ebimodeling / empirical_yield_model Goto Github PK

@phenolphtalein

How might you answer the following ...

Given the Miscanthus and Switchgrass yield data that @sidewallme has used in his analysis, plus any other data in the yields table, determine:

Are there species that haven't been grown in agronomic trials (i.e. we don't have yield data in betydb.org/yields.csv) that might be good candidates for high yielding crops?

Notes:

• in the plants table (betydb.org/species.csv) is full of categorical variables - for all species in plants.usda.gov
• there are also many sparse rows that we have created, mostly for poplar and willow hybrids (bred across species) and non-US plants.

• rename tree.Rmd to methods-and-results.Rmd, or else move content to manuscript.Rmd
• code updated to use both swithcgrass and miscanthus
• update the methods section
• update the results section
• propose new analyses as github issues

Define models, method of evaluation (e.g. regression tree, ANN, linear regression), and model skill metrics (e.g. RMSE, r2, sd_model/sd_mean, Taylor Skill, etc.)

For any data set you want to use, either add to this repository or post as a gist.

• Sugarcane (NASS?)
• Multicrop
• Miscanthus https://gist.github.com/dlebauer/0ce86601f1c4ddedfdc6

@sidewallme

Download each fof the 19 BioClim variables listed here http://www.worldclim.org/bioclim using the getData function of the raster package

other variables: http://www.worldclim.org/formats

library(raster)
x <- getData('worldclim', var='tmin', res=0.5, lon=5, lat=45)

note that bioclim are long-term climate means; also would be useful to find precip. and temp during harvest year (e.g. 'year' field in miscanthus.csv from #2).

Try out this script for downloading dayment: https://bitbucket.org/khufkens/daymetr