set up a treatment profile. think of the profile as a template ie I want to use the following chemicals on this specific type of crop. And when it comes to applying the chemical(s) record in a similar table but for a specific instance (field/location). the template could include a profile eg n grams per M2, say as per the manufacturers recommendation. and your table containing details of the actual treatments applied could record what you did use. when it comes to recording the information you have a table that contains the field or location and the type of crop and then from that you limit what treatments are available. In principle the actual seeding of the crop could be a treatment, and arguably so could the prep of the field as well as harvesting.
and so on. note the above is a 'back oif fag packet' sketch NOT a design model that is refined, designed and proven. its to give soem ideas, it is not neccessarily 'the' way to do it, a 'right' way to do it. its just an approachthat may go partway to reaching yourgoal, and offered more in hoep that it may make you think about the problem
look up intersection tables
..an intersection table provides a means of implementing a many to many relationship (a specific treatment (chemical) may be used on one or more crops, and a crop may may be treated with one or more chemicals
the key issue that I see is finding a way to constrain what treatments can be applied to what crops, thats the 'template' bit. think of it as a 'recipie' ie Monsanto to recommends the following chemical treatment plans...
chemical A at drilling
chemical B at seeding
chemical C at 6 weeks after germination
chemical D if insect infestations
...and so on....
so once you have identified that a specific location (up to you how you model that, it could be a GPS co-ordiante, a boundign box of co-ordinates or just a field / geographical name). you identify the crop for each location that year
at this point the nice theoretical model falls apart as you match the theory (ie manufacturers recommendation for each crop) with reality, IE what treatments you actually applied to each location. (say you decide partway through the growign season there is insufficienrt nitrate, so you could apply another treatment of nitrates, or there is a local infestation of weeds, so another treatment of roundup is called for.
so you need a mechanism to store what actually happened
I'd rather be riding on the Tiger 800 or the Norton