Sunday, February 4, 2024

RIMpro – to spray or not to spray using ‘virtual’ vs. weather station weather data

 RIMpro from their website (rimpro.cloud) “is a decision support system (DSS) for the sustainable management of pests and diseases in fruit and grape production. Every day, the cloud service together with the weather data system help thousands of growers and consultants worldwide to make the best decisions to protect their crops.” 

In 2023 I subscribed to RIMpro using both a hardware based weather station connected to NEWA (newa.cornell.edu, as Belchertown-2) and their virtual weather data service Meteoblue (meteoblue.com). My intent for using both was to be able to make a comparison of the weather station vs. virtual weather data ‘after the fact’ in terms of making spray decisions to manage apple scab, fire blight, and codling moth. It’s obvious the two sources of weather data  are going to differ, but that was not the point in making these comparisons. So I won’t belabor that, but I will discuss how using the two different sources of weather data may have made an impact on the decision-making process to “spray or not to spray” to manage the above-mentioned pests. The answer, of course, and as always, is “it depends.” Realize also this comparison was made at the end of the growing season when all the weather data was “in.” No attempt made to compare the forecasts in the midst of the growing season. (That’s a whole nother story.)

Apple scab

I just looked at primary scab using 1-April as a green tip date. Weather data that factor into the scab model include temperature (I presume but it does not show up in the RIMpro data file?), rainfall (to trigger spore release), and leaf wetness (duration). Looking at timing of spore maturity and last available spores to cause infection, there is not much difference there in the RIMpro chard (Figures 1 and 2) with an end date of approximately 10-June. I don’t care so much about that as long as they are close, but what I do care about is the RIM value(s) which are a measure of infection risk, being low, medium, or high. In Figures 1 and 2 I placed the arrows where I figured, based on what I know about RIMpro and primary scab management with fungicide sprays of some sort -- be it protectant, or protectant plus kickback fungicide – would have been required to manage a primary apple scab infection. Count how many fungicide spray black arrows there are in either plain old UMass Orchard or UMass Orchard-MB (MB = Meteoblue, virtual weather data). I count seven for each weather data source. No difference in number of sprays, right? (I'm sure my assumptions about spray timing will be argued.) But, what bothers me a bit, is it appears the RIM value is much higher for MB in several infection “events.” Looking at the raw weather data RIMpro uses, it’s clear that MB  “overestimated” by a factor of approximately two times both rainfall and leaf wetness compared to the weather station. “Overestimated” is assuming the weather station is correct? In particular, it’s likely the leaf wetness hours were the primary contributor to the difference in magnitude (as measured by the RIM value) of the infection events. Begs the question, which is correct, the leaf wetness sensor on the weather station – and we know there are issues there – or the modeled leaf wetness by MB? Good question. (Here's more on how RIMpro models leaf wetness duration for virtual weather station.) Just look at the duration of light blue which signifies leaf wetness in the weather data charts (Figures 3 and 4). What’s my inclination here? Well, looks like the MB output is a bit more conservative in the approach to manage apple scab = better safe than sorry? Know what I mean? Although not affecting the timing of fungicide sprays, the rate of fungicide might be adjusted upwards? I can see the plant pathologists out there shuffling in their seats.

Figure 1 - apple scab model output using weather station data

Figure 2 - apple scab model output using MB virtual weather data

Figure 3 - weather data chart using weather station

Figure 4 - weather data chart using MB virtual weather data. Note the extended periods of wetness (light blue) compared to Figure 3

Fire blight

I am no fire blight expert. We did have a significant fire blight outbreak in 2023 at the UMass Orchard. I don’t quite understand why MB showed that two more infection thresholds were reached compared to the weather station? Using the MB approach, three streptomycin sprays should have been applied vs. one for the weather station  (Figures 5 and 6). Which do you think would have been the better defense? Well, after the fact, and given our observed fire blight outbreak by early summer, following MB might have been the better strategy? I note that the prediction of visible symptoms was pretty much right on in RIMpro, as fire blight was indeed seen on June 5 (Figure 7).

Figure 5 - fire blight model output using weather station weather data. Date of strep spray is estimated
 
Figure 6 - fire blight model output using MB virtual weather data. Dates of strep spray are estimated

Figure 7 - fire blight symptoms observed 5-June, 2024 at UMass Orchard

Codling moth

I am no entomologist. But I do know something about codling moth management based on adult flight, mating, and larvae hatch, the latter being the ideal time to control codling moth with a targeted insecticide. So, looking at both outputs I see little difference in suggested insecticide timing between the two sources of weather data (Figures 8 and 9). Likely because the model is largely based on temperature, and we know that virtual weather data is pretty good at predicting/adjusting temperature such that over time the two sources of weather data end up being pretty closely aligned. That is about all I have to say about codling moth, I’d feel comfortable using MB there. And you have to love RIMpro’s approach of tagging ‘virgin’ females and ‘mated’ females. Maybe I am the only one that finds that amusing?

Figure 8 - codling moth model output using weather station weather data

Figure 9 - codling moth model output using MB virtual weather station data

A final note about Meteoblue, you can subsctibe to a free daily email from them that gives you a succinct graphical image of the weather forecast specific to your location. I get it daily, and have found the forecast to be about as good as anything else, although far from the perfect forecast. I particularly like their graphic which gives you a one-shot-in-one-place-in-your-face picture of the forecast as seen in Figure 10.

Figure 10 - Meteoblue Meteogram forecast for Belchertown, MA on 19-May, 2023 

So, although RIMpro is not for everyone as it's a bit dense unless you are really into it, I do highly recommend it as a 'precision' pest management tool. RIMpro has other models too numerous to mention here, but the stone fruit brown rot model might be worth mentioning. And, it can be site-specific (anywhere in the world) using the MB virtual weather data. Colleague Srdjan Acimovic at Virginia Tech University is a big proponent of RIMpro, and he is hosting a webinar about RIMpro on 20-February, 2024. I will be there. Rimpro also just launched on the App Store for iPhone and iPad...



No comments: