ProjectNotes

2021-08-06

Silas - These are my notes from a few years ago.  Please:

1. Please look up the MARSS package for guidance
2. I added two .R files to your code folder with some of the code for the presentation I showed you for 1997 - 2017. 
The files are essentially copies of the MARRS vignette (link: https://cran.r-project.org/web/packages/MARSS/vignettes/UserGuide.pdf) 
chapters 8 and 9 (used to be chapters 7 and 8), hence my file names.
3. You could try to updated this code or start from scratch, up to you. I recall that I didn't get all of the Ch8 file to run.

Thanks!

######################################################################################################
##### Old notes below from 2017-2018, may be useful to understand motivation behind existing code ####
######################################################################################################

=Seal project outline

BREEDING SEASON EMPHASIS

1. population trends 
   overall and by site
   
2. get survivorship (NO GOOD)
      year on year adult + pup

3. get fecundity = 
      pups/adults 
      assume 50:50 sex ratio
         = so pups/(0.5*adults)
         but males in water more during peak pupping season.  check with sarah.....

4. What number of surveys needed for monitoring?
   top_n with regressions
      compare estimates of population growth
      
5. Covariates on fecundity, pups, etc.
   ENSO
  
   double cpoint crash in 2016 with slides and coyote
   how much move to other colonies?
   
   phenology cline by site?
   
   if lose 2 big centers what happens to PORE colony?
   
   check with Codde on coyotes
   
6. MARRS
make files with top count of year for 
  1) adults, 
  2) pups
  3) adults + pups

H1: DP+DE+BL vs TB+TP
H2: all independent  (basic models show this is lowest AIC by far)
H3: Estuary vs Coast
      DE+BL+TB vs. DP + TP
don't count DUX, PB, or PRH
WELL, maybe count PRH and 
   
 -> how include ENSO covariate
 -> how code inverse relationship between sites withing a hypothesized group?  B-matrix, but performs poorly in converging.
   
   OTHER IDEAS FOR MARRS
   weigh station of 3 small sites during molt?
   
COMPARE: PORE Breed and Molt VS NMFS Molt 
  plots and population parameters.
  
  
7. basic MARRS file show that 5 independent pops are most parsimoniuos
next going with GAMMS for pops over time and MEI index

8. fix plots
report MARSS
report GLMMs including pup/adult
report rstanarm models


2021-09-20
TO DO
1. update plots to ggplot
2. get MEI to run
3. Solve linear plotting issue for best model

2021-10-08
Quick git update on bb computer test




2024-01-12
Hypotheses for adult-molt time series  1982 present
1. one adult pop vs one molt pop   
  A. (use for calculating overall regional trends for molts and adults)
2. one all stages pop
3. 12 individual pops
  A. include coyote
  B. Use to understand impacts of coyotes on sites.
4. North, South, and Bolinas pop
  A. can sum coyote for Drakes and DP ?


5. Perhaps build models with all years data from 1997 on?  then go back to the older data for just molt and adult?


2024-01-16

1. Growth rate by time - DONE
2. insert lag effect for PDO from previous fall (embryo implant) (DONE --> 2024-05 --> using Upwelling instead from previous spring)
3. add coyote leg effect from previous year or cumulative years with coyotes DONE = 3 year sum of presence/abselce 
4. add 2003 eseal to coyote disturbance(with and without)  (ADDED to PRESENT = 1) = DONE
5. Make sure future pop growth not too optimistic  =I THINK DONE
6. NOAA appreciates if molt helps predict breeding season - measures how in sync
7. corr plot    
  A. group sites close together --> BL, DE, DP and TP, TB            ####STARTED 2024-02-06   DONE
    i. can include or exclude PRH try both ways  INCLUDED
  C. include DR and PB in model if BL, DE, DP, TP, TB over time vs DR, PN, PRH  DONE
  D. 


2024-05-26
1. adding general UI Mar_MAY for 39N, including lag  - use as main covariate  - DONE
  A. upwelling data from here: https://oceanwatch.pfeg.noaa.gov/products/PFELData/upwell/monthly/upanoms.mon

TODO: put onto bbecker1000 repo

2024-05-29
Ben and Sarah A.
1. Determine the effect of column on row is within year OR year t to year t+1?
2. SARAH COMMENTS
  A. Add Human Disturbance
  B. Add Water Temperature (SST) April-June for pup feeding. 
  C. Consider MOCI - WILL USE MOCI AND MOCI LAG INSTEAD
  D. BEUTI - biologically effective upwelling index - WILL USE BEUTI AND BEUTI LAG INSTEAD  DONE
3. Add to the network matrix.  DONE
  Hypotheses to add to matrix.
  A. TB-TP
  B. DE-DP
  C. BL-DP
4. BB check the data for Double Point between 1982 - 2002 - DONE - checked with Sarah A.  looks good.
5. include all MAJOR disturbances in the coyote effect ONLY for LONG-TERM dataset.
   A. Live Rockfish (just plot, no stats) - disturbance from boats and fish depletion (NEED DATES FROM SARAH A.)
   B. gillnet fishery --> affected weaned pups (lag effect to next year)...just plot...no stats.  Add in YEARS that were active.  SA checking data.
   C. Shooting --> plot, no stats.
   D. sharks (qualitative)
   E. competition between weaned eSeal and phoca.  DONE!!
    i. Quantitative data for weaned eSeal pups in Drakes Bay for same year.  SUMMED
    ii. Just for DE, DP, PRH --> All three age classes (Adult, Molt, Pup).   DONE
    iii. eSeal data-> SUM the IMMATURE MOLT and the WEANED PUP count as for all PR sites (total pop) covariate.  DONE
   

PAPER STORY:
A. Env effect have a huge effect overall
B. Ability for populations to absorb disturbances

Story of recovery of an ecosystem



2024-06-06

SC: for MOCI 
1. YEAR OF: MOCI JFM (as is)  UPDATED THRU 2023. MEAN OF CENTRAL AND NORCAL
2. Implant sustain: MOCI OND previous year DONE
3. Lag previous year conditions For MOCI AMJ leaves good food conditions for fall  DONE
4. run with 2023 data included. DONE (except used mean # juvenile eSeals from 2021&2022 = 2023...need new data from Sarah C.)



2024-06-11
1. Think about changing state R to vary by year.  probably not...
1A. Ut Site x Class OR Class?   RUNNING CLASS ONLY!
2. Derive pup:adult ratio from output results by site any by year 
   A. can we look at covariates explaining this.
3. Incorporate 3 years of lag into the coyote rate.   DONE
   A. Use the weighted count of the past 3 years
   B. Weight 50-30-20 for years t, t-1, t-2
   C. BB needs 2023 disturbance data from SC.         
4. Conceptual model of the relationships


2024-06-13
1. Updated eSeal IMM, WNR, YRLNG through 2023
2. No Coyote rate at DP 2020 (no surveys) so used mean of 2019 and 2021.
3. Added coyote weighted 3 year through 2023.
4. Class uT model.


2024-06-17
1. For the B matrix, see MARSS manual section 14.2
   The diagonals are interpreted differently than the off-diagonals since
the diagonals are (bi,i − 1) so subtract off 1 from the diagonals to get the effect of
species i on itself. If the species are density-independent, then Bi,i would equal 1.
Smaller Bi,i means more density dependence.
2. B matrix is effect of t-1 on t.  (from MARSS manual)
3. hard to estimate U and B in same model...see if can run with steady state U?

2024-06-18
1. add human disturbace covariate by site.  SA -> all age classes effected the same, so single covariate.  DONE But can try single age classes variable.
2. Pup:Adult ratio and then what sort of model
  a. Not sure about covariates since already used in the MARSS model
  b. let's just show graphically since already was modeled.
3. Next meeting in July.

2024-06-23
1. Models with human disturbance run well
2. Currently just year of disturbance.  Should it be 3 year weighted like Coyote?
3. scenarios for projections (see if can plot cleanly, or supplement?)
  A. good MOCI, low coyote, low dist
  B. poor MOCI, low coyote, low dist  
  C. poor MOCI, High coyote, low dist
  D. poor MOCI, Low coyote, High dist
  C. poor MOCI, High coyote, High dist
  D. **organize on the management leverage points (coyote, disturbance)
  
2024-07-19
  A. add ratio plots - DONE
  B. Create disturbance plots - DONE 
  C. Plots for site independent effects: MOCI on PUP, MOCI OND pup, MOCI Molt
  D. Update predict projections - 
  
2024-07-30
A. include plot ratios for pre-1997 data.
B. plot by site back to 1970s (model without all the covariates?)
C. How do pup:adult ratios change over time?  (Long data series)
D. After add the long-term data, calculate rates of change over time for overall and by site. 

2024-08-21
A. removed eSeal covariate but then kept
B. check that ratio being calculated correctly - done
C. check why new model has some odd estimates for pop size. fixed - was b/c molt date filter removed year 1 for TB.  expanded date windo 5 days.

D. Fix up the B matrix since it is one year in the future.