As part of my decision modeling class, a group of us set out to apply the principles of modeling to an area that didn't immediately come to mind... a number of ideas came up ranging from efficient bike routes to stopping nuclear proliferation (although figured our emails back and forth may put us on a terrorist watch list and thus bailed on that idea). Instead we focused on producing a model that would tell us what plants to put where in a home garden using a cool color coded map - this is our presentation of results.
2. Home Gardening 31% of US households participate in food gardening (2008) Approximately 36 million households In 2009, 21% of food gardening households will be new to gardening. 11% of currently active households plan to increase both the amount and variety of vegetables they grow in 2009.
4. The Problem What should we plant and where in a basic backyard garden? Time horizon Size of the garden Number of plants that fit in a given plot Yield per plant Retail price of the particular plants Nitrogen fixing plants (minimum)
5. Time Horizon Considered five years Short enough that price estimates and user preferences are stable Long enough to offer sufficient planning benefits
6. Size of the Garden Attempted to match common garden sizes Median household garden size: 96 sq. ft. Set max 12’ x 12’ garden area (144 sq. ft.) Broke garden into 3’ x 3’ sections for planting 9sqft. plots User can customize garden size per above
7. Number of plants in a given plot Required plant spacing to determine number of particular plants that fit in a given plot Adjustable by the user Potential source of uncertainty
8. Yield per plant Compared various sources to find ranges Applied Pert probability distribution to estimate yields
9. Retail Price of Plant Researched local retail pricing of selected vegetables to include in the model Assumed prices are constant over the given time horizon Potential source of uncertainty Price changes will most likely be correlated
10. Nitrogen Fixing Plants Green Beans and Peas are Nitrogen-fixing plants Maintain the quality of the soil Nitrogen is necessary for plant growth Rotation of these plant types maintains a balance of Nitrogen in the soil.
11. Decision Variables Variable for each vegetable, each year and each plot: Gbi,j = Green Beans in Year i and Plot j Used 10 vegetables common for household gardening 800 decision variables 10 vegetables * 5 years * 16 plots
12. Objective Function Maximize the value of the yield over the five year horizon Expected output of each given vegetable Price of each vegetable MAX((Price-per-PoundGB * #-of-PlantsGB * Pounds-per-PlantGB) + … + (Price-per-PoundH* #-of-PlantsH * Pounds-per-PlantH))
13. Constraints Gb and Pe must be grown in each plot at some point during time horizon Each plot must be used in given year No fallow soil Maximum 1 herb plot each year Number of plants must be an integer User preferences for min/max of any given plant
14. Friendly User-Interface User can customize certain aspects of the model to match their own gardening criteria Garden size Estimated price of each vegetable Set a minimum and maximum preference for each vegetable
15. User Preferences User can define certain preferences to customize the model to fit their own garden.
16. Spreadsheet Model Decision variables grouped by plant in order of year. Plot numbers appear or disappear according to user settings
17. Yield Table A backend to the model handles how each plant is incorporated into the model Size of plants, expected yield and the appropriate retail values are pulled from the yield table
19. Planting the Seed Garden planning map Color coded plots indicate what type of plant goes in which plot Quick, easy and understandable reference Different map for each year
National Gardening Association: http://www.gardenresearch.com/index.php?q=show&id=3126
Gardening is already an interest for a large number of people… and its only growing.Figuring out what to plant and where is a question that a lot of people are asking. Being the innovative thinkers that we are, we thought this was ripe… ah… ah… for a decision model.
Median household garden: 96 sq. ft
Why Pert? Because HuybertGroenendaal said to use it. It provides a better, more realistic shape than the triangular probability distribution…?
Objective function is in terms of dollarsUsed dollars because it offered an objective means of quantifying the value of each plant
Price per pound of herbs was extremely high effected the weighting inside the objective function… no one wants only basil.
Assuming user has appropriate knowledge and software…