Sensor research and algorithm development for corn in ND L.K. Sharma, D.W. Franzen, H. Bu, R. Ashley, G. Endres and J. Teboh North Dakota State University, Fargo, ND
INTRODUCTION Corn acreage in North Dakota is increasing at a very high rate in the last 10 years. • Area and production 2011- 2.06 million acre and 5.5 MT 2012- 3.2 million acre and 8.5 MT
Corn N timeline Period of greatest uptake Application Day 1 Day 45 Day 80 Day 120
Early Growth Rapid Growth Late Loss Maturing 100 75 Seasonal N Uptake, % 50 25 Over 80% of N required after V8 0 Sept May June July Aug The first 6 weeks of growth, little N is needed Source: Dr. Jim Schepers, NUE conference presentation, Fargo-http://nue.okstate.edu/Nitrogen_Conference2012/North_Dakota.htm
In high clay soils • Leaching is not an issue. • Downward movement of water in a high clay soil (Fargo soil series) is about 0.015 inches per hour, or about 1/3 inch per day.
Image taken June 28, 3 days after area was covered by 6 inches of water
Active optical sensors have been identified as a tool to increase nitrogen-use efficiency GreenSeeker™ (Trimble) Holland Crop Circle Sensor™ (Holland Scientific)
User Selected Filters Holland Crop Circle-470 TARGET Red Red Edge NIR LED ACS-470 SENSOR Source: Dr. Jim Schepers, NUE conference presenattion, Fargo-http://nue.okstate.edu/Nitrogen_Conference2012/North_Dakota.htm
Greenseeker emits two bands visible and near infrared: • NDVI= (NIR – Red)/(NIR+Red) • (774nm reading – 656nm reading)/(774nm + 656nm) • Or • (774nm reading – 710nm reading)/(774nm + 710nm) • (New GreenSeeker) • Crop Circle-470 emit three bands visible, red edge, and near infrared: • NDVI= (NIR – Red/(NIR+Red) • (760nm reading – 670nm reading)/ (760 + 670) • Or • NDVI= (NIR – Red Edge/NIR+Red Edge) • (760nm reading – 730nm reading)/ (760 + 730)
Materials and Methods • Experimental design: Randomized complete block design with four replications. • Plot size: 20 feet long by 10 feet wide • Soil was sampled to 2-feet in depth for residual nitrate-N preplant. • P and K applied, if found deficient and cooperator application not practical Locations and Treatments • 51 sites were selected in 2011-2013. • Six nitrogen treatments: 0, 40, 80, 120, 160, and 200 lb/acre.
Crop History & Soil Texture • Previous crop • Tillage history • Surface-subsurface soil texture Sensor readings • Approximately 45 samples/row • The NDVI values were averaged for each plot as well as for each treatment. • Both sensors Crop Circle and Greenseeker were used • 8 and 12 leaf stage over the top
Location segregation All research Sites Western sites Medium Textured Sites No till Sites East High Clay Sites Conventional till Higher yields/lower yields
Corn yield difference in kg/ha. X 1.25 % N in corn grain divided by efficiency factor 0.6 = N rate in kg/ha Reference Yield Field Yield estimate Yield Reference INSEY INSEY in field INSEY
Example- Reference yield predicted- 120 bushels In-field yield estimated- 60 bushels difference = 60 bushels X 56 lb N/bushel = 3360 pounds X 0.0125 = 42 lb N 42 /0.6 efficiency factor = 70 lb N at that location.
Wavelength evaluation of two ground based active optical sensors to detect sulfur deficiency in corn using N rich within field areas
Tillage system, soil type, planting date and date of the first and second sensing of experimental sites.
Relationship between N rate and Crop circle red edge INSEY (Crop circle red edge wavelength reading/growing degree-days), V6 at Arthur Relationship between N rate and Crop circle red edge INSEY (Crop circle red edge wavelength reading/growing degree-days), V12 at Arthur
Relationship of Crop Circle red edge INSEY (sensor red edge NDVI/growing degree-days from planting to sensing) and N rate, V6 stage at Arthur Relationship of Crop Circle red edge INSEY (sensor red edge NDVI/growing degree-days from planting to sensing) and N rate, V12 stage at Arthur
Overall Conclusion • Multiplying INSEY by the corn height improve the relationship between INSEYS and Yield. • Red edge NDVI is better at 2nd stage than Red NDVI. • Crop circle red edge was found better as compared to Greenseeker 2nd stage. • V12 leaf stage was found better in predicting yield as compared to V8. • Inseason N rate algorithm was successfully build with help of sensors.
This algorithms is a starting point for growers. NDSU Computer Science (Anne Denton- a ICPA presenter) are developing a ‘machine learning’ tool, which will help growers to add their data into the existing algorithm to make the algorithm their ‘own’.
ACKNOWLEDGEMENTS • Special Thanks • Dr. Dave Franzen (PhD Advisor) • Dr. Tom DeSutter (PhD Committee member) • Dr. R. J. Goos (PhD Committee member) • Dr. Joel Ransom (PhD Committee member) • Thanks to the North Dakota Corn Council, IPNI and Pioneer Hi-Bred International for their support of this project. Also to Dr. Anne Denton, NDSU Computer Science Department • and the National Science Foundation. Also to • Honggang Bu, Brad Schmidt and Eric Schultz.