Monday, January 14, 2019

MFGA meeting and Malusim app

Last Thursday, January 10, 2019 the Massachusetts Fruit Growers' Association met for their Annual Meeting at the Great Wolf Lodge in Fitchburg, MA. The meeting program included presentations by UMass Extension faculty and staff as well as Dan Donahue from Cornell's Eastern New York Commercial Horticulture Program.

MFGA Annual Meeting, 10-Jan 2019, Great Wolf Lodge, Fitchburg, MA
But, as a result of some feedback, I wanted to highlight one of my presentations, "Precision thinning using the Malusim app: trials and tribulations." I am going to follow with the individual "slides" and what I should have said during my presentation, but of course then I did not have time to smooth it out like I will here. Not to mention I only got 15 minutes. So here goes...

slide 1
Slide 1 - Today I want to talk to you about the Malusim app and how I used it (hence trials and tribulations) at the UMass Orchard in 2018 to practice "precision thinning."

slide 2
Slide 2 - So exactly what is precision thinning? It uses the fruit growth rate model co-developed by Duane Greene at UMass, and Alan Lasko and Terence Robinson at Cornell University to help predict if chemical thinners have been effective and thus achieve a target crop load per apple tree. AKA predicting fruit set. (See: A Method to Predict Chemical Thinner Response in Apples.) You can see the required steps here, which include: 
  • Determine desired crop load
  • Count flower clusters at bloom (7 trees per variety per orchard block)
  • Tag and mark fruitlets at about 5 to 6 mm (7 trees times 15 spurs times 5 fruits per spur equals 525 fruits)
  • Begin measuring each fruit with a caliper and record results, keep track of each fruit with each measurement, go home and enter into Excel spreadsheet 
  • Spray thinner, repeat above (several times) until desired crop load achieved (number of fruit or % fruit set), and additional thinners (if necessary) have been applied.
Are we having fun yet? Is anyone actually doing this?

slide 3
Slide 3 - If you really do want to use the predicting fruit set procedure, it is all outlined here in a 7 page document: Just keep in mind too it really needs to be done on every variety in every block! I ask again: are we having fun yet? Is anyone out there actually doing this?

slide 4
Slide 4 - Enter the Malusim app. An app developed by Poliana Francescatto and co-workers at Cornell University to help put precision thinning/predicting fruit set and the in the palm of your hands. Literally. First note that the Malusim app works in your browser where you create an account an set up your orchard blocks, as can be seen here in a browser window. As for the iOS and Android apps, we have been beta testing them but the app should be available to download this spring in the respective app stores. Note that Malusim also includes an Irrigation Model and has the ability to keep chemical thinner spray and irrigation application records.

slide 5
Slide 5 - So what did I do this year? Note that I have experience predicting fruit set and have wrote two articles on highlighting my results. But this year, I used the Malusim app to (hopefully) facilitate the process. Which included:
  • Selecting six (6) varieties in six different blocks: Pazazz, Fuji, Gala, Honeycrisp, McIntosh, and Empire
  • All trees were dwarf, tall-spindle (more-or-less) except Empire (slender-spindle)
  • Five trees were selected, five (only five!) spurs per tree selected and measured on 4 measurement dates
  • All data entered using Malusim app on Android (Google) phone using (the experimental) voice input
  • Only a petal fall thinning application was applied: NAA (or Maxcel) plus carbaryl

slide 6
Slide 6 - All fruit measurement data was entered using the Malusim app on a mobile device either using voice input or the device keyboard. I want to say it was easy, so simple even UMass Stockbridge School of Agriculture undergraduate students could do it! (Thanks Cam and Lindsey!)

slide 7
Slide 7 - Here are some screen shots using the phone app. On the left, the Locations menu in the ?. Middle, fruit diameter data entry screen, including the voice input icon - press and your are prompted to speak the measurement which is automatically entered into the Fruitlet # field. And last on right, the results showing target fruit number and % fruit set, and predicted set (number and %) based on each measurement date. At bottom of this screen you can also see when the chemical thinning spray was applied, also indicated by the vertical gray line on chart (to left).

slide 8
Slide 8 - and here, for example, is the interface (browser window) you would get when logging into from your computer. When logged into your account, the app completely displays and syncs across phone, tablet, and computer.

slide 9
Slide 9 - Now, let's look at the results. First up Pazazz, a managed variety grown in WA, MN, WI, NY, and Nova Scotia. These are screen captures from the browser window. At the bottom, you can see when the thinning spray was applied, which is also depicted by the vertical gray line on the left of the chart. You can also see there the Potential fruit per tree and the Target fruit per tree (number of fruit and % set). These are determined when the block is initially set up, after counting the flower clusters and deciding how many fruit per tree is wanted. Then, four fruit measurements were made, with predicted fruit set shown by the blue bars beginning/after the second measurement date on 30-May. At the top, I counted the number of fruit per tree left at harvest, averaged across the 5 trees. So, on the last measurement date (8-Jun) predicted fruit set was 159 fruits per tree. Actual fruit per tree at harvest was 117. Humph. Definitively over-cropped were these trees, which affected quality -- look, I only wanted 50 fruit per tree on these smallish Bud.9 rootstocks. That's how it works. Now let's look how the rest of the varieties worked out?

slide 10
Slide 10 - Next up, Fuji. Target 80 fruit. Predicted (after last meausurement) 137 fruit. Actual, 125. Close, but over-cropped.

slide 11
Slide 11 - Gala. These Gala trees were a bit odd, with variable bloom and final crop. So, not putting much stock in it, but Target 65, Predicted 50, Actual 61. Not bad. I have come to the conclusion you almost need to be just below your target upon the final measurement to come up right.

slide 12
Slide 12 - Honeycrisp. We love to hate it. Target 50 fruit, predicted 90 apples, harvest 101 apples. Ugh. Way over-cropped = lousy tasting Honeycrisp. More chemical thinner should have been applied, need to see that final blue bar BELOW the pink shade?

slide 13
Slide 13 - McIntosh. 100 fruit target. 114 fruit predicted. 160 apples at harvest. What? McIntosh are different. But small Macs are good, right? Not sure I would waste my time doing this on Macs, which habitually crop every year and did I say small Macs aren't a bad thing? But do they make any money? 😟

slide 14
Slide 14 - Empire, I add here only because there is a side story (in two slides). This was not my experiment you will see. I don't have much to say here, except wait for slide 16.

slide 15
Slide 15 - So, the general tendency is for the trees to be over-cropped at harvest, and that has a tendency to be indicated by the predicting fruit set results. All is good on that front, but you have to wonder why too many fruit? Don't forget the Malusim app can also display the results of the Carbohydrate (CHO) Balance model at your orchard. Wait a minute, that is if you have a NEWA site because the app pulls the results of the CHO model real-time from NEWA. Note here the CHO balance is not particularly severe when the chemical thinners were applied. Therefore, one would expect modest chemical thinning (at best). Another/more chemical thinners should have been applied.

slide 16
Slide 16 - Now the Empire story, here Paul O'Connor, UMass PhD student works with a technician from Carnegie Mellon University at the UMass Orchard to visualize fruitlet growth using a hyperspectral(?) camera. Over the course of a week they took several visualizations of fruit growth using this special (and expensive and heavy) camera with the goal of seeing if indeed growth rate can be visualized and calculated based on these images. Indeed, preliminary results suggest that this is the case, and last I heard, they are working to see if the same growth rate learning model can be applied ot images taken with, shall I say it, a smart phone? An idea I have had for quite some time...😲

slide 17
Slide 17 - Conclusions. The Malusim app has the potential to make the job of precision thinning and predicting fruit set notably easier. One needs to be a bit of a technophobe, however, and it's not for everyone. Plus it's kind of in beta. (Cornell, please figure out the future path of Malusim.) Don't forget, however, there is a whole irrigation model built-in too. I encourage you to go to, sign-up with an account, and give it a try in 2019. I will say at the very least you will learn a lot by going out and looking at a small sub-set of your growing (ot not growing) fruitlets, which will make you "seat of the pants" chemical thinning decisions a little less so. Good luck and feel free to submit feedback to the Malusim team, whoever that is...