Loading…
6 A Comparison of Methods for Estimating Forage Mass
Abstract A summer internship project was created to increase our understanding of how new technologies may increase efficiency of estimating forage yield while not sacrificing accuracy along with providing hands-on learning in Animal Science Extension programming. Forage mass (FM) data were collecte...
Saved in:
Published in: | Journal of animal science 2018-03, Vol.96 (suppl_1), p.3-4 |
---|---|
Main Authors: | , , , |
Format: | Article |
Language: | English |
Online Access: | Get full text |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Summary: | Abstract
A summer internship project was created to increase our understanding of how new technologies may increase efficiency of estimating forage yield while not sacrificing accuracy along with providing hands-on learning in Animal Science Extension programming. Forage mass (FM) data were collected in 2, grazed tall fescue (Festuca arundinacea) fields totaling 6.5ha, late May, using 1) traditional rising plate meter (RP), 2) Garmin LIDAR-Lite 3 (Garmin, Olathe, KS) connected to an Arduino Uno (open-source hardware) microcontroller (LIDAR) with data logged to a computer while simultaneously logging GPS position, and 3) vegetation indices derived from a red+nir band camera (Survey2, MAPIR, San Diego, CA) attached to a DJI Phantom 4 drone (DJI, Shenzhen, China). Both RP and LIDAR were calibrated independently using clipped forage samples within a 0.19 sq-m frame. Images were imported, georeferenced and overlapping images were stitched (Photoscan, Agisoft, St. Petersburg, Russia) to create a single image. The image was also calibrated using MAPIR’s calibration target. Non-calibrated image data contained red, green, and nir band data; whereas, the calibrated image only had adjusted red and nir band data. A raster calculator (QGIS, www.qgis.org) was used to calculate NDVI = (nir - red)/(nir + red), SAVI = ((nir - red)/(nir+red+L))*(1+L); use L of 0.2, GNDVI = (nir-green)/(nir+green), and GRVI = nir/green. Images were further aggregated by a factor of 2^n where n ranged from 2 through 9. Forage mass did not differ (P = 0.63) between RP (N = 44, mean = 1826.4, and cv = 47.2) and LIDAR (N = 986, mean = 1892.4, and cv = 63.1). Without image aggregation GRVI, GNDVI, NDVI, and SAVI exhibited very little correlation with LIDAR point estimated FM (r(pvalue) =0.076(0.02), 0.076(0.02), 0.028(0.39), and 0.028(0.39), respectively). At an image aggregation factor of 79 (representative of 4.6m GPS error), GRVI, GNDVI, NDVI, and SAVI exhibited small positive correlation (P |
---|---|
ISSN: | 0021-8812 1525-3163 |
DOI: | 10.1093/jas/sky027.006 |