Bird abundance and distribution data for emergency response.
Tracking data
- Get metadata
- Build kernels 2 usage grid function
Abundance data
- Get metadata
- Kernels vs. rqss GAMs
##PROS and CONS of different methods
###GAMs
PROS
-
Easily interpretable value (group size; number of birds seen when going to a site, by quantiles)
-
Independent of the number of observers/checklists (but needs a certain number of observations to give a meaningful result)
CONS
-
Can be difficult to implement
-
Sensible to smoothing value
-
May give extreme results when few observations are available and quantile is high (e.g. 90%)
-
Need cells for predictions
###KERNELS
PROS
-
Easier to implement and to use (e.g. risk: high, medium, low)
-
Easy to output to polygons
CONS
-
Not easy to interpret real meaning (group size, number of birds concerned?)
-
Strongly depends on the amount of observers and checklists (lightly mitigated by a weigthing method based on numbers)
-
Sensible to bandwith value and different parameters
##NOTES and IDEAS
-
Use GAMs to predict quantile group size in each kernel polygons
-
Need to sum species by groups when doing GAMs
-
Integrate ECSAS, tracking and ebird data within the same kernel output
##TODO
- Winter Eiders, check email for using whites and adding browns which are in % ?
- Take out groups and observations flagged red in FB dynamic excel file (in downloads)
- Add EPOQ data, but only what is older than 2012 (?) to reduce overlap with EBIRD data
- Verify which seasons to use
- Add data from SOMEC (2012 and over ?) to the ECSAS data
- Use 50-70-90% kernels
- Shanti wants tracking data? (email 01-27)
- BIOMQ never was included in the kernel output (email 01-27)
- Adjust seasons according to (email 01-27)
- Check list of things left to do on google drive
- Make sure last version of ECSASconnect is used to produce RData with ECSAS data
- Check FB commentary on putting seabird survey data with ECSAS and shorebird data with ebird