Find research datasets worth reusing

Figure 5 in Cloudy with a chance of speciation: integrative taxonomy reveals extraordinary divergence within a Mesoamerican cloud forest bird

Figure 5. Phylogenies of Aphelocoma unicolor based on mitochondrial DNA and ultraconserved elements (UCEs). For the Bayesian time-calibrated mitochondrial DNA phylogeny generated in BEAST, the mean estimated split dates are provided on the nodes, with the 95% highest probability density shown below in square brackets. For both phylogenies, nodes with perfect support are shown with black dots.

Figure 4 in Cloudy with a chance of speciation: integrative taxonomy reveals extraordinary divergence within a Mesoamerican cloud forest bird

Figure 4. Results of a discriminant function (DF) analysis and normal mixture models on all morphological and plumage traits. A, differences among all five Aphelocoma unicolor subspecies in the first two DF axes. B, differences between only the A. u. unicolor and A. u. griscomi subspecies in the third and fourth DF axes. C, D, results of normal mixture modelling to determine the objective number of phenotypic clusters among individuals west (C) and east (D) of the Isthmus of Tehuantepec, with inset showing the assignment of individuals to each cluster with respect to their a priori subspecies assignment.

Figure 3 in Cloudy with a chance of speciation: integrative taxonomy reveals extraordinary divergence within a Mesoamerican cloud forest bird

Figure 3. Scatterplot of hue and colour saturation (chroma) for the five Aphelocoma unicolor subspecies.

Figure 2 in Cloudy with a chance of speciation: integrative taxonomy reveals extraordinary divergence within a Mesoamerican cloud forest bird

Figure 2. Differences among the five Aphelocoma unicolor subspecies for six morphological traits (measured in millimetres).

Figure 1. A, a in Cloudy with a chance of speciation: integrative taxonomy reveals extraordinary divergence within a Mesoamerican cloud forest bird

Figure 1. A, a unicolored jay (Aphelocoma unicolor unicolor) from Reserva de Biósfera Sierra de las Minas, Guatemala (Macauley Library ML85163771, photograph by Daniel Aldana). B, specimens representing A. u. guerrerensis (MLZ 45972), A. u. concolor (NMNH A9096), A. u. oaxacae (MLZ 33558), A. u. unicolor (MLZ 45360) and A. u. griscomi (AMNH 327521). C, distribution map of A. unicolor subspecies drawn from eBird observations.

ExtendedData Fig. 9 in Climate-driven variation in dispersal ability predicts responses to forest fragmentation in birds

ExtendedData Fig. 9 | Correlationbetweenseasonalityanddisturbance. At thelocallandscapelevel (a), seasonalityiscalculatedasthestandarddeviationof meanmonthlytemperaturevaluesthroughouttheyearatthelandscapecentroid (n = 31). Highdisturbancemeans 50% of thestudylandscapeareaoverlaps areasofhighnatural (forexamplestorms,glaciers,fires) orAnthropogenic (for exampleforestloss).Boxplotsshowmedian,interquartile range,andwhiskers toextremevalues (outliersaredatapoints>1.5x quartiles).Statisticsarefrom atwo-sided Wilcoxon test.Atthespecieslevel (b), communitymeanvalues (n = 31), arecalculatedusingspecies' distributionalseasonalityanddisturbance scores.Disturbanceiscalculatedastheproportionof thespeciesbreedingrange whichoverlapsareasofhighnatural (forexamplestorms,glaciers,fires) or anthropogenic (forexampleforestloss) disturbance.Seasonalityiscalculated asthestandarddeviationof meanmonthlytemperaturevaluesthroughoutthe year,averagedacrossallgridcellsinthespecies' breedingrange.Statisticsare fromalinearregressionwith Gaussianerrors;purplelineshowsmodelfit;shaded areais 95% confidenceintervals.

ExtendedData Fig. 8 in Climate-driven variation in dispersal ability predicts responses to forest fragmentation in birds

ExtendedData Fig. 8 | Relationshipbetweendispersallimitation (nHWI) anddiet. Datashownfor (a) 276 birdspeciessampledacross 18 temperate studylandscapes,and (b) 817 birdspeciessampledacross 13 tropicalstudy landscapes.Dietaryclasseswith <5 specieswereremovedfromtheanalysis.Diet classificationsarefrom Tobiasand Pigot110. F-statisticand P-valuearecalculated withatwo-way ANOVA.Boxplotsshowmedian,interquartile range,andwhiskers toextremevalues (outliersaredatapoints>1.5x quartiles).

ExtendedData Fig. 7 in Climate-driven variation in dispersal ability predicts responses to forest fragmentation in birds

ExtendedData Fig. 7 | Predictorsofdispersallimitationinbirds. Results shownareoutputsof phylogeneticleastsquaresmodelpredictingdispersal limitation (nHWI) acrossallbirdspeciessampled,includinglong-distance migrants (swallowimage,dark bars;n = 1034); onlyresidentspeciesandshort distance/partialmigrants (thrushimage,medium bars;n = 921); orresident speciesonly (pittaimage,palebars;n = 858). Panelspresentthreesetsofmodels withincreasingcomplexity:aunivariatemodelwithsinglepredictor (a,d), and multivariatemodelswithtwo (b,e) andthree (c,f) predictors.Eachpredictor iscalculatedatthespecieslevelbyaveragingacross landscapeswhereeach speciesispresent.Disturbance (red) iscalculatedastheproportionofspecies breedingrangewhichoverlapsareasofhigh natural (e.g. storms,glaciers,fires) oranthropogenic (e.g. forestloss) disturbance.Absolutelatitude (yellow) is calculatedasthecentroidlatitudeof thespeciesbreedingrange.Seasonality (blue) iscalculatedasthestandarddeviationof meanmonthlytemperature valuesthroughouttheyear,averagedacrossallgridcellsinthebreedingrange. a–c, EffectsiZeestimatesaregivenwith 95% confidenceintervals;anegative effectindicatesreduceddispersallimitation (thatisincreased dispersalability). R2 and AICvaluesarecalculatedforfullsamplemodelsonly.d–f, Proportion of independentvariationexplainedbyeachmodelcovariate,calculatedusing hierarchicalpartitioning.

ExtendedData Fig. 6 in Climate-driven variation in dispersal ability predicts responses to forest fragmentation in birds

ExtendedData Fig. 6 | Driversoffragmentationsensitivitywithnatural disturbances. Resultsof Bayesianphylogeneticmixedeffectmodelspredicting fragmentationsensitivityfor 1564 birdpopulations (n = 1034 species). Populationswereclassifiedasfragmentationsensitiveiftheywereidentifiedas 'Forest-core' by BIOFRAG. Restrictedanalysisassignedfragmentationsensitivity onlyto 'Forestspecialists' (a); Expandedanalysisassignedfragmentation sensitivitytoboth ' Forestspecialist' and ' Forestassociated' species (b; see Methods).Bayesianposteriordistributionisshownabovetheline;effectsiZe estimateswithcredibleintervals (CI) belowtheline (68%: thickerrorbars; 95%: thinerrorbars).HigheffectsiZesindicateapositiveassociationwith fragmentationsensitivity;loweffectsiZesindicateanegativeassociation. Finchandhawksilhouettesindicatethatbothmodelswererunonacomplete sample. Historicaldisturbanceisabinaryvariable (1/0) calculatedusingnatural disturbance (forexamplefires,storms & glaciation) layersonly.

ExtendedData Fig. 2 in Climate-driven variation in dispersal ability predicts responses to forest fragmentation in birds

ExtendedData Fig. 2 | Thelatitudinalgradientinaveragedispersallimitation ofbirdassemblages.Datapoints (colouredbylevelofhistoricaldisturbance) showthecommunitymeanvaluesforavianassemblagessampledat 31 study landscapesmappedin Fig.1. Theoverallgradientisnotexplainedbylandscape disturbancehistory.Absolutelatitudeisthecentroidlatitudeofallsampling pointsineachstudylandscape.Mean dispersallimitationisthenegative (thatis inverse) hand-wingindex (nHWI) averagedacrossallspeciesintheassemblage; nHWIislogarithmicallyscaled (log(1/HWI)) forvisualiZation.Statisticsarefroma linearmodelwith Gaussianerrors;purplelineshowsmodelfit (R2 = 0.44); shaded regionshowsthestandarderrorof theregressioncoefficient.

Fig. 3 in Climate-driven variation in dispersal ability predicts responses to forest fragmentation in birds

Fig. 3 | Fragmentationsensitivityincreaseswithdispersallimitationinbird assemblages.a, Variationin fragmentationsensitivity anddispersalability plottedonaconsensusphylogenetictree.Eachbranchrepresentsagenus (n = 441), withdataattipsaveragedacrossfamilies (n = 115) forvisualiZation. Branchcoloursindicatedispersallimitation (leastdispersivespeciesin red); tipcoloursshowtheproportionoffragmentation-sensitivespeciesineach family (expandedanalysis;mostsensitivein yellow).b, Datapoints (coloured bylevelofhistoricaldisturbance) aremeansfor 31 studylandscapes.Foreach assemblage,fragmentationsensitivityisassignedtoforest-corespecies withhighforestdependency (Restrictedanalysis),andmeandispersal limitationisthenHWIaveragedacrossallspecies;nHWIislogarithmically scaled (log(1/HWI)) forvisualiZation.StatisticsarefromageneraliZedlinear modelwithquasi-binomialerrors;purplelineshowsmodelfit (R2 = 0.180); shadedregionshows 95% confidenceintervals.Boxplotsin b showthesame distributionswithmedianvalue,interquartilerangeandwhiskerstoextreme values (outliersaredatapoints>1.5× quartiles).Resultsfortheexpandedsample areshownin ExtendedData Fig. 4.

Fig. 2 in Climate-driven variation in dispersal ability predicts responses to forest fragmentation in birds

Fig. 2 | Globalpatternsoflandscapedisturbanceanddispersallimitation. a, Thepresenceofnatural oranthropogenichistoricaldisturbancesrecorded ineachgridcell.Naturaldisturbancepressures (bright red) includemajorfires, stormsandglaciation;theseeventshavetypicallypersisted forlongerperiods oftimeandmaycausecompleteremovalofforestbiota.Anthropogenicforest loss (palered) representsmore recentdisturbancethatoftenalterscomposition oflocalassemblageswithoutcompleteeradication.b, VariationinnHWI averagedacross speciesoccurringineachgridcell,rangingfromlow (blue) to high (red) dispersallimitation.Dispersallimitationdataarecalculatedfrom measurementsof 10,562 birdspecies,logarithmicallyscaledforvisualiZation (log(1/HWI)).Yellowdotsshowstudylandscapes (21 from BIOFRAG;10 from additionalsampling).Gridcellsin a and b are 2.5 arcminutes. c,d, Hypothetical relationships:extinctionfilterspredictthatfragmentationsensitivityis negativelyassociatedwithhistoricaldisturbance (c), whiledispersal-related mechanismspredictthatfragmentationsensitivityispositivelyassociated with dispersallimitation (d).

Fig. 1 in Climate-driven variation in dispersal ability predicts responses to forest fragmentation in birds

Fig. 1 | Hypothesespredictingthedistributionoffragmentation-sensitive species. Thetoppathway (a) illustrateshow ' extinctionfilters' linkedto historicaldisturbances (forexample, fireandanthropogenicforestloss) canbenon-random,removingspeciestraitsassociatedwithsensitivityto disturbanceandretainingmoreresilientsurvivors.Tropicalbirdcommunities thathavelargelyavoidedseverehistoricaldisturbancetheoreticallycontain morespecieswithdisturbance-sensitivetraits (suchaspoordispersaland ecologicalspecialiZation),accentuatingtheimpactsofforestfragmentation (b). Backgroundturnoverofspecies,shownin (b) butpresentinallpathways, israndomwithrespecttodisturbance-sensitivetraits.Adifferentmechanism involvestheevolutionofflightadaptationstocopewithseasonalfluctuations intemperatureandresources (including vegetation,insects,flowersand fruits).In birds, thepredominantadaptationtoseasonalityinvolvesincreased mobility (fromlocaldispersaltolong-distancemigration),sohighly seasonal communitieslackdispersal-limitedspecies,potentiallyincreasingtheir resiliencetoforestfragmentation (c) incomparisonwithclimaticallystable regions (b). Relativespeciesrichnessisshownbythenumberofbirdsilhouettes inthecommunity.

ExtendedData Fig. 5 in Climate-driven variation in dispersal ability predicts responses to forest fragmentation in birds

ExtendedData Fig. 5 | Driversoffragmentationsensitivitywith Anthropogenicdisturbances.Resultsof Bayesianphylogeneticmixedeffect modelspredictingfragmentationsensitivityfor 1564 birdpopulations (n = 1034 species).Populations wereclassifiedasfragmentationsensitiveif they wereidentifiedas 'Forest-core' by BIOFRAG. Restrictedanalysisassigned fragmentationsensitivityonlyto 'Forestspecialists' (a); Expandedanalysis assignedfragmentationsensitivityto both 'Forestspecialist' and 'Forest associated' species (b; see Methods). Bayesianposteriordistributionisshown abovetheline;effectsiZeestimateswithcredibleintervals (CI) belowtheline (68%: thickerrorbars;95%:thinerrorbars).HigheffectsiZesindicateapositive associationwithfragmentationsensitivity;loweffect siZesindicateanegative association.Finchandhawksilhouettesindicatethatbothmodelswererunon acompletesample.Historicaldisturbanceisabinaryvariable (1/0) calculated usinganthropogenicdisturbance (forestloss) only.

ExtendedData Fig. 1 in Climate-driven variation in dispersal ability predicts responses to forest fragmentation in birds

ExtendedData Fig. 1 | Correlationbetweendisturbanceandlatitude. Study landscapesexposedtohighlevelsofhistoricaldisturbance (n = 16 landscapes; red) tendtobefoundathigherlatitudesthanlandscapesexposedtolowerlevels ofhistoricaldisturbance (n = 15 landscapes;blue).Disturbancelevelisestimated fromglobalmapsof majorhistoricaldisturbance (forexamplefire,glaciation). Absolutelatitudeisthecentroidlatitudeof allsamplingpointsineachstudy landscape.Boxplotsshowthemedian, interquartilerangeandwhiskers extendingtoextremevalues.Statisticsshowresultsof two-sided Wilcoxon ranksum testindicatingthatdisturbanceandlatitudearecorrelated (without accountingforspatialauto-correlation).

ExtendedData Fig. 4 in Climate-driven variation in dispersal ability predicts responses to forest fragmentation in birds

ExtendedData Fig. 4 | Correlationbetweenfragmentationsensitivityand dispersallimitationinbirds. Datapoints (colouredbylevelofhistorical disturbance) arecommunitymeanvaluesforavian assemblagesat 31 study landscapesmappedin Fig.1. Foreachassemblage,fragmentationsensitivity isassignedtospecieswith 'Forest-core' habitatpreferenceandeitherahigh or mediumforestdependency (Expandedanalysis).Meandispersallimitationis thenegative (thatisinverse) hand-wingindex (nHWI) averagedacross allspecies intheassemblage;nHWIislogarithmicallyscaled (log(1/HWI)) forvisualiZation. StatisticsarefromageneraliZedlinearmodelwithquasi-binomialerrors;purple lineshowsmodelfit (R2 = 0.270); shadedregionshows 95% confidenceintervals. Adjacentboxplotsshowthesamedistributionwithmedianvalue,interquartile range,andwhiskerstoextremevalues (outliersaredatapoints>1.5x quartiles).

ExtendedData Fig. 3 in Climate-driven variation in dispersal ability predicts responses to forest fragmentation in birds

ExtendedData Fig. 3 | Correlationbetweenfragmentationsensitivityand latitudeinbirds. Datapoints (colouredbylevelofhistoricaldisturbance) are communitymeanvaluesforavianassemblagesat 31 studylandscapesmapped in Fig.1. Foreachassemblage,fragmentationsensitivityisassignedto (a) Forest-specialistspecieswith 'Forest-core' habitatpreference (Restrictedanalysis),and (b) Forest-associatedspecieswith ' Forest-core' habitatpreference (Expanded analysis).Absolutelatitudeistheabsolutecentroidlatitudeofallsampling pointsineachstudylandscape.StatisticsarefromgeneraliZedlinearmodels withquasi-binomialerrors;purplelineshowsmodelfit (Restrictedanalysis:R2 = 0.2559, Expandedanalysis:R2 = 0.3208); shadedregionshowsthe 95% confidence intervals.

Fig. 5 in Climate-driven variation in dispersal ability predicts responses to forest fragmentation in birds

Fig. 5 | Predictorsofdispersallimitationinbirds. Resultsshownareoutputs ofphylogeneticgeneraliZedleast-squaresmodelspredictingdispersal limitation (nHWI) acrossallbirdspeciessampled,includinglong-distance migrants (swallowimage,dark bars;n = 1,034), onlyresidentspeciesandshort distance/partialmigrants (thrushimage,medium bars;n = 921) oronlyresident species (pittaimage, palebars;n = 858). Panelspresentthreesetsofmodels withincreasingcomplexity:aunivariatemodelwithsinglepredictor (a,d), and multivariatemodelswithtwo (b,e) andthree (c,f) predictors.Eachpredictoris calculatedatthespecieslevelbyaveragingacrosslandscapeswhere eachspecies ispresent.Disturbance (red) isthelocalbinarydisturbancescore,latitude (yellow) istheabsolutelatitudeof thelandscapecentroids andseasonality (blue) isthestandarddeviationof meanmonthlytemperaturevalues.a–c, Effect-siZeestimateswith 95% confidenceintervals;anegativeeffectindicates reduceddispersallimitation (thatis,increaseddispersalability).R2 and Akaike informationcriterion (AIC) valuesarecalculatedforfullsamplemodelsonly. d–f, Proportionof independentvariationexplainedbyeachmodelcovariate, calculatedusinghierarchicalpartitioning.

Fig. 4 in Climate-driven variation in dispersal ability predicts responses to forest fragmentation in birds

Fig. 4 | Dispersallimitation (nHWI) explainsvariationinfragmentation sensitivity.Resultsof Bayesianphylogeneticmixed-effect modelspredicting fragmentationsensitivityforall 1,564 birdpopulations (n = 1,034 species). Populationswereclassifiedasfragmentationsensitiveiftheywereidentifiedas 'Forest-core' by BIOFRAG. Restrictedanalysisassignedfragmentationsensitivity onlyto 'Forestspecialists' (a); Expandedanalysisassignedfragmentation sensitivitytoboth ' Forestspecialist' and ' Forestassociated' species (b; see Methods).Bayesianposteriordistributionisshownabovetheline;effect-siZe estimateswithcredibleintervals (CI) arebelowtheline (thickerrorbars,68%; thinerrorbars,95%).HigheffectsiZesindicateapositiveassociationwith fragmentationsensitivity;loweffectsiZesindicateanegativeassociation. Historicaldisturbanceisabinaryvariable (1/0) calculatedusingalldisturbance layers (forestloss,glaciation,stormsandfires).