Identify appropriate multi-species datasets

[nicholasjclark]

There is considerable information (with example code) provided by this preprint and the accompanying Github repo

[caseyyoungflesh]

I've taken a bit of a poke around in the Johnson et al. resources (summarized well here in the repo associated with the publication). They compiled the following datasets for their analyses:

• RivFishTIME
  • freshwater fish abundance time series
• North American BBS
  • North American bird abundance time series from breeding season (already in use)
• BioTIME
  • assemblage abundance time series from various biomes/species
• Living Planet Index
  • abundance time series from various biomes/species
• CaPTrends
  • mammal abundance time series from various biomes
• ReSurvey Germany
  • vegetation plot resurvey data
• Environment Agency NFPD Fish Counts
  • freshwater and marine fish abundance time series in UK
• FishGlob Global Bottom Trawl Survey Data
  • marine fish abundance time series from bottom trawls
• TimeFISH
  • reef fish assemblage abundance time series from southwestern Atlantic
• Pilotto
  • a study that aggregated a number of datasets from "long-term ecological research sites"

There are a few others that come to mind (and have been brought up) that we might consider as well:

• Christmas Bird Count
  • North American bird abundance time series from non-breeding season
• UK Breeding Bird Survey
  • UK bird abundance time series (I believe this is request-only)
• Swiss Common Breeding Bird Survey
  • Swiss bird abundance times series (I believe this is request-only)

[caseyyoungflesh]

If we're interested in functional relationships, AVONET is a great resources for birds. This contains morphological (and other) measures for nearly all bird species across the world.

[caseyyoungflesh]

I suppose the question is which of theses datasets are the most relevant to the goals of this project/are the most robust? Are there some taxonomic groups that are 'better'? I would argue that birds are advantageous because they are relatively conspicuous and survey methods are fairly standardized -- maybe this is just my bird-bias though. Birds also have great information for functional traits, though there are some data available for mammals as well.

I think we should also keep in mind whether these are closed populations. The strict answer is no for nearly all, but some are going to be more closed than others. Or maybe it doesn't matter because any variation in abundance across time that might be driven by movement across space is still interesting? Thoughts on this? Or do we even have expectations on which taxa/sampling strategies are captured closed vs. not populations?

[caseyyoungflesh]

I'm looking into the BioTIME datasets more closely now to see what the breakdown is in terms of # of species, sampling type, etc.

[caseyyoungflesh]

Potentially relevant functional fields in AVONET:

• Beak.Length_Culmen
• Beak.Length_Nares
• Beak.Width
• Beak.Depth
• Tarsus.Length
• Wing.Length
• Kipps.Distance
• Secondary1
• Hand.Wing.Index (though to be indicative of dispersal capabilities)
• Tail.Length
• Mass
• Habitat
• Habitat.Density
• Migration (resident/partial migrant/migrant)
• Trophic.Level
• Trophic.Niche
• Primary.Lifestyle
• Range.Size

Estimated generation lengths (a nice catch-all metric for 'pace of life') are available for all bird species from Bird et al. 2020. As far as 'functional traits' go, I'd say this is an important one

PHYLACINE has data on a number of traits for mammals

AmphBIO has data on a number of traits for amphibians

TRY has data on a number of traits for plants

More trait databases can be found on the Open Traits Network but from my understanding all of the above are relatively well curated (but with various levels of completeness across species/traits).

[caseyyoungflesh]

While there are many datasets in BioTIME...

Many have limited spatial/temporal coverage. When using the following filters (at least 10 locations, 10 species, 25 time points of data, and specific taxonomic groups),

biotime_meta_f <- dplyr::filter(biotime_meta, 
              number_lat_long >= 10,
              number_of_species >= 10,
              #time points
              data_points >= 25,
              taxa %in% c('Birds',
                          'Mammals',
                          'Amphibians',
                          'Terrestrial plants',
                          'Fish'))

we're left with this.

These are the dataset titles (which includes the NA Breeding Bird Survey):

 [1] "Seabirds of the Southern and South Indian Ocean (Australian Antarctic Data Centre)"                                           
 [2] "DFO Maritimes Research Vessel Trawl Surveys Fish Observations (OBIS Canada)"                                                  
 [3] "South Western Pacific Regional OBIS Data provider for the NIWA Marine Biodata Information System (South Western Pacific OBIS)"
 [4] "Pacific Shrimp Trawl Survey (OBIS Canada)"                                                                                    
 [5] "ECNASAP - East Coast North America Strategic Assessment (OBIS Canada)"                                                        
 [6] "Breeding birds survey North America"                                                                                          
 [7] "Long-term growth mortality and regeneration of trees in permanent vegetation plots in the Pacific Northwest 1910 to present"  
 [8] "Vegetation Plots of the Bonanza Creek LTER Control Plots Species Count (1975 - 2004)"                                         
 [9] "Upper Little Tennessee River Biomonitoring Program Database - LTWA Biomonitoring Database"                                    
[10] "Pelagic Fish Observations 1968-1999"                                                                                          
[11] "South Western Pacific Regional OBIS Data provider for the NIWA Marine Biodata Information System (South Western Pacific OBIS)"
[12] "DFO Maritimes Research Vessel Trawl Surveys Fish Observations (OBIS Canada)"                                                  
[13] "Long-term stem inventory data from tropical rain forest plots in Australia"                                                   
[14] "The New Zealand Freshwater Fish Database - Electric fishing - Backpack"                                                       
[15] "The New Zealand Freshwater Fish Database - Traps- Gee Minnow traps"                                                           
[16] "Trawl fisheries in Israeli Mediterranean" 

I'd argue that [1] isn't such a great dataset as this is derived from seabird sightings taken from ships. Seabird density can be impacted by windspeeds, they follow ships, etc. I'm unfortunately not very knowledgeable about marine fish surveys but there do seem to be a few datasets available.

[nicholasjclark]

Thanks for all this digging @caseyyoungflesh, it is good to get a sense of what datasets we should consider as possibles. Perhaps @GitTFJ can also give some thoughts on which of the datasets that were used in his paper could be most appropriate here?

[nicholasjclark]

Potentially relevant functional fields in AVONET:

• Beak.Length_Culmen
• Beak.Length_Nares
• Beak.Width
• Beak.Depth
• Tarsus.Length
• Wing.Length
• Kipps.Distance
• Secondary1
• Hand.Wing.Index (though to be indicative of dispersal capabilities)
• Tail.Length
• Mass
• Habitat
• Habitat.Density
• Migration (resident/partial migrant/migrant)
• Trophic.Level
• Trophic.Niche
• Primary.Lifestyle
• Range.Size

Estimated generation lengths (a nice catch-all metric for 'pace of life') are available for all bird species from Bird et al. 2020. As far as 'functional traits' go, I'd say this is an important one

PHYLACINE has data on a number of traits for mammals

AmphBIO has data on a number of traits for amphibians

TRY has data on a number of traits for plants

More trait databases can be found on the Open Traits Network but from my understanding all of the above are relatively well curated (but with various levels of completeness across species/traits).

Good to see. @shubhi124081 also mentioned a recent paper that inspected axes of variation for birds, which seemed to show that a few specific traits can inform clusters of bird species. I don't remember the exact details, but perhaps she can give that reference here

[nicholasjclark]

Also just saw this: https://twitter.com/thiagotoyoyo/status/1785048029550248071?t=54H3fCW-SknAVoFouA8piA&s=19. Haven't looked carefully yet though

[GitTFJ]

Hi @caseyyoungflesh great summary. Agree with all of the above. I think the bird monitoring schemes are the best place to start. I am currently trying to setup a forecasting challenge for the European chapter of EFI and we are in the process of trying to compile all of the open (on-request) bird datasets for the challenge. If successful, and data providers approve, we could embed the data in the same pipeline.

Another dataset is the waterbird suveys in this paper. Heavily towards protected areas though

[caseyyoungflesh]

@nicholasjclark @shubhi124081 was that paper in question this one (came to mind since it uses what would become AVONET)?