Bridging gaps in demographic analysis with phylogenetic imputation

Phylogenetically informed imputation methods have rarely been applied to estimate missing values in demographic data but may be a powerful tool for reconstructing vital rates of survival, maturation, and fecundity species conservation concern. Imputed could used parameterize models explore how populations respond when are perturbed. We standardized rate estimates 50 bird assess the use phylogenetic fill gaps data. calculated accuracy focal excluded from set either singly or combination with without phylogeny, body mass, life-history trait imputed calculate metrics, including generation time, validate analyses. Covariance among other provided strong basis guide birds, even absence information. Mean NRMSE null differed by <0.01 except no were available high signal (Pagel's ? > 0.8). In these cases, mass compensated lack information: mean normalized root square error (NRMSE) adult survival <0.04 maturation rate. Estimates metrics sensitive rates. For example, time doubled response inaccurate time. Accurate such as needed inform planning processes, example through International Union Conservation Nature Red List assessments population viability analysis. useful this context but, any estimated model parameters, awareness sensitivities outputs is essential. Cerrando Brechas en los Análisis Demográficos con Imputación Filogenética Los métodos de imputación guiados filogenéticamente se han aplicado poca frecuencia para estimar valores faltantes datos demográficos, aunque pueden ser una herramienta poderosa la reconstrucción tasas vitales supervivencia, maduración y fecundidad especies importancia conservación. Las imputadas podrían usarse generar parámetros modelos demográficos explorar cómo responden las poblaciones cuando perturban vitales. Utilizamos estimaciones estandarizadas aves analizar el uso filogenética llenar vacíos demográficos. Calculamos certeza focales excluidas del conjunto por sí solas o combinación sin filogenia, masa corporal características historia vida. Usamos calcular medidas demográficas, incluyendo tiempo generación, así validar análisis La covarianza entre otros proporcionó base sólida orientar aves, incluso ausencia información filogenética. El medio nulo filogenético difirió salvo hubo disponibles señal alta (? Pagel En estos casos, inclusión vida compensó falta filogenética: cuadrático raíz normalizada media filogenéticos supervivencia adulta tasa maduración. demográficas fueron sensibles imputadas. Por ejemplo, generacional duplicó respuesta imprecisas certeros, como generacional, son necesarios procesos planeación conservación; través valoraciones Lista Roja Unión Internacional Conservación Naturaleza viabilidad poblacional. útiles este contexto, pero cualquier tipo parámetro modelo estimado, conocimiento sensibilidades rendimiento demográfico es esencial ??????????????????????????????, ????????????????????????????????????????????????????????, ??????????????????????????????????????????, ??????????????????????????????????????????????????, ???????????, ?????????????????????????, ??????????????????????, ??????, ???????????????????????, ?????????????????????????????????????????, ??????????????????????????????????????????????? 0.8) ?????, ???????????????????????????? 0.01 ???????, ?????????????????????????:?????????????????????????????????? 0.01, ????????? 0.04 ???????????????????????????????, ?????????????????????????????????????????? (?????) ???????????, ??????????????????????????????????????????????????, ?????????????????????????????????????????: ???; ??: ???? ???????????????????? Understanding responses human-induced threats, habitat loss degradation, climate change, overexploitation (Brook et al. 2003; Parmesan 2006; Maclean & Wilson 2011; Maxwell 2016), crucial identifying at-risk interventions (e.g., Bruna 2009; Dahlgren 2016; Lunn 2016). Population parameterized development, reproduction can generate predictions about will pressures that cause changes (Selwood 2015). Obtaining necessary populate requires investment resources which lacking critical setting. The most those information (Beissinger Westphal 1998; Coulson 2001; González-Suárez 2012), due geographical, taxonomic, biases recording (Roberts Troudet 2017; dos Santos 2020) logistical barriers collecting complete (Menges 2000; Weimerskirch Pike 2008; Clutton-Brock Sheldon 2010). Consequently, empirical only small biased subset (Lebreton 2012; Salguero-Gómez 2015, Conde 2019). When species, ad hoc commonly modeling 1998). Parameter derived based on relatedness (Heinsohn 2004; Koenig 2008) similarity (McCarthy 1999; Valle 2018). Other approaches include combining form representative (Sæther Bakke 2000) parameterization range plausible (Rodríguez 2004) captive individuals Young 2012). Such produce bias (Schafer Graham 2002), their raises concerns reliability ability make robust conclusions Engen 2002; Ellner Fieberg McGowan 2011). Therefore, formal estimating quantifying uncertainty needed. Many imputing expectation similar closely related (Felsenstein 1985; 1999). By accounting more formally evolutionary history, it possible improve Phylogenetic together an describing divergence (Martins Hansen 1997; Freckleton species-based Traits less labile, leading differences well predicted relationships (Freckleton Blomberg 2003). signal, measure strength dependence (Pagel Garland determine benefit using (Penone 2014). If strong, phylogenetically potentially performance. has proposed filling functional plants (Swenson 2014) mammals (Guénard 2013; Penone data, although hierarchical incorporating taxonomy parameters fish (Thorson al 2017). focused traits, namely fecundity. plants, single suggests neither nor different life stages strongly phylogeny species-level traits (Che-Castaldo 2018), reflecting weak plant (Burns vertebrates, characteristics covary (body size, morphology, traits) interpreted being informative stabilizing selection lability (Blomberg see Revell 2008). Whatever exact processes involved, tendency size (Stearns 1983) age at maturity clutch Sæther they also would setting infer species. inclusion covarying allometric Shine Charnov 1992; Brawn 1995). provide means derived. Demographic interest growth its sensitivity elasticity underlying (Benton Grant 1999) Sensitivity analysis identifies capacity change valuable making well-founded interventions. Generation international bodies, (IUCN), indicators decision-making (Mace sparse, proxies reproductive lifespan (Fung Waples Staerk 2019) directly (Fagan Cooke 2018) (Pacifici Bird 2020). over alternative methods. existing vital-rate birds feasibility Although many avian sets compiled Lebreton concern 82% derive apply multivariate framework incorporates covariance impute values. determined accurately imputed, combination. Further, we assessed value (clutch female maturity) original performance extinction risk. All analyses carried out R (version 3.6.3, Core Team extracted matrix COMADRE Animal Matrix Database 3.0.1, sources 2000). screened avoid errors construction (Kendall ensure valid structure subsequent (Appendix S1). resulting represented across 15 orders histories. identified prebreeding postbreeding census categorized each history early (individuals mature breed after 1 year) delayed remain nonbreeding juveniles years; Fujiwara Diaz-Lopez Allowing representation early- delayed-maturation models, collapsed prereproductive (Salguero-Gómez Plotkin 2010) representing first-year (), () matrices. To full analysis, restricted main 40 S2). combined (Wilman 2014; Myhrvold 2015) transformed all variables satisfy requirements downloaded sample 1,000 trees (Hackett backbone) BirdTree website (www.birdtree.org, Jetz pruned match tree topology was supported (3 nodes posterior probability <0.95), so least squares consensus method (Lapointe phytools version 0.7-20, 2012) create average S3). This creates sum-of-squares patristic (node-to-node) distances minimized. compared results distribution demonstrate our insensitive S5). pattern comparing observed distributions expectations specified evolution. Pagel's transformation obtained maximum likelihood, produces best fit Brownian motion takes 0 (complete independence) (patterns proportional shared history) above (traits than expected) 2002). 0.7-20) (Revell account residual branch lengths. addition, Rphylopars 0.2.12) (Goolsby 2016) characterize taking into conditioning problems included both phenotypic variation excluding not detrimental estimation S6). (Table 1). set, (Eq. 1) compare inspected (systematic differences) increased variance. (0.246 [SD 0.013]), intermediate (0.532 [0.018]), (0.889 [0.016]) (0.923 [0.116]). (sa: 0.817 [0.019]; m: 0.934 [0.094]). High suggested should successful targets variance greater 0.488 increasing 0.702 added decreasing 0.684 included; These indicate improves characterization rates, do further improvement act slightly negatively signal. Adult data: 0.169 (SD 0.039) 0.172 (0.019), respectively. (mean NRMSE: 0.248 [0.010]) 0.346 [0.055]) accurate. fecundity, accurate (Fig. 2). However, helped rate, particularly multiple Including improved 2) reduced difference between survival. had 0.075 0.011]) 3), despite higher (0.140 [0.073]) markedly included, 2 outliers underestimated, overestimation S17). because assumed juvenile equal relatively characterized 0.041 0.007]) 3). 0.234 [0.035]) and, metric overestimated S18). unknown 0.121 0.007] 0.118 [0.011], respectively) influenced adding matched reasonably 0.051 <0.001]) 4). 0.125 [0.010] 0.126 [0.014], respectively). arose 0.221 [0.039]) driven S20). varied calculation Responses consistent elasticities; 0.042 0.009]) 0.060 [0.019]) 0.105 [0.013]) 0.161 [0.027]) Errors elasticities unbiased S21 S22). Detailed understanding species’ global address current biodiversity crisis, limited predict trajectories (Kindsvater 2018; Efforts IUCN (IUCN designed (Rodrigues Mace 2008), compensate result under- risk high, long-lived important reliable found applying accounted reproduction, allowed us some well, cases. auxiliary tended Imputation did reflect ranking strongest (2014) linked influence carnivores much covaried size. deteriorated suggesting patterns important. overestimate (Appendices S9 S13). discrete stage-based step 1, whereas onset <1. bimodal severely non-normal, transformation. normally distributed unusual distribution. Our finding contrasts studies showed minor effects does per capita 2013), largely fail predictability typically assessment purposes, depended Moreover, simplified cycle approach introduce (Fujiwara researchers advise caution interpretation parameter Reed 2002); care availability partiality 2019), strengthening random (MNAR) concern, comparative skewing (Nakagawa geographical latitudes) latitude sufficient control (Jetz Scholer Future broader coverage investigate affect taxonomic groups. success rests validity structure. known covariates, Thus, quantity suggest exploring impact input by, varying within reasonable limits outputs. Uncertainty explored way sampling trees. poorly handled distributional assumptions method. approach, two-component mixture capture qualitative translate metrics. A quantify propagates novel bridging imputation. cannot replace detailed age-specific available, sparse data-limited avoids associated assuming family- genus-based indicators, List, informs spatial prioritization 2006), Green assessing recovery (Akçakaya data-driven essential guiding business supporting sustainable development goals (Brooks 2015; Bennun thank R.O. Martin, G.H. Thomas, K.J. Simpson guidance editorial advice. H.R. Akçakaya, R. Robinson, 4 anonymous reviewers comments manuscript. T.D.J. funded Adapting Challenges Changing Environment, NERC doctoral training partnership (ACCE DTP; NE/L002450/1), additional CASE funding World Parrot Trust. R.S.-G. (R/142195-11-1). O.R.J. grant Danish Council Independent Research (DFF – 6108-00467). work benefitted interactions feedback sDiv working group sAPROPOS (Analyses PROjections POpulationS) led German Centre Integrative Biodiversity (iDiv). Please note: publisher responsible content functionality supplied authors. Any queries (other content) directed corresponding author article.