May, 1999
How to predict blossom blight using disease models
By Deborah I. Breth, Cornell Cooperative Extension
and
Timothy Smith, Washington State University
There are many prediction systems used to predict diseases in agricultural crops. They are usually referred to as models. All models are based on general biological principles, packaged in a computer program or a set of rules to make predictions or decisions.
In New York, the Maryblyt 4.3 model is used for fireblight (FB) blossom blight prediction. It was field tested in 1992 through 1994 after the epidemic experienced in 1991. It has been demonstrated empirically over the past five growing seasons to accurately predict blossom blight risk.
Many times, models developed in one part of the country are not appropriate for use in other climates. So other models for the same disease are developed based on regional conditions.
Cougarblight is another fireblight prediction model developed in the Northwest. Based on growers concerns that Maryblyt "may not be perfect," I have been running the Cougarblight model alongside Maryblyt to determine its value for predicting blossom blight in New York. Although my testing has shown some parallel recommendations between the models, it is important to understand the differences in how the two models are used.
Biological principles
Predicting blossom blight requires a general understanding of the biology of Erwinia amylovora (Ea), the bacteria that causes fireblight, and the susceptibility of the apple blossom. Bacteria reproduce by cell division: one cell becomes two, two cells become four, etc. A single cell of Ea will not result in fireblight symptoms unless conditions for multiplication are met. The bacteria colonize the surface of the stigma on the flower. For colonization of the blossoms, the stigmas must be exposed with petals open. Some studies have shown as many as 10,000 bacteria can be present on one stigma. Multiplication of bacterial cell numbers can occur between 39-90oF, but most rapid multiplication occurs between 75-84oF. For infection of blossoms to occur, the nectaries at the base of the flowers must be open and functional (before petal fall). In general, flowers have an expected lifetime and are susceptible to infection for about four days.
The bacteria have flagella, which help them swim and propel themselves through a continuous film of water after a wetting event. The bacteria swim from the stigma and are attracted to simple organic acids in apple nectar. If pear or apple blossoms are wetted through dew, rainfall or spraying, the number of fireblight strikes and the extent of damage they cause depends on three factors.
Knowing the bacteria multiply at different rates depending on the temperature, a predictive system of degree hour accumulation was developed in California using colonized pear blossoms to determine the threshold for predicting a significant number of blossoms colonized by bacteria after the start of bloom.
Cumulative heat units expressed as degree hours (CDH) for blossom blight are calculated over a threshold temperature of 65°F. Daily degree hours were calculated by recording the daily maximum, minimum, and average temperature, subtracting the threshold temperature of 65° from each value (if value is <0, use 0). It is assumed that the maximum and minimum daily temperatures last for six hours, and the average temperature occurs over a 12-hour period. This research showed that there is an apparent threshold of 198 CDH before enough flowers are colonized for an epidemic. The difference in models lies in how these cumulative degree hours are calculated.
All models start at first bloom in the orchard and must continue through complete petal fall, including rattail bloom on many varieties. Based on the above principles, the obvious factors that must be monitored for prediction of blossom blight in all models include: maximum, minimum, and average daily temperature, wetting events in the form of rain, dew or spraying, the presence of blossoms and the weather forecast for the next couple days.
Using Maryblyt
The Maryblyt model was developed by Steiner, and Lightner under Maryland conditions, and tested in many other states. They took the basic principles and adjusted the daily distribution of temperatures using a sign wave that more closely represented the degree hours gained on a daily basis. They combined the expected lifetime of blossoms into the model. They adjusted the model for differences in population growth in cold temperatures. If temperatures drop to less than or equal to 32°F and the CDH is less than 400, the CDH becomes 0. If the maximum temperature is less than 65°F for one to three days, the CDH is reduced by a third the first day, half the second day, and the remainder the third day.
Maryblyt is a DOS-based computerized model that relies on daily data entry requiring about five minutes per day. The model assumes the presence of the bacteria in the orchard. It is up to the growers to determine how accurate this is and reminds growers that it only requires a few active cankers in the orchard to provide bacteria for a major epidemic under the right weather conditions.
This assumption will sometimes overestimate the risk prediction and predict infections that do not actually occur. It is unlikely that an infection of blossoms will result when low risk is predicted, unless other factors such as high winds or hail that result in "trauma" blight also occur. This model, like all other models estimates the risk of blossom infection. It should be used in combination with weather forecasts to project risk potential over the next couple days. If the model predicts a "high" risk of infection, the forecasted temperatures are critical to determine the potential of blossom blight infection. This model will also predict the appearance of symptoms of canker blight, blossom blight and shoot blight to help monitor infections that need to be removed before further spread of the disease.
Maryblyt predicts potential risk of blossom blight infection by Ea based on the occurrence of certain conditions in sequence:
Maryblyt predicts a low risk of infection when only one of the conditions is met, moderate when two of the conditions are met, and high when three are met. It predicts a blossom blight infection when all four conditions occur.
Cougarblight
This model was developed during the 1980s and has been the standard fireblight risk model in Washington and Oregon throughout the 1990s. The model is simple to learn and use, requires no computer or other special equipment, and is freely distributed to users.
The model requires the user to recognize specific and ever-changing local events and aspects of their orchard that may increase or decrease fireblight risk relative to other orchards in the region. The model requires the user to assume there is risk of fireblight infection whenever blossoms are present on the trees, especially during the petal fall and post-bloom period, when scattered blossoms may remain on many apple and pear varieties. The model user is asked to carefully assess the situation on their specific site and to initiate control measures if blossoms are present, risk levels are "high" or "extreme," and blossom wetting is likely to occur sometime during the next 24 hours.
Cougarblight has three major considerations: temperature, wetness, and local conditions.
Temperatures and wetness: The Cougarblight model estimates bacterial growth rate with degree hours based on a specific growth rate curve. This growth curve is based on the growth rate of Ea bacteria at various temperatures in laboratory tests. The degree hour values are accumulated each hour of the day that temperatures rise above 60°F. The hourly values increase as temperatures rise from 60 to the mid-80s, decline at higher temperatures, and reach zero for any hour with temperatures over 105°F. The model user does not actually calculate these hourly values, but uses daily high and low temperatures and a look-up chart to assign daily degree hour values. Estimated degree hour values on the look-up chart and actual degree hour values are similar, and repeated tests have shown no practical differences in risk ratings when automated monitoring systems have provided actual daily degree hour values to the model user.
The Cougarblight model user sums the degree hour values of the four days leading up to the time of potential blossom wetting to evaluate the degree of infection risk. Each day during the blossom period will have a new four-day degree hour total. Risk potential in the near future may be evaluated with forecasted temperatures.
In the Pacific Northwest, blossom wetting while degree hour totals are below 300 rarely results in blight, except on flowers very near active cankers. Wetting while degree hour totals are between 300 and 500 may lead to light, scattered fireblight outbreaks. These outbreaks of blossom blight are usually in orchards with near-by active carry-over cankers. Fireblight damage increases in the region as degree hour totals rise over 500, and severe, widespread damage can be expected if degree hour totals are near 800 when blossoms are wetted.
Local conditions: Growers may set personal risk thresholds at lower than those recommended on the model if the site in question has higher than usual risk of fireblight damage. The block may be young, an especially fireblight sensitive variety, on blight sensitive rootstocks, have high flower numbers, and be in an area that seems to be fireblight prone.
Use of the model: Starting when blossoms first open in the orchard, write down the degree hour value indicated on the look-up table for the known high and low temperature the prior day, and the predicted high and low for the current day. Write the next three days degree hour values based on temperature forecasts. Then give each day a value based on the current day plus the three previous days. Count only those days where open blossoms are present. Update the numbers daily. Spray appropriate materials as the degree hour totals approach "high" risk. Maintain control until degree hour totals drop below your chosen threshold, or blooming stops. If degree hour totals are well above the 500 threshold, nearing or over the "extreme" risk numbers, and flowers are numerous in the orchard, carry out all possible control measures to their maximum degree.
The model will most often indicate that sprays are not necessary. If you find that you are spraying very often when infection conditions are not relatively obvious, you are probably using a too-low personal risk threshold.
Discussion
To say one model is more conservative than the other requires a definition of conservative. Does the model overestimate the risk and require you to spray antibiotics more often, just in case? Or does it err on the side of limiting streptomycin applications to prevent resistance of the bacteria to streptomycin?
Experience with both models shows correlation in both models when we are at low to moderate risk. The difference lies in the "high" risk prediction, where in Maryblyt, it means there are three of the four requirements satisfied for an infection, and the missing factor may be either insufficient heat units accumulated or no wetting.
In Cougarblight, the "high" and "extreme" risk prediction is made when there are open blossoms and the heat units are accumulated. Infection is assumed to take place if there is a wetting period while the risk is "high" or "extreme."
In 1997, in blocks where fireblight cankers were present in 1996, five days were "high" risk days according to Maryblyt, but because the missing condition for infection was sufficient heat units, they were only rated as "low to moderate" risk using Cougarblight.
There were three days when Maryblyt predicted a high risk with the missing condition being a wetting event. On those days, Cougarblight predicted a "high" or "extreme" risk and would have recommended a streptomycin application. In these cases the CDH calculated by Maryblyt was between 100-312, and Cougarblight accumulated 391-503 CDH during the time period. The models appear to be well correlated with each other when temperature thresholds for bacterial growth are exceeded.
In 1998, Maryblyt predicted four "high" risk days, all of which were missing the required heat units, and Cougarblight rated those same days in the "low to moderate" risk level. Only late in bloom on late blooming varieties did enough heat units accumulate to result in a high-risk prediction by Cougarblight. Fireblight blossom blight did occur in those late blooming varieties at about a 5% infection level. Early blooming varieties with predicted "high" risk periods but with low CDH had no blossom blight where streptomycin was not applied.
Models are guidelines that should be part of our management scheme to be integrated in other parts of production systems. Using Maryblyt has certainly helped growers in western New York obtain greater control of blossom blight by fine-tuning the timing of streptomycin applications and has increased grower awareness of when symptoms will occur to help reduce spread of the disease in the summer. Because we have been limiting streptomycin applications to as few as one and occasionally as many as three per season, the Maryblyt model has been helping us with our resistance management program for antibiotics. Resistance to streptomycin is well established in most other apple growing regions of the U.S., but still no significant resistance levels have been detected in New York to date.
Limitations
There are limitations to all models because of the limitations in implementing control strategies. The variability we see in results of blossom blight control using antibiotics is a combination of several factors, including
Streptomycin is effective by stopping multiplication of the bacteria. It does not kill the bacteria. Streptomycin can only do its work when it is absorbed by the plant parts, such as open blossoms, shoots or wounds caused by hail or wind. Therefore, application timing and coverage to the day is critical. The model relies on accuracy of information gathered - remember the phrase "garbage in, garbage out." A weather monitor only reports weather conditions that exist in the spot it is located in. Variability of weather conditions in different parts of the farm can produce those surprise outbreaks such as in the low spots where the dew has settled, or there is later bloom than the rest of the block. And finally, you need to use a range of conditions in temperatures when forecasting potential risk for a couple days ahead. Predict the range, plan for the worst, and be ready to go if it happens.
But in the absence of any models, we can only be uncertain, apply costly controls more than necessary, or miss critical control timing only to result in epidemics and economic tree loss. With high-density, highly susceptible plantings being the trend in the apple industry, we need to use all the tools we have to maintain a minimum fireblight incidence.
