Spring break meeting - Scott, Sean, Tom present
- think about abiotic variables - should they change over time? e.g., snowdepth
- our variables interact with each other (slope + aspect affects snowmelt, e.g.)
- use plot-level variation (spatial random effect)
- thinking about interacting effects (e.g., slope interacting with snowdepths)... interaction terms in models
- start with null models and model with temporal autocorrelation (nothing else) - what variables explain the residual variation? This is a good way to go. Plotting trends would be helpful.
- e.g., we have a lot of plots with Kobresia, what plot-level predictors may predict KOMY presence/abundance
- Look only at the most DYNAMIC plots, what about THOSE is changing? as a way to assess important trends
- probably most if the information there is in space
- subset by vegetation type, look to see if there are changes within plot type
- if so, look to see what predicts those changes
- but how to capture those trends? one option is coefficient of variation, but that doesn't capture directionality
- could look at a regression, or look at differences between first (x) years and last (x) years
- moving windows regression?
- Note: 76 plots have only one veg type classification, 6 plots have two; most plot-year combos do not have a veg class (last one is probably in 2008)
- Veg classes: wet/moist/dry meadow, fellfield, SB, SF, ST?
- have annual measurements, but not plot level measurements (that's better than nothing though)
- Growing degree days? Tom says there are experts for alpine plants and using temp -> growing season length (hopefully)
- note: Cliff has evidence that growing season length became decoupled over time from temperature... investigate (read manuscript?)
- where kobresia is present, there's a lot of it
- kobresia appears to be present more often where easting is negative (west facing? - drier?)
- look at dry meadows versus everything else... if there are more dry meadows, perhaps there will simply be more kobresia as a consequence
- loook at trends in kobresia... is there anything there worth modeling or interesting?
- maybe we could just include Kobresia abundance as a predictor instead of a response?
- Sean will talk to Cliff to where see any differences may come from
- it's probably not just competition between these two species...
- could do this with a multinomial regression?
- note: this is complicated with bare ground, litter, etc.
- to account for that note: could have an other plants, other non-veg
- Tom: look at trends (by plot, by veg type)
- Sean: talk with Cliff about slope, aspect, DEM, etc., also look at trends in bare ground, etc.
- Scott: snowmelt models, try multinomial models, thinking about temperature