a Four-Year Precision Agriculture Study. In Annual Meetings Abstracts [CD-ROM]. ASA, CSSA, and SSSA, Madison, WI.
- Derby, N.E., Casey, F.X.M. and Franzen, D.W. 2004. Methods of Zone Delineation and Water Quality Impacts of Precision Agriculture. In Annual Meetings Abstracts [CD-ROM]. ASA, CSSA, and SSSA, Madison, WI.
- Gautam, R.K., and Panigrahi, S., 2004. Development and evaluation of nutrient prediction model using neural network architecture, Abstract submitted for the ASAE annual meeting to be held in Ottawa, Canada, Aug. 1-4, 2004.
- Franzen, D. and Nanna, T. 2004. A weighted classification system to delineate nitrogen management zones. In 2004 Agronomy Abstracts. ASA- CSSA-SSSA. Madison, WI
- Franzen, D.W. and Nanna, T. 2004. Delineating N management zones. In Proceedings of the 7th International Precision Agriculture Conference, July 26-28, 2004, Minneapolis, MN. Univ. of MN, St. Paul.
|
Progress 10/01/03 to 09/30/04
Outputs
September, 2004 was the official end of this USDA-IFAFS project.However, the project was given a one-year no-cost extension to summarize the project and provide a final report. New data was not generated during 2004 from Crookston, or Renville, MN locations. However, data was generated and collected from Montana (Dan Long), and all of the North Dakota locations. The Montana location was again in winter wheat and yield, quality, and N input and residual nitrate data have been received from that location for 2004. Variable rate application was applied using three treatments- uniform, variable based on landscape, and variable based on multiple layers. Spring wheat was grown at Valley City, ND. Variable rate broadcast urea application was conducted on 4/21 using three treatments- conventional uniform application based on a composite soil nitrate test; variable based on topography only; variable based on a weighted, classified zone with topography, yield mapping and
satellite imagery as data layers. The field received one inch of rainfall within three days of application. Spring wheat was seeded on 4/26. Wheat quality samples were obtained 8/24, and the field was harvested with a yield monitor 9/2. At Oakes (Francis Casey, researcher, Nate Derby, technician), variable-rate urea application was applied using three treatments- uniform based on a field composite nitrate test, variable based on landscape, and variable based on k-means clustering technique using satellite imagery (NDVI from Landsat 7), soil EC, and topography. Corn yield was obtained using a yield monitor. At Mandan (Vern Hofman, researcher), three fields received three treatments through a variable-rate air-seeder- uniform based on a composite soil nitrate test, variable based on topography, and variable-rate based on a neural-network clustering technique developed by Dr. Sarinjian Panigrahi and his graduate student Ramesh Gautam. The technique combined satellite imagery, topography,
yield map data and soil EC in an algorithm and tested against residual soil nitrate values. The fields were seeded to either winter wheat, spring wheat or sunflower. At Williston (Jim Staricka), a cold, wet spring delayed fertilizer application and seeding until early June. The field received three variable-rate ammonia treatments- uniform based on a composite soil nitrate test, variable based on topography, and variable based on the neural network technique of Dr. Panigrahi and Ramesh Gautam. Although heading of the wheat was delayed by the cool summer weather until early August, quality samples were obtained in September and the late, dry fall allowed an early October harvest using a yield monitor.
Impacts
The results of this study should make zone nutrient management easier and more effective in the region, resulting in more efficient use of N, and less environmental hazards for surface and ground water as the result of over-application of N to sensitive soils. Each of the methods, k-clustering, neural network clustering, yield frequency mapping, and weighted-classification, resulted in zones developed in a highly automated manner, compared to the old way of hand-examination and subjective line drawing used in the past.
Publications
- Lamb, J.A., Bredehoeft, and S.R. Roehl. 2004. Nitrogen management on a field scale. p. 110-115. In 2004 Sugarbeet Research and Extension Reports. Vol. 34. Sugarbeet Research and Education Board of Minnesota and North Dakota.
|
Progress 10/01/02 to 09/30/03
Outputs
During 2003, data was gathered from each of the eight project locations (Montana, Dan Long researcher; Minnesota, Crookston, Albert Sims, researcher; Minnesota, Renville, John Lamb, researcher; North Dakota, Valley City, Dave Franzen, researcher; Oakes, Francis Casey, researcher; Mandan, Vern Hofman, researcher; Minot, Mark Halvorson, researcher; Williston, James Staricka, researcher). In the fall of 2002 or spring of 2003, soil samples were collected on one-quarter to one-half acre grids. Sample depth was either a 2 feet or 4 feet. Residual nitrate termination serves as evaluation criteria for the past N applications, and also as a basis for determining the residual nitrate patterns. In 2003, six fields received variable-rate N treatments based on a favorite zone delineation method, while the second treatment was based on either a neural network analysis of the site (Montana, Williston, Mandan), or a classified method (Minnesota (2), Valley City, Oakes). Yields were
recorded in each field with a yield monitor, while sugarbeets were hand harvested to evaluate fertilizer treatments. Aerial photographs and satellite images were obtained of each field. Zone delineation techniques were evaluated by three graduate students under the guidance of Francis Casey, Dave Franzen and Suranjian Panigrahi. Frank Casey's group using data from the Oakes site is using a k-means clustering technique to delineate zones using soil EC, topography and imagery. A classified technique is also being evaluated. Another technique being evaluated is dividing the yield map into small grids, applying a mean yield to the grid, and then classifying the grid as above the mean, around the mean or below the mean. This technique appears to add meaning to yield maps. This method is being automated and there are plans to have software available on a web page within the next few months. Dr. Panigrahi's student, Ramesh Gautam, is applying data from each site to develop zones using one of
several neural network techniques. Within a site, the application of the neural network is quite descriptive, with very high ( r2 > 90%). The next step will be to use the algorithm between sites to validate its general use across unknown fields. Franzen and graduate student Tania Nanna are using a weighted classification technique. The zones are being developed using combinations of soil EC, yield, topography, imagery (aerial and Landsat 5) and Order 1 soil survey. Satellite imagery, topography and yield appear to be most useful nitrogen zone delineation layers to date. Use of multiple layers of data increases the correlation of the zones with base-line soil nitrate data. This technique will be automated within 2003 for general public use. Background materials on previous precision agriculture economic studies have been forwarded to Ag Economist Dwight Aakre. Data from 2001-2003 is being organized and will be presented this spring for economic analysis.
Impacts
The results of this study should make zone nutrient management easier and more effective in the region, resulting in more efficient use of N, and less environmental hazards for surface and ground water as the result of over-application of N to sensitive soils.
Publications
- Panda, S. S., Panigrahi, S., and Gautam, R. K., 2003. Learning vector quantization (LVQ) based neural classification for soil nutrient management. ASAE Paper No. 033066, St Joseph, MI.
- Gautam, R. K., and Panigrahi, S., 2003. Genetic algorithm for optimization of nitrogen application in the field. ASAE Paper No. RRV03-0022, St. Joseph, MI.
- Panda S.S., and Panigrahi, S. 2003. Heuristics integrated self organizing map (SOM) network for clustering of aerial images. Engineering Applications of Artificial Intelligence (accepted).
- Franzen, D.W. and Nanna, T. 2003. Comparison of nitrogen management zone delineation methods. p. 114-118. In Proceedings of the North Central Extension-Industry Soil Fertility Conference, Nov. 19-20, 2003, Des Moines, IA. Potash & Phosphate Institute, Brookings, SD.
- Nanna, T, and Franzen, D.W. 2003. A weighted classified method for nitrogen zone delineation. p. 177-184. In Proceedings of the North Central Extension-Industry Soil Fertility Conference, Nov. 19-20, 2003, Des Moines, IA. Potash & Phosphate Institute, Brookings, SD.
- Gautam, R. K., and Panigrahi, S. 2003. Image processing techniques and neural network models for predicting plant nitrate using aerial images. Proceedings of, International Joint Conference on Neural Networks. Gautam, R. K., and Panigrahi, S., 2003. Comparative analysis of different techniques for predicting soil nitrate using remote sensing images. Proceedings of the 30th International Symposium on Remote Sensing and Environment, Hawaii.
- Panigrahi, S., Yao, X., Gautam, R. K., and Franzen, D. 2003. Automated color assessment and classification of remotely sensed images of crops. Proceedings of the 30th International Symposium on Remote Sensing and Environment, Hawaii. Gautam, R. K., Panigrahi, S., Franzen, D., and Sims, A., 2003. Prediction of soil nitrate using image textural properties. ASAE Paper No. 033065, St Joseph, MI.
- Franzen. D. 2003. Variable N management in a sugarbeet rotation. 2003 ASA Agronomy Abstracts.
- Derby, N. E., Casey, F.X.M., Ralston, D.P.V., and Franzen, D.W. 2003. Precision agriculture: zone delineation and water quality. Long, D.S., Engel, R.E., Gessler, P.E., and Westcott, M.P. 2003. Determining field management zones using high resolution information from proximal sensing, remote sensing, and terrain modeling. ASA Agronomy Abstracts.
- Gautam, R. K., and Panigrahi, S., 2003. Plant nitrate determination from aerial image using neural network models. Submitted for publication in IEEE Transactions on System, Man and Cybernetics, Part c: Applications and Reviews, paper ID: SMCC-03-07-0077.
- Gautam, R. K., Panigrahi, S., Franzen, D., Sims, A. L., Lamb, J., and Smith, L., 2003. Soil nitrate prediction using imagery and non-imagery information. Submitted in the Transaction of the ASAE, paper ID: IET 723.
- Franzen, D. 2003. Delineating nitrogen management zones in a sugarbeet rotation. 2003 ASA Agronomy Abstracts.
|
Progress 10/01/01 to 09/30/02
Outputs
During the past year, data was gathered from each of the eight locations (Montana, Dan Long researcher; Minnesota, Crookston, Albert Sims, researcher; Minnesota, Renville, John Lamb, researcher; North Dakota, Valley City, Dave Franzen, researcher; Oakes, Francis Casey, researcher; Mandan, Vern Hofman, researcher; Minot, Mark Halvorson, researcher; Williston, James Staricka, researcher). Yields were recorded in each field using a yield monitor, except in Crookston, where sugarbeets were hand harvested to evaluate fertilizer treatments. Zones at Valley City, Montana, Mandan and Crookston were delineated using a method used previously by the researchers and fertilized accordingly in a randomized block design. Yields obtained in 2002 will be used to evaluate effectiveness of those first zones. Zone delineation techniques were evaluated by three graduate students under Francis Casey, Dave Franzen and Suranjian Panigrahi. At Oakes, a clustering technique was used to
delineate zones using soil EC, topography and imagery. At Valley City, a weighted classification technique is being evaluated using combinations of soil EC, yield, topography, imagery (aerial and Landsat 7) and Order 1 soil survey. Dr. Panigrahi's student, Ramesh Gautam, is applying data from all sites to develop zones using a neural network technique. Soil samples have been obtained at all sites to evaluate residual N following variable fertilizer applications using a uniform compared to a variable rate application of nitrogen.
Impacts
The results of this study should make nutrient management easier and more effective in the region, resulting in more efficient use of N, and less environmental hazards for surface and ground water as the result of over-application of N to sensitive soils.
Publications
- Franzen, D.W. and Nanna, T. Nitrogen zone delineation. 2002 Agronomy Abstracts. ASA-CSSA-SSSA, Madison, WI.
|
Progress 10/01/00 to 09/30/01
Outputs
During the past year, each state and each location within the state has made progress towards the goal of finding better ways to delineate nutrient management zones. Fields for study were established in Montana (Dan Long, researcher, two fields near Malta, 120 acres total), Minnesota (Larry Smith and Albert Sims researchers, 40 acre field near Crookston; John Lamb, 30 acre field near Renville), and North Dakota (Dave Franzen, 40 acres near Valley City; Jim Staricka, 40 acres west of Williston; Mark Halvorson, 11 acres south of Minot; Vern Hofman, 50 acres west of Mandan; Frank Casey, 40 acres near Oakes). Fields at Montana, Crookston, Valley City and Mandan had been previously investigated to some degree and a variable rate N application was made. The other fields were characterized using various methods. Fields were sampled to 4 feet in depth at sampling density of 0.25 to 0.5 acre grids and analyzed at least for nitrate-N. The Veris EC sensor was purchased and used
to map EC on each field except one at Malta, MT. Yield data was obtained from all fields. Quality information was collected from variable rate plots in the fields receiving variable rate application. An Order 1 was obtained from Valley City, Mandan and the fields in Montana. Elevation data has been obtained from each location except Minot and Crookston. Aerial photos were taken of all locations except Renville, where overcast skies were a problem. Satellite data was obtained from all locations. Graduate students have been placed as planned. Analysis of data is underway. Results to date have shown similar patterns of crop growth and soil parameters within fields. This indicates that delineation of nitrogen management zones may be improved using combinations of these data layers.
Impacts
The results of this study should make nutrient management easier and more effective in the region, resulting in more efficient use of N, and less environmental hazards for surface and ground water as the result of over-application of N to sensitive soils.
Publications
- No publications reported this period
|
|