confirm data cleaning choices

[sdtaylor]

All the long term datasets have various protocols which required decisions on when to assign budburst/flowering. Need to go thru all this again and document what I did for methods and confirm that decisions make logical sense and match the NPN data as closely as possible.

[sdtaylor]

Harvard

Budburst and flower doy are when the percentage of each (which are observations for a single tree) are >= 10

HJ Andrews

Bud break is the first doy when they were marked as "bud break". Flowering was the first doy that was marked "Flowers open". The raw data has codes for each species, see the document `raw_data/hjandrews/phenophase_info.pdf' for info on them.

Jornada

Observations are, for each zone (essentially a transect), the percent of plants in each phenophase. I made budbreak the first day where <90% of plants are observed as dormant, and flowers the first day >10% of flowers are seen.

Hubbard brook

The data is the average, among 3 trees, of some ordinal descriptions of phenophase

0: winter conditions
1: bud swelling
2: small leaves or flowers
3: leafs 1/2 of final length, leafs obscure half the sky as seen thru crown
3.5: leaves 3/4 expanded, sky mostly obscured, crown not yet in summer condition
4: fully expanded, canopy in summer conditions

For budburst I chose a cutoff of >=1.6. This is where the most likely values for the 3 trees are c(1,2,2).

[sdtaylor]

#25