Progress report for GNE21-272
Estimating grass content of mixtures has not yet been attempted on any hand-held NIR instrument. The introduction of near infrared (NIR) technology in a small portable unit has the potential for a cost-effective solution to the problem of evaluating and managing variability in alfalfa-grass composition. The technology could improve the ability for alfalfa-grass producers to optimize field management and reduce variability in dairy rations, resulting in more environmentally and economically sustainable farming systems. Harvest management and rotation of alfalfa-grass stands is often a function of the grass percentage of the mixture, and an on-farm NIR unit can provide a record of stand composition, required for correct nutrient management decisions. This proposal includes a focused evaluation of the Neo Spectra Scanner where the main objective is to collect spectra and develop calibrations of alfalfa-grass compositions using wet, chopped samples. Farmers and farm consultants are very interested in the collection of independent research data on the functionality and feasibility of on-farm hand-held NIRS devices and this project will provide the research needed urgently in the agricultural industry. Outreach activities are essential to disseminate results to the agricultural community and will involve publication of articles in regional farming magazines, presentations at local field days and national conferences, and publication in a peer-reviewed journal.
# 1. Create a database of NIR spectra with the Neo Spectra Scanner, using alfalfa-grass samples with the full range of grass% from 0 to 100%.
# 2. Develop calibrations for the Neo Spectra Scanner for estimating grass percentage in alfalfa-grass fresh mixtures.
The purpose of this project is to collect spectra and develop calibrations for the Neo Spectra Scanner for estimating grass percentage in alfalfa-grass fresh mixtures. The Neo Spectra Scanner is a hand-held near infrared reflectance spectroscopy (NIRS) unit that has been recently developed in Egypt and it is one of the few hand-held NIR units that allows users to collect spectra and develop calibrations. The Neo Spectra Scanner currently has no forage calibrations, but has great potential with a wide NIR spectral range of 1,350 to 2,500 nm. All other hand-held instruments have a narrower NIR scanning range. On-farm NIR analysis will be particularly important for alfalfa-grass producers, to improve field management, optimize nutrient management by reducing variability in dairy rations, thus creating opportunity for more sustainable dairy farm production systems (Cherney et al., 2021b).
Alfalfa-grass mixtures are commonly integrated into the Northeast agricultural dairy systems. Over 84% of alfalfa sown in New York State is done in combination with a perennial grass (Karaylilanli, et al., 2016). Binary alfalfa-grass mixture stands can provide a more balanced nutritional forage for livestock than a pure crop of either alfalfa or perennial grass (Cherney, et al., 2002; Cherney, 2020). Relatively flat land with optimal soil conditions for alfalfa production can result in excellent stands of high yielding pure alfalfa with sufficient stand persistence, but hilly land, or land with suboptimal soil conditions for alfalfa, however, can benefit from alfalfa-grass mixtures (Cherney and Cherney, 2020; Cherney et al., 2020b). Interest is steadily increasing across the northern US towards incorporating alfalfa-grass mixtures into forage-dairy systems. Alfalfa-timothy (Phleum pratense L.) mixtures in eastern Canada increased both yield and nutritive value over pure stands of alfalfa (Bélanger et al., 2014). This unusual combination of increased yield and increased nutritive value is because grasses have higher neutral detergent fiber digestibility (NDFD) than alfalfa, and alfalfa-grass mixes tend to have higher yields than pure alfalfa stands (Cherney, et al., 2002; Cherney, 2020).
Diversity of plant functional types provide additional benefits in agricultural production systems by supporting multiple ecosystem services, including reduced weed invasion compared with monocultures (Picasso et al. 2008; Frankow-Lindberg et al. 2009; Sanderson et al. 2012). Competition from non-legume species also has been shown to increase forage legume N fixation (Nyfeler et al. 2011). Besides being beneficial for forage yield, milk production and weed control, legume-grass mixtures can also benefit soil health. Improved productivity of belowground microbial biomass was similar to the improved productivity of aboveground plant biomass, indicating that grass-legume mixtures have a positive impact on soil productivity and fertility resilience, which leads to improved soil health in the long-term (Dakhal and Islam, 2018). Surface erosion and leaching also will be reduced compared to pure alfalfa stands due to the fibrous root system of grasses. While alfalfa-grass mixtures are preferred on land that is suboptimal for pure alfalfa, the greatest obstacle to overcome with mixtures fed to dairy cattle is not forage quality, but forage variability.
Portable NIR instruments have the potential to improve both the economic and environmental wellbeing of forage and dairy farmers, but only if they provide relatively accurate and precise information on forage composition. On-farm NIR analysis will be particularly important for alfalfa-grass producers, to improve field management and reduce variability in dairy rations. We have purchased three Neo Spectra Scanner ‘Alpha’ version instruments. Our hypothesis is that the Neo Spectra Scanner, with its wide NIR range, will provide accurate and precise alfalfa-grass composition information on farm.
Approach and Methods #1
a. Sample Collection
The first set of fresh alfalfa and grass samples were collected during the summer of 2021 from Cornell Caldwell farm. Additional fresh alfalfa and grass samples will be collected over the summer of 2022 from a minimum of 20 dairy farms in NY, with multiple alfalfa varieties, and grass species and varieties. With fresh samples, we need to immediately refrigerate samples in the field and transporting samples back to a laboratory setting for processing. Freezing of fresh samples would preserve them for analysis, but a thawed out frozen alfalfa sample is physically quite different from fresh, and likely does not have the same spectral characteristics as a fresh sample. Pure alfalfa and pure grass samples as well as mixtures will be collected. Mixtures will be separated and recombined in known proportions. Pure species samples also will be combined in known proportions. Fresh samples will be chopped using a Hege 44 stationary sample chopper (Wintersteiger Inc., Salt Lake City, UT). The unit has a band conveyor for feeding forage, and chopping length is easily set by a stepless adjustment of the conveyor speed. We plan on collecting at least 300 fresh chopped samples from dairy farms to combine in known mixtures for scanning and laboratory analysis. Over the summer and fall of 2021, we harvested, combined and scanned 150 fresh chopped samples. We are currently developing a calibration model in MatLab with the help from statistical consultant, May Boggess at the Cornell Statistical Consulting Unit and will continue with this work during the winter of 2022.
Each fresh or silage sample will be distributed uniformly in a 410 x 40 x 13 cm deep rectangular container, maintaining a thickness of at least 5 cm. Samples will be remixed in between each of the four scans for each method. We will collect both stationary and sliding spectra scans with the Neo Spectra Scanners to compare the two methods. We have already demonstrated that a laboratory NIRS instrument can be calibrated to predict the grass percentage of mixed alfalfa-grass forage that is dried and finely ground (Karayilanli et al., 2016). This calibration is currently being used by DairyOne Forage Laboratory to provide farmers with species estimates in alfalfa-grass binary mixtures, assisting with both stand nutrient management and crop rotation decisions. To our knowledge, no one has attempted to calibrate a hand-held NIRS instrument for alfalfa-grass species composition using wet, chopped samples. We will use pure alfalfa and pure grass samples, mixed in known fractions, to calibrate the Neo Spectra Scanners, and assess the success of these calibrations.
Si-Ware Systems will be turning out the production version of the Neo Spectra Scanner in 2021. We currently have three of the ‘Alpha’ versions of the scanner for testing. These will be evaluated with fresh chopped alfalfa and grass in known mixtures. Preliminary evaluations have determined that four scans of four seconds each provide repeatable scan data from a wet, chopped alfalfa-grass sample. Although Si-Ware Systems does not have a protocol for moving the scanner during the scanning process, they have confirmed that the instrument is fully capable of integrating the spectral data received during a moving scan. The protocol requires that the instrument remain in direct contact with the sample throughout the scanning process, even a small gap will dramatically impact the recorded spectra. All samples will be scanned with three Neo Spectra instruments from three different production batches, using both stationary and sliding scans, four scans for each scanning method.
Approach and Methods #2
a. Calibration and Validation
Calibrations will be developed by relating near-infrared spectra to oven moisture reference values. Spectra will exported from the Neo Spectra Scanner and subsequently imported into Wolfram Mathematica (Version 12.1, Wolfram Research, Inc., Champaign, IL, USA). Mathematica will be used for all subsequent data analysis, including mean centering, math pre-treatment, partial least square regression (PLS), cross-validation, and computation of model performance statistics.
b. Evaluation of the Hypothesis
Collecting a set of samples with a sufficient range in composition for the purpose of NIRS calibration, with subsequent evaluation of calibrations, does not involve a classical experimental design. The key to successful calibration is in collecting a range of diverse samples for a robust dataset. Hypotheses are tested using evaluation criteria for calibrations. Near infrared reflectance calibrations will be evaluated with multiple criteria (Malley et al., 2005). The coefficient of determination (R2) is the proportion of variability explained by the model. The root mean square error of prediction (RMSEP), also known as the standard error of prediction (SEP), is the average difference between measured and NIR-predicted values, and is used to calculate two additional criteria. The residual prediction variation (RPD) is the standard deviation (SD) of the reference data divided by RMSEP, and is considered a better measure than RMSEP, as it relates RMSEP to the range of the reference measurements (Foster et al., 2013). Additionally, the range of the reference data divided by RMSEP (RER) has been used to assess the success of NIR calibrations (Ward et al. 2011).
Education & Outreach Activities and Participation Summary
- A journal article will be generated assessing the accuracy and precision of the Neo Spectra Scanner for on-farm use to determine grass% in alfalfa-grass mixtures.
- Outreach presentations:
- ASA-CSSA-SSSA: International Annual Meeting - American Society of Agronomy, Crop Science Society Of America, Soil Science Society Of America (Nov 6 - 9, 2022)
- Cornell Nutrition Conference (November 2022)
- Northeast Agribusiness and CCA advanced training, 2021 & 2022
- Northeast Organic Farmers Association Conference, 2022
- Outreach publications:
- Interpretative publications based on the research results will be developed and published as two traditional “fact sheets”.
- We will also develop popular press articles for national distribution in magazines (Progressive Forage, Hay & Forage Grower, Hoard’s Dairyman, etc.).
- Post results on a well-established forage crop website, www.forages.org.