Defining Relevant Beekeeping Regions

We’ve been thinking of changing up the arbitrary HiveCheck reporting regions based on feedback from respondents, and our own pressure to make more significant beekeeping “zones” that have both climate and cultural relatedness specifically for beekeepers. Last week we asked our HiveCheck respondents to give us a hand in this pretty important task and this post details the process.

Step 1) A Starting Point

For any new project it’s important to get your hypothesis down on paper to test against in later exercises. So where to start? We needed a basic grouping of states that had enough relevant attributes to beekeepers to qualify them for a “region”. The most basic, and probably the most important, attribute to look at is climate so we started there.

Heading over to the USDA’s Revised Plant Hardiness Zone Map we can get a pretty clear picture of the high and low extremes across the U.S.. Interestingly, the revised version of the map from 2012 is the first revision since 1990 and there are some stark differences.

The temperature data from the 2012 Plant Hardiness Map has been taken from more recent years (1976 – 2005) and the map itself is entirely GIS which includes higher resolution and better accuracy. Of course climate change may be a factor but the methods used in generating the Hardiness Zone Map most likely is not the best tool to measure that.

None the less, this proved as a great starting point and we eyeballed basic color distributions in each state to make the initial groupings.

We also wanted to take into consideration Ayers & Harman’s Honey Bee Forage Map from 1992. This map is broken into 14 regions which is primarily based on forage types in each region which we used to compare states within the same climate zone to help make some tougher decisions. We did not use this data across climate regions since states like Montana and Texas may be in the same foraging zone but are certainly worlds apart when it comes to seasons.

Here’s what we came up with to start:

New England: ME, NH, VT, MA, CT, NY, RI

Mid Atlantic: PA, DE, NJ, MD, VA, WV

Souther Region: NC, SC, GA, FL, AL, LA, MS, TX

Western Tier: WA, OR, ID, CA, NV, UT, AZ, NM

Middle Interior: OH, KY, TN, IN, IL, MO, KS, AR, OK

North Mid West: MT, WY, CO, ND, SD, NE, MN, IA, WI, MI


Step 2) Test It

So how do we know if we’re on the right track? Last week our HiveCheck respondents were asked “Which states other than your own are you most interested in seeing data from?”. As a result we got more 400 recommendations for states with a significant relation to a respondent’s own state. We simply ran each pair (respondents state and requested state) across our initial state groupings to find how much we were in agreement with over 280 beekeepers. If both the respondent’s state and the requested state were in the same region we considered that an agreement, anything else was a disagreement.

Initial Agreement Score: 79%
80/20 consistently applies to all things and numbers, it never ceases to amaze me… We can certainly strive for better though, to improve the results we compared the frequency of non-matching state pairs to see if there were any specific patterns.

These were the top states that respondents asked for that were not in a matching region. Note (We only looked at states with 3 or more hits)

Respondent’s State: { Requested States}
NC : {‘TN’: 9, VA’: 10,}
VA : {‘NC’: 7, ‘MD’: 3}
NY : {‘PA’: 6, ‘NJ’: 4}
TN : {‘NC’: 3, ‘MS’: 4, ‘AL’: 4, ‘GA’: 3}

It was fun to see some states really liked each other like NC/VA, but what about the states who’d be lonely if we moved a state from one region to another? Luckily we had the statistical model so we cold simply move states around and see how are score changed. In theory, if we moved a state out of a group where it was already popular our agreement score would go down. If we made a better match the score would go up.

Here’s what we found

Moving Virginia from  Mid Atlantic to Southern Tier : +2% Score
Moving New York from New England to Mid Atlantic: + 2% Score
Moving Tennessee from Mid Interior to Southern: +5%

Tennessee was a big surprise, considering it was not as polarized as NC or VA. I think this data really shows the cultural ties between states which is an important factor in making meaningful beekeeping regions. I’m happy to say our final agreement score was 88%, not bad at all.

What’s Next

Coming up, we’ll be implementing these new regions in our HiveCheck reporting starting this week. We’ll also be reviewing comments from our respondents and the public about the changes to continue getting feedback. Feel free to leave comments on this post and let us know what you think!

Written By: Alex Jones

Michael Wilson has written 13 post in this blog.