While sampling and research for the 2016 season has been rolling along, we’ve also compiled data and preliminary analysis to share from 2015.
2015 was an incredibly dry year for the vast majority of agriculture producers across the Great Plains of the US, California and the Canadian Prairies. The widespread drought was said to be a major contributing factor for a number of large forest fires, increased consumer costs for produce, and added challenges for livestock producers who grew short on feed. By late August, the Government of Alberta declared an agricultural disaster in order to free up additional funds for producers in need of insurance compensation.
Across our four ranches, we saw conditions and precipitation measurements consistent with the reports of drought, particularly in the months of May and June (as seen in the chart below). The one exception to this was at the Porcupine weather station near Claresholm, which read 200mm of precipitation for May and June, roughly 82% of what we’d expect.
Through July and August, precipitation increased (especially in the McLaughlin and Jenner regions, where they received close to normal levels of precipitation in those months) but for many producers it was too little, too late—the damage had already been done.
On the ground, the effects of low precipitation were obvious. Especially early in the year, green grass was minimal and much of the material collected was carryover from last year. But things grew greener in July and August, which allowed our researchers to test readings for dry and healthier green conditions. Thanks to these varied conditions, we will be able to determine whether NDVI correlation to growing material exists regardless of moisture levels.
NDVI and green material correlation
The purpose of our study is to see if we are able to use satellite readings of pasture as an indicator of pasture growth. To do this, we are measuring the amount of green grass and forbs by the pound, and comparing what we see to an NDVI reading for the same area that is collected by satellite and with our handheld spectrometer. NDVI stands for “Normalized Difference Vegetation Index” and is calculated by measuring light absorbed relative to light given off the surface of the Earth to determine the amount of living vegetation on the ground.
Long story short: we are seeing good correlation between NDVI satellite values and the amount of green material on the ground, especially when we average five sample points (NDVI vs. green grass) together. Averaging helps to reduce errors that occur in any field sampling. While it is too early to make a call on the accuracy of this correlation, we like what we’ve seen so far. One area we will want to watch over the next year or two is the correlation of smaller values. As you’ll see in the charts below, the X/Y values of NDVI and lbs/acre increase together, but more obviously with NDVI readings over 0.5.
To solve for low production values, we’ll be selecting more sites of low production rather than solely relying on the approach of random selection, which we employed over our first year. That will give us more data in the lower range of NDVI values and should help us better understand the kind of growth/NDVI correlation we have across the dry-green spectrum.
Over time, with confidence in the correlation between NDVI and growing material, we can start to estimate how much material is growing per acre on the ground based on the score we receive by satellite (see charts below). For example, if we see an NDVI score of 0.75, we can assume that approximately 1,200 lbs/acre of green material is being produced on the ground. This information could help insurance designers calculate the 10-year average of pasture production for a specific producer (historical base-line insurance coverage) and also determine whether they are seeing unusually low growth in any given year (settling an insurance claim).
One year of field samples is not sufficient to demonstrate confidence in a correlation between an NDVI and grass production. In research, more is always better. In this case, more years of field sampling research will give us more confidence in a correlation, if one truly exists.
Results of the first years’ research has returned encouraging results. For example, an insurance design would be much simpler if there was a single relationship between NDVI and pasture production for: winter vs. summer pasture, different months in the growing season, different ranches, different grass species, etc. If that was the case, then an NDVI value would indicate the same pounds per acre of pasture regardless of differing parameters. Preliminary results indicate this could be the case—the difference among these various “production situations” has (so far) been negligible.
On another note, we are pleased to annouce that the primary contributor to this project, AgriRisk Initiatives, has expanded our research budget for the 2016-2017 season to increase the pasture samples we can collect, sort and analyze, extending the project for one more year. This will increase the confidence we have in any NDVI to pasture-production relationship that we see develop.