-
Notifications
You must be signed in to change notification settings - Fork 3
Data Model: Activity Type
As previously stated, activities are the heart of Business Process Modeling (BPM) or Data Warehouse modeling. This is a fractal pattern where an activity is made up of many smaller activities. A critical part of modeling is determining the boundaries of what to capture and what to leave alone; what is conceptually and logically there, but not worth the time and effort (or interest) to capture. In a large commercial agriculture operation there would be many details like maintenance records, employee time charting and energy consumption that are critical to the profit or loss of the business. However, with a small growth chamber; while there is maintenance and energy consumption, these are things that are usually not significant to capture and hence are left out of the database and data model.
BPM comes in three 'flavors': templates, plans and actions. Templates define abstract actions. They are the 'recipe' of how to bake a batch of cookies, not the actual action of baking real eatable cookies. Plan is the proposed realization of a recipe; it is the schedule of when the action will take place, and identifying the physical ingredients that will be used. Action is the execution of the plan, what was actually done, by whom and with what. For the MVP, we will focus on capturing actions, and secondarily look at trial recipes. Planning and scheduling are out of scope for the database.
Agriculture (focusing on annual plants) centers around the life-cycle of the plant, from planting of the seed through harvest. The life of the plant can be sub-divided into smaller cycles of germination, vegetative growth, flowering and fruiting. These life-cycle events are usually captured as phenotype observations. Accompanying these plant cycles are agronomic activities, things done by the grower to the plant and its environment. Many of these activities are planned and scheduled, while others are done in response to changes in the plant or environment. These activities can include:
- Environment preparation
- Planting
- Thinning
- Watering (Irrigation)
- Fertilizer application
- Insecticide application
- Light application
- Chemical adjustment (adjusting pH)
- Harvest
- Clean-up
These activities are usually manually recorded, or created from derived data (see Actuator Activities).
Actuator activities are similar to sensor readings, but different; they are both devices controlled by the 'brain' (Raspberry Pi), but used for different purposes. If environmental observations (sensor readings) are recording what the environment is doing to us (or the plants), actuator activities are what we do to the environment. These are the detail records that are often used to determine larger agronomic activities. An actuator activity it the change of state of a device; the turning on or off of a light, the opening or closing of a solenoid valve. By knowing when a light turned on and off, we can determine the photo-period, and knowing the properties of the lights we can determine the total PAR (light radiation) applied to the plants. Knowing when a valve opened and closed (along with the flow rate) allows us to calculate the amount of fertilizer or water added to the reservoir. From a modeling perspective, the actuator is both the subject and the participant (as it is often reporting about itself), the state is the attribute, and the value is the state change (ON or OFF).
Observations are notes about the plants and the environment. These activities may be scheduled around the plant life-cycle, or they may be done ad-hoc. Observations fall into two main groups, those that are of the plant (Phenotypic Observations) and those of the plant environment (Environment Observations).
Phenotypic observations are recordings about a plant, or sometimes a population of plants. For the purpose of the MVP it will be assumed these are individual plants. We want to avoid problems of population/mass observations. Generally the subject will be the plant, with attributes being things like height, width, length or weight. From my experience, this category often becomes the most confused and difficult areas of consistent data collection. Once we get beyond the initial basic categories, things get complicated, especially when you want to compare and aggregate data records that were not carefully thought out or designed to be used beyond the initial experiment. It is hard naming phenotypic characteristics. I remember subjects such as: 'Fruit color three days after harvest', 'Number of mature insects between the first and third node (of the plant)', 'percentage of flowering three days after initial flowering'. There is a plant ontology which is a part of the OBO ontology effort, but even with all the work of many organization it has no formal approval due to its flaws. While we may not want to use OBO ontology identifiers, we should at least be aware of them (they are good for identifying plant parts) and proceed carefully - beyond here be dragons. If you think this will be easy, try mapping the growth stages of corn and soybeans, then add in strawberries. Phenotypic observations are tied to a particular plant or plot (thus a location identifier), and will be tied to a specific trial and experiment.
Environment observations are independent of the Experiment, Trial or live cycle of the plant; though correlated to them for analysis. I think of this in terms of weather data. The National Weather Service (NOAA) makes recordings of weather and publishes daily summaries (min, max, average temperature; min, max, average wind speed; humidity; cloudiness; amount of sun; amount of precipitation; precipitation type; sun rise, sun set). This data is not made for any one group of people (farmers, sports stadiums, ...) but exists as an independent set of data. People of all interests then access this data for different needs: farmers plan their planting and harvest (and irrigation); individuals plan their vacations, golf games and whether to take an umbrella to work; pilots determine their flight plans, etc. In a similar manner I see the sensor data of the MVP being of a similar nature. I can collect sensor data regardless of whether there is anything growing in my box or not. As long as there is a timestamp on the observation, I can correlate it to the time and dates of an experiment. There are two main subjects of environmental observations:
- Air
- Water (or nutrient solution)
Temperature, Humidity, percent CO2 are all attributes of the air. Water has attributes such as temperature, pH, EC and any other chemical properties. Both of these have the problem of being mass units.
- Experiment and Trial
- Agronomic Activities
- Actuator State
- Environment Observation
- Phenotype Observation
2018/05/31 webbhm