Models are generally at their best when interpolating, not when extrapolating. Although models are often used to predict the future, one of their greatest strengths is their ability to “fill in the gaps” of our monitored data (which are often sparse). Those of us who grow cold-tender varieties of wine grapes in the eastern U.S. need to make winter/spring pruning decisions based on an informed understanding of the risks of winter-injury given the weather conditions in a given year. Temperature monitoring and cold-hardiness models can combine to give us a leg up on planning our pruning decisions.
Sub-zero temperatures in early January had us a bit on edge, reminding of February 2015 when we saw temperatures down near -10°F at our vineyard in Central Virginia, and suffered widespread bud damage in a 2nd year vineyard block, with crown gall apparent in nearly 20% of the vines by year 3. When our weather station showed temperatures down near -5°F last week, we were feeling gun shy and wondering if we should be tweaking pruning strategies this winter and shoot thinning this spring. Not entirely confident in my ability to inspect buds for damage and really know what I am seeing, I decided to run a cold-hardiness model.
The Cold Hardiness Model
Washington State Viticulture and Enology program studies cold hardiness to determine the ways in which winter temperatures to date affect the grapes abilities to withstand extreme lows, in essence: the colder the winter, the hardier wood. If you have mean daily temperatures since September, their model let’s you estimate the “critical temperature” beyond which 50% bud loss is expected (for DIYers, see here: http://wine.wsu.edu/extension/weather/cold-hardiness/model/ ). Figure 1 shows the results for our vineyard – we were within 1° of the critical temperature twice during the first week of 2018, bottoming out at -4.1° on the 7th. This suggests that we were still in the safe zone (just barely), but to be sure we headed outdoors anyhow.
Models Can Guide the Extent of Sampling
Thankfully, a sample of 30 buds and several canes from various spots in our vineyard showed no obvious cold damage. But we might ask, if we can just sample the buds, why bother running a model? I think that models can be useful to:
- Save time – sampling/analyzing buds takes time, and the larger the farm, the more work. If models show a low risk, we may choose to sample only known trouble spots.
- Fill in the gaps – if you’re unsure of your ability to interpret what you get out in the field, or don’t have time to sample everywhere, models can help you interpret what you see.
- Alert us to situations we might not otherwise recognize – Maybe we didn’t think it was that cold? For example, this winter has been a cold one, but what did that warm spell during mid December do to the vines?
Had the models predicted greater risk of damage, we would have returned to the field for another set of buds to insure that we got a representative sample. Had we found some damage, we can adjust winter pruning to keep more buds as an insurance policy later in the season.
Models contain the distilled understanding of years of scientific and practical experience generalized into algorithmic form — and they can be a fantastic teacher as we gain our own experience.