jmcextman: PACMAN: Malusim vs. Fruit Growth model app vs. Farm Vision

Precision Apple Crop load MANagement (PACMAN) is the topic of our time. Seriously, it is. Although I have to be a jack of all trades (pest management included) my heart lies in horticulture. And technology. (Although I am no computer geek like some people think I am!) So, I have been interested in the fruitlet growth rate model per Duane Greene et al for quite some time including onerous measuring of fruitlets for quite a few years.

The fruitlet growth rate model is one part of PACMAN which also includes precision pruning, blossom thinning with caustics, chemical fruitlet thinning, and finally – if necessary – hand thinning. Why? Because, particularly when growing high value varieties like Honeycrisp – or where size matters even more, like Gala – it’s money you are otherwise leaving on the table. (Thanks Rod Farrow for that bit of wisdom).

So, carry on, here is what I did for PACMAN in 2022. My main objective was to compare the fruitlet growth rate model when measurements were put into Malusim (MS app) vs. the Ferri-Fruit Growth model ‘FG’ app. In addition, and to make it more complicated of course, Farm Vision (FV) was in the mix. (And I later realized it’s a bit like comparing apples to oranges though. More on that later.) I’m going to lay it out here pretty tersely for you to digest and form your own opinion on the value of using the fruitlet growth rate model as part of your PACMAN strategy going forward. Which I do recommend, with caveats as you will see.

METHOD

At the UMass Orchard (UMO) I chose Gala, Fuji, and Honeycrisp. Per my RECIPE for predicting fruit set, five representative trees were selected at bloom, number of flower clusters counted on each tree (for potential fruit set), fourteen flower clusters tagged per each of the five trees for fruitlet measurements beginning when fruitlets were app. 6-7 mm in size on 23-May and continuing subsequently on 26-May, 29-May, and lastly on 3-June, 2022. Unfortunately, although chemical thinners were applied, the details are sketchy and date of application(s) not included here for UMO.

Ditto for Tougas Family Farm (TFF) Northboro, MA for Gala and Honeycrisp, although over only three measurement dates: 21-May, 25-May, and 27-May, 2022. Here, chemical thinner application dates are noted in the results.

At both UMO and TFF, sometime between August and harvest the number of apples left on each tree were counted for the final fruit set. All make sense so far? Oh wait, on each measurement date I also did a Farm Vision (FV) scan using their provided equipment: smart phone, stereo video camera, and enhanced GPS location. And also did a final FV fruit scan to determine the fruit set in August.

FV scan on 3-Jun, 2022 at UMO, Fuji

There, now the results are presented in both tabular and chart form (Figures 1. through 5.). Ought to be self-explanatory, but we will see 🙂 and more explanation will be included as it’s not always really pretty…

RESULTS

Figure 1. - Predicted fruit set Gala at UMO

A bit of explanation here which applies to all the Figures. There was one measurement date previous to all the initial measurement date(s) noted in Figure 1. (and Figures 2. through 5.). But that is not included because in Malusim there is no prediction. (No growth rate calculated on first measurement.) Otherwise, on the dates noted, those numbers in the tables are predicted fruit set (number of apples per tree) based on Malusim, the FGM app, and Farm Vision. Actual (count) in the Table was the number of apples hand counted (per tree) in mid-August. (Except for UMO Honeycrisp in mid-September.) Actual (FV scan) was what FV 'saw' on the trees mid-August. So you can compare that Actual fruit set to the Predicted fruit set on the last date of fruitlet measurements. Charts are table data visualized. All make sense now?

Gala-specific at UMO (Figure 1.), these trees are on M.9 rootstock, fully mature, tall-spindle. Bloom was spotty but five trees with approximate equal (and good) bloom density were selected. The overall objective here was to compare Malusim and FGM app when the same numbers were used. You can see they differed a bit. Why? Measurements were made by different people (on at least one date), and that could have accounted for some 'increased' predicted fruit set noted on 29-May. (That should not happen!) Advice: same person should measure on each date to keep that source of measurement error down with more consistent measurement technique (hopefully). Farm Vision here was in the ball park. Oh, one more thing, not sure I am using the trend line appropriately, but I like it here...😉

OK, onward with the rest of the RESULTS (Figures 2. through 5.), with brief variety-location-specific comments.

Figure 2. - Predicted fruit set Fuji at UMO

Same story as Gala here with these Fuji (M.9 rootstock), with that uptick of predicted fruit set on 29-May, different measurement people, student interns? Malusim and FGM app were quite different at the end (3-Jun), FGM being right on compared to Actual (count). I should note that Malusim has instituted some error correction and will exclude measurements deemed to be outside specific error limits (too large, too fast growth rate). Not sure I like it because I think at least here it might have been discarding measurements that were indeed valid? Not sure, but perception is reality? Again, Farm Vision, at least on the last measurement date, predicted more fruit setting than Actual. Is this necessarily a bad thing as some fruit will continue to fall off, June drop, etc. I told you it was a bit ugly, but I am not casting any shadows yet after just one season of evaluation of Farm Vision. (BTW, they are now part of Meter Group/Pometa.)

Figure 3. - Predicted fruit set Honeycrisp at UMO

At least Malusim, FGM app, and FV align pretty good here with these Honeycrisp on G.11 rootstock in their 9th-leaf. Actual (count) may be an underestimate as there were some drops which were not counted. But not so many to skew the results too badly IMHO.

Figure 4. Predicted fruit set Gala at TFF

At TFF there were only three (total) measurement dates for both Gala (Figure 4.) and Honeycrisp (Figure 5.). Hence only two dates of Predicted fruit set. These Gala were mature tall-spindles on G. 41. I do have some thinning information here, thinners were applied to these trees on 12-May (Promalin + AmidThin, bloom) and 20-May (Maxcel). But I was not worried about using the Predicted fruit set to help with thinning times and rates, at least here, where the objective was to just compare what the three tools were doing in predicting the fruit set. Got it?

Figure 5. Predicted fruit set Honeycrisp at TFF

Thinning regimen here on these Honeycrisp/G.41 was NAA (4 oz.) on 12-May, NAA (4 oz.) plus carbaryl (one pint) on 18-May, and NAA (2 oz.) on 27-May. No Farm Vision here, the equipment was not behaving properly. (Except for the Actual FVT scan.) Malusim and FGM app were close, but not close enough. As an aside, bitter pit was really bad in this light cropped block of Honeycrisp on G.41. Did I say G.41? (Friends don’t let friends plant Honeycrisp on G.41.)

I have to say that 2022 was overall just plain weird, I mean the weather during the post-bloom period made chemical thinning applications a nail biting experience. What we saw and what happened kind of defied expectation. It was not a 'normal' year. (As if we ever have one?) Can't wait to see what 2023 brings... 😱

RECOMMENDATION

There you go, make of it what you will, there is lots more to dive into. But, I do know this is what I would do: follow the RECIPE using the FGM app. iPhone only unfortunately. Malusim has issues, although I do like the website interface. The app, not so much. But it does more than predict fruit set, and I hear it is being updated to include fixed voice input recognition in 2023. After the initial set up, it takes less than an hour (easily) to do the measurements (per variety per block) using the RECIPE instructions. Preferably with two people. And you get instant results via the FGM app, i.e. predicted fruit set to better inform further chemical thinning decisions. Period. One side effect often observed, is the actual physical process of measuring fruitlets gives better insight over time as to what is going on -- what is growing, and what is not -- that you otherwise get from simple observation from the truck window!

But, Farm Vision (now Meter Group/Pometa) has merits in its early stages. Plus it gives you a crop estimate later in the season for better harvest management. Lots of potential there, and it – or similar tools (FruitScout?) – are the future of precision apple crop load management. Well, then there are rovers and drones, but I don't go there, I let others. Flip-flop, check out Orchard Robotics. And here's a couple videos from Farm Vision you might get a kick out of. I did, reviewing them here in January 2023 when it is otherwise pretty bleak outside. Good luck in 2023...