The information that we are gathering from MODIS satellites is fantastic, but we need to ensure the data aligns with what producers are seeing on the ground. As we mentioned in an earlier post, the only way to accurately determine this is to get people into some pasture and count the blades of healthy grass.
Luckily, we have three researchers crazy enough to dedicate their summers to the pursuit of grass samples and scientific truth. This is their story.
Our Research Strategy
Armed with a satellite-equivalent spectrometer and a variety of tools, the team of Richard McConnell, Tom Crozier and Mackenzie McConnell have been collecting and analyzing grass samples from four locations across Alberta.
Last year, our friends at the Canadian Cattlemen’s Association (CCA) sent word of our project around the community of cattle producers and we’ve been fortunate to have four volunteers lend portions of their pasture to us for sampling. That gives our team field locations on the Eddleston Ranch (northeast of Wainwright), Osadzchuk Ranch (northwest of Medicine Hat), Hargraves Ranch (at the Saskatchewan border east of Medicine Hat) and Burke Creek Ranch (on the eastern slopes west of Claresholm).
At each field location, the team has selected four 250-meter-square areas and laid fifteen metal rings (33 in. / 85 cm in diameter) within those areas. The pattern of our sampling arrangement in each 250-meter-square sampling area is shown to the left. Wire cages are used to cover each ring in areas of open grazing in order to eliminate the risk of having our samples devoured by livestock.
Once the rings and cages are set, the team gives the grass time to grow before returning. They then take spectrometer readings and grass clippings from each ring throughout the summer months to determine the health of the grass and compare to MODIS NDVI readings taken at the same time.
To first determine whether the satellite and handheld spectrometers are calibrated, the team begins each data collection mission by taking the handheld spectrometer around the perimeter of the 250 x 250 meter sample site. Readings are taken with the spectrometer on the ground, which are then averaged and compared to the reading of the 250-meter-square pixel reading provided by satellite. Once the team was confident that their manual tool was calibrated, they could begin their work.
Step 1: Pre-clip NDVI readings
At each sampling ring, the team took a preliminary NDVI reading with the handheld spectrometer to get a sense of the “greenness” of the grass before being cut. Remember, one of our primary challenges is to understand how much healthy grass is represented by a particular NDVI score. So, in addition to sampling 250-meter-square sections and getting a feel for how much grass is represented by a large-scale NDVI score provided by satellite information, the research team has been trying to see if the difference between pre-cut and post-cut NDVI numbers equates to a weighed quantity of grass. As of today, that theory remains inconclusive.
Step 2: Sample inspection
At each ring, a researcher inspected and documented the contents of that ring. He or she would document whether there was grass, forbs, bare ground, etc. present within the ring. A photo of the contents of the ring is also taken.
Step 3: Grass clipping
The contents of each sample ring is clipped approximately 1.5 cm off the ground (to simulate the height of material left by grazing cattle) and carefully placed into a marked paper bag. The paper bags are then put into “onion bags” to allow the sample material to dry before the cataloging process. To complete the drying process, all samples were put through drying ovens located at the AAFC Research Station in Lacombe, AB for 72 hours prior to sorting.
Step 4: Post-clip NDVI readings
Once a ring has been clipped, another spectrometer reading is taken and documented. Again, the hope is that the difference between pre- and post-clip NDVI readings will show correlation with the amount of green material collected.
Step 5: Grass sorting
After drying is complete, the samples are taken to the McConnell acreage in Central Alberta (shown left) for sorting. The samples are then emptied into a foil tray and then sorted into three separate categories—green grass (this year’s growth), carryover, and forbs/herbaceous material.
Each sample sorted is recorded in a ledger that includes the sample clip site, clip area and date the sample clip was taken along with the weight (in grams) of each component of the grass sample—green, carryover, forbs. The information is also stored in an Excel sheet with the handheld spectrometer readings.
Step 6: Comparison
As sampling continues throughout this summer and next, close attention will be paid to the correlation between NDVI readings and the amount of grass clipped and weighed from our different locations.
Hopefully we can develop a clear relationship between NDVI values taken by the handheld sensor (the “X”), and green grass production collected at the sample locations (the “Y”). If we can, and we can prove the handheld sensor is giving similar readings as the satellite, then this “X” to “Y” relationship built from sensor information can also be used to measure pasture production with the satellite.
The relationship between the satellite’s NDVI values and green pasture could form the basis of an insurance program. A low “X” from the satellite would mean a “low Y” or poor pasture production and could mean that an insurance payment should be made.
Our goal is to find a relationship between the amount of forage found in a pasture and NDVI readings provided by satellite. We think that if we can accurately measure the amount of forage in a pasture by reading satellite data, we can find average growth for specific fields over years past and set premiums that are viable for both the producer and insurance provider.