Macroecology …characterizing and explaining patterns of abundance, distribution, and diversity
The Feasible Set: A New Understanding of Constraints on Ecological Patterns of Abundance
CHAPTER 1: How species richness and total abundance constrain the distribution of abundance CHAPTER 2: Efficient algorithms for sampling feasible sets
Rank-abundance curve (RAC) Rank in abundance Abundance Frequency distribution Species abundance distribution (SAD) Abundance class frequency
Frequency distribution The ubiquitous hollow-curve Abundance class frequency
Rank-abundance curve (RAC) Rank in abundance Abundance
Rank in abundance Abundance Predicting the SAD Observed Predicted
Rank in abundance Abundance N = 1,700 S = 17
Rank in abundance Abundance How many forms of the SAD for a given N and S?
Integer Partitioning Integer partition: A positive integer expressed as an unordered sum of positive integers e.g. 6 = = = Written in non-increasing order e.g
Rank-abundance curves are integer partitions Rank-abundance curve N = total abundance S = species richness S unlabeled abundances that sum to N Integer partition N = positive integer S = number of parts S unordered +integers that sum to N =
Combinatorial Explosion NSShapes of the SAD > 886 trillion > 302 trillion trillion
Random integer partitions Goal: Random partitions for N = 5, S = 3: Nijenhuis and Wilf (1978) Combinatorial Algorithms for Computer and Calculators. Academic Press, New York.
SAD feasible sets are dominated by hollow curves Frequency log 2 (abundance)
The SAD feasible set ln(abundance ) Rank in abundance N=1000, S=40
Question: Can we explain the SAD based solely on how N and S constrain observable variation?
DATA Ethan P. White, Katherine M. Thibault, and Xiao Xiao Characterizing species abundance distributions across taxa and ecosystems using a simple maximum entropy model. Ecology 93:1772–1778 DatasetNumber of sites Christmas Bird Count1992 North American Breeding Bird Survey2769 Gentry’s Forest Transect222 Forest Inventory & Analysis10356 Mammal Community Database103 TOTAL15442
DatasetNumber of sites Indoor Fungal Communities128 Terrestrial metagenomes Chu Arctic Soils, Lauber 88 Soils 128 Aquatic metagenomes Catlin Arctic Waters, Hydrothermal Vents 252 TOTAL METAGENOMES512 GRAND TOTAL15954 Microbial metagenomic datasets obtained from MG-RAST metagenomics.anl.gov
TOOL LOGO COOLNESS Sage mathematical software 8 Amazon Web Services 2 Weecology Servers (in-house) 10 TOTAL COMPUTING CORES 180 Generating random samples of the feasible set
Datasettotal sitesanalyzable sites Christmas Bird Count (6.5%) North American Breeding Bird Survey (57%) Gentry’s Forest Transect (82%) Forest Inventory & Analysis (71%) Mammal Community Database10342 (41%) Indoor Fungal Communities (97%) Terrestrial metagenomes (72%) Aquatic metagenomes (19%) TOTAL (60%)
The center of the feasible set ln(abundance) Rank in abundance N=1000, S=40
R 2 = Observed abundance Abundance at center of the feasible set North American Breeding Bird Survey (1583 sites)
Abundance at center of the feasible set Observed abundance
Abundance at center of the feasible set
DOI: /ele.12154
Public code and data repository
General Conclusions Feasible set: A primary way to account for how variables constrain ecological patterns…before attributing a pattern to a process
General Conclusions Extending the feasible set approach: ○Spatial abundance distribution ○Species area relationship ○Distributions of wealth and abundance The ubiquitous hollow curve
0.91 Observed Urban population sizes among nations ( , rescaled) Oil related CO2 emission among nations ( , rescaled) 0.92 Center of the feasible set
Observed home runs
General Conclusions ●The integer partitioning approach needs improvement
CHAPTER 2: Efficient algorithms for sampling feasible sets
Generate a random SAD for N=5 and S=
Combinatorial Explosion NSSAD shapes > 886 trillion 10001,...,1000> 2.4x10 31 Probability of generating a random partition of 1000 having 10 parts: <
Generate a random SAD for N=5 1) 5 2) 4+1 3) 3+2 4) ) ) )
Task: Generate random partitions of N=9 having S=4 parts
1.Generate a random partition of N with S as the largest part 2.Conjugate the partition A recipe for random SADs N = total abundance S = species richness
Generate a random partition of N with S as the largest part Divide & Conquer Multiplicity Top down Bottom up
Un(bias) Skewness of partitions in a random sample Density
Speed Number of parts (S) Sage/algorithm N = 50N = 100 N = 150N = 200
Old Apples: probability of generating a partition for N = 1000 & S = 10: < New Oranges: Seconds to generate a partition for N = 1000 & S = 10: 0.07
Integer partitions S positive integers that sum to N in without respect to order What if a distribution has zeros? subplots with 0 individuals people with 0 income publications with 0 citations
Abundance class frequency Intraspecific spatial abundance distribution (SSAD) N = abundance of a species S = number of subplots
SSAD N = total abundance S = no. subplots S non-negative abundances that sum to N without respect to order (weak) Integer partition N = positive integer S = number of parts S non-negative integers that sum to N without respect to order = Intraspecific spatial abundance distribution (SSAD)
Abundance class Frequency Abundance class
Frequency SAD “…frequency distributions of intraspecific abundance among sample sites resemble distributions … that have been used to characterize the distribution of abundances among species” (Brown et al. 1995) Species abundance = 1K Subplots = 100 Community abundance =1K Species = 50 SSAD Abundance class
Conclusions How do empirical SSADs compare to the feasible set of possible SSAD shapes? Other ecological patterns/distributions: – Occupancy frequency distribution – Collector’s curve – Species-area curve – Species-time relationship
Public code repository PeerJ Preprint Locey KJ, McGlinn DJ. (2013) Efficient algorithms for sampling feasible sets of macroecological patterns. PeerJ PrePrints 1:e78v1
Acknowledgements For collecting, managing and providing datasets: North American Breeding Bird Survey Christmas Bird Count Gentry’s Forest Transect Data Forest Inventory and Analysis dataset Microbial metagenomic datasets accessed from MG-RAST Mammal Community Database My committee: Morgan Ernest, David Koons, Jeannette Norton, Jacob Parnell Past: Mike Pfrender, Paul Cliften Colleagues: Justin Kitzes, James O’Dwyer, Bill Burnside, Jay Lennon, Paul Stone and the Stone Crew Faculty and Staff of the Biology Dept: esp. Brian Joy, Kami McNeil Funding: W. L. Eccles Graduate Research Fellow James A. and Patty MacMahon Scholarship Joseph E. Greaves Scholarship in Biology Dissertation Fellowship CAREER grant from NSF to Ethan White ( DEB ) Research grant from Amazon Web Services American Museum of Natural History Theodore Roosevelt Memorial Grant
Weecology I you guys
Sampling the SAD feasible Set Density Evenness Density Sample size = 300Sample size = 500Sample size = 700
Future Directions in Feasible Sets
Evenness and diversity metrics
The ubiquitous hollow-curve
New feasible sets: integer composition: all ordered ways that S positive integers can sum to N
New feasible sets: integer composition: all ordered ways that S positive integers can sum to N