Credibility of Fire Blight Forecasts

Given that apples and pears are blooming again in the Hudson Valley, it’s time to think about fire blight. The primary risk from fire blight occurs during bloom, as everyone should know by now. The fire blight bacterium, Erwinia amylovora, is spread by insects and/or rain splash to the stigmas of open flowers where the bacteria multiply very rapidly at temperatures above 65 F. If rain occurs after populations reach infective levels, then the bacteria are washed from the stigmas to the base of the flowers where infection occurs through the flower nectaries.

Two fire blight risk models, MaryBlyt and Cougar Blight, have been widely used in North America to predict when infections are likely to occur and thus, when sprays are needed to protect blossoms from fire blight. Last year (2015) both of these models indicated that temperatures were favorable for bacterial multiplication on stigmas throughout most of the bloom period. Given those conditions, infections were expected to have occurred on days when there was enough rain to move the bacteria down to the nectaries. The very small amount of rain needed to make pavement look wet is considered sufficient for moving bacteria into the nectaries. In the lower Hudson Valley, we had three light rains during bloom. Based on the model outputs, I expected severe fire blight outbreaks would occur where streptomycin was not applied at least twice during the season. In fact, I sprayed the plant pathology blocks at our research station three times to cover the three potential infection periods that occurred during bloom in 2015.

What happened? Very little fire blight occurred in the Hudson Valley last year, and it appeared that any blocks that were sprayed at least one time during the bloom period were completely protected from blight regardless of when that spray was applied whereas blocks that had received no strep sprays sometimes had a few strikes. At the end of the season, I was left asking how the risk models (and my own predictions based on those models) could have been so wrong.

We still do not have a definite answer to that question, but I suspect that blight failed to develop last year because the relative humidity throughout bloom was abnormally low. In a research paper published in 2000 (Phytopathology 90:1352-1357), Dr. Larry Pusey reported that very little blossom infection occurred in growth chamber trials with detached flowers when the relative humidity was below 80 to 85%. In most years, wetting periods in eastern United States are associated with at least short periods of high relative humidity, but that was not the case during any of the bloom-time wetting periods at the Hudson Valley Lab in 2015. Unfortunately, no one knows the minimum duration of high RH required for fire blight infections, nor do we know how the sequence of high RH periods (i.e., before, during, or after rains) might impact infection potential under the varying conditions that we find in the field.

So how should we interpret fire blight risk models in 2016? The conservative approach will be to continue following risk models just as has been done for the past decade or more and ignore RH as a factor until researchers can sort out the details. Extreme caution is required if one chooses to ignore strep sprays when our current models indicate fire blight risk is high because (1) we still do not have data indicating how to incorporate RH into the existing models, and (2) missing a strep spray when it is really needed will be far, far more costly than applying an extra strep spray when perhaps it was not needed. Furthermore, the amount of inoculum that is present will be very important in determining the severity of blossom blight, and right now we have no way to evaluate the presence of inoculum other than basing our estimates on orchard history.  That means that a very conservative approach is especially important in blocks that had a lot of blight last year.

Nevertheless, for those who wish to consider RH as a factor in fire blight forecasting, I suggest the following process as a starting point:

  1. Follow MaryBlyt or Cougar Blight models as one normally would during bloom.
  2. When the models predict either high risk with rain expected (Cougar Blight) or infection in MaryBlyt (which only occurs when wetting is added to other factors), then check relative humidity levels.
  3. Finally (and this is the part where we are really flying by the seat of the pants), I would suggest that if RH does not exceed 85% for at least four hours either during the wetting event or continuous with it (i.e., immediately before or after the wetting period), then the risk of infection may be rather low even if the models suggest otherwise.

One way that I have habitually crosschecked conclusions about blossom blight risks over the past 30 years is by applying what I have called the “personal discomfort index” or PDI. The PDI works as follows: “In early spring, if outdoor work that requires a modest exertion causes a person to feel uncomfortably hot and sticky, then one should check to see if apples and pears are in bloom. If flowers are present and strep has not been applied in the past two or three days, then begin strep sprays immediately.” If a PDI event is triggered, it is almost always because warm temperatures have combined with high RH to create that feeling of personal discomfort. Thus, in retrospect, my PDI for identifying blight infection periods may have been ahead of it’s time. (OK, I can hear all of my colleagues laughing! And I admit that what works for me in the Hudson Valley may not work for other folks in other places.  This is certainly true for those who never get out of the office or for those whose body temperature sensors fluctuate wildly 🙂

Last year when our risk models were predicting severe blossom blight conditions, I was a bit concerned because the conditions never triggered a PDI-event. I decided the models had to be more scientific than my subjective PDI analysis of blight risk and therefore warned about the potential for severe blossom blight if strep sprays were not applied. In retrospect, I wish I had trusted my PDI analysis a bit more. Since we now know that the models did not work well in the Hudson Valley in 2016, I’ve opted to give you both a quasi-scientific “seat of the pants” approach for adjusting model output this year with RH considerations, and I’ve also risked explaining my PDI approach for evaluating blight risk. As you can imagine, I will refuse to take responsibility for control failures that might be associated with using either of these unproven approaches!

What are the models showing now for the lower Hudson Valley? When I input into MaryBlyt a first bloom date of April 20 (Wednesday) along with the Accuweather forecasted temperatures for today (Thursday) and next week, MaryBlyt suggested that we could have an infection event with rains tomorrow (Friday) and again next Tuesday. I think that the risk of infection on Friday of this week is virtually nil, both because many orchards did not have many flowers open yesterday and because the predicted highs of 78 F for both today and tomorrow may never be reached. Furthermore, both Cougar Blight and RIMpro discount infection predictions during the first few days of bloom. The predicted infection for next week, which shows up in both MaryBlyt and the RIMpro fire blight models, is more “iffy” and bears watching. However, the current forecast for next week suggests only scattered showers are likely, and those often occur without generating extended periods of high RH. Thus, in my opinion, fire blight risk is relatively low in the lower Hudson Valley for at least the next few days and quite possibly for all of next week.

This entry was posted in Seasonal notes. Bookmark the permalink.

Comments are closed.