Data from: Demographic responses to population recovery illustrated by 30-years of monitoring a once critically endangered parrot

Understanding how demographic rates change during population recovery is critical for evaluating conservation success and determining when populations become self-sustaining. However, documenting such transitions requires long-term monitoring across substantial changes in population size - conditions that are rarely met. We leverage a 29-year capture-mark-recapture dataset (1994 - 2022) from the echo parakeet population (Alexandrinus eques), a vulnerable endemic parrot of Mauritius island that recovered from fewer than 20 individuals to over 500 during the study period. Using multistate capture-mark-recapture models, we analysed encounter histories of 2,734 known-aged wild individuals to estimate age and state-specific survival probabilities while accounting for temporal variation, transitions between breeding states (PB = pre-breeder, B = breeder, NB = non-breeder) and controlling for un-equal recapture probabilities. Demographic rates changed substantially during population recovery. Recruitment into the breeding population declined across all age classes as the population approached carrying capacity. Survival showed contrasting patterns: juveniles (< 2 years old) exhibited a declining temporal trend (from 0.68 in 1994 to 0.57 in 2020), whereas adult survival remained stable across years, with breeding adults showing consistently higher rates (0.96-0.97) than non-breeding individuals (0.76-0.90). Synthesis and applications. These findings highlight the dynamic nature of demographic processes throughout population recovery and emphasise the value of continued monitoring beyond apparent recovery to detect density-dependent shifts in demographic rates of wildlife populations. Field biologists from the Mauritian Wildlife Foundation systematically check all known nest boxes and natural cavities to identify breeding pairs and record their reproductive success during the austral summer breeding season (September to March). While ringing records began in 1991, the first chicks to successfully fledge were marked in 1994, establishing this cohort as the starting point for our capture-recapture analysis. Individuals are ringed in the nest before fledging, enabling long-term tracking and identification of birds within the population. Individual encounter histories were compiled using various sighting sources (field observations, feeding observations, ringing data, breeding data), and bird IDs were cross-referenced with the studbook to confirm their origin and age. From the master dataset of resighting histories, some individuals were excluded from the analysis if we believed they could bias survival estimates due to their atypical life history or conservation management actions: individuals may have been ringed as adults (unknown age), never successfully fledged (died in the nest), translocated (N =151), taken into captivity (N = 4), or known to have bred prematurely <2 years (N = 2). The last cohort was removed (N = 120) due to no recapture events after 2022. This filtering process resulted in a final dataset of 2,734 known-age individuals, of which 997 (36.4%) were resighted in at least one subsequent breeding season following their initial ringing event. Encounters at supplementary feeding stations were excluded to maintain the assumption of equal-resighting probability among individuals (Pollock et al., 1990) that could potentially bias demographic parameter estimates if violated. Birds nesting closer to feeding stations aggregate around them (Fogell et al., 2019), creating spatial heterogeneity in detection probabilities. Accurate state assignment is critical for multistate models (Conn and Cooch, 2009). In our study, individuals were classified as breeders (B) only when positively identified as parents at a monitored nest and actively participating in reproduction (where a breeding attempt is characterised by at least one egg being laid); all other resighted individuals were classified as non-breeding (pre-breeders (PB) if never recorded breeding, non-breeders (NB) if previously recorded breeding but not in the current season). Several factors support the reliability of this classification. First, the scarcity of natural cavities means that the majority of breeding attempts occur in monitored nest boxes, which were visited multiple times by field biologists each breeding season to confirm parent identities through unique colour ring combinations. Second, echo parakeets exhibit high nest site fidelity and low divorce rates, allowing cross-validation with records from previous years. Third, continuous surveys across the species' limited range (~41 km²) throughout the breeding season minimised the probability of undetected breeding attempts. Fourth, natural cavities outside monitored areas are unlikely to support successful breeding due to lack of maintenance during the non-breeding season and competition from invasive species for cavity access. Thus, we expect breeder resightings probabilities to be very close to 1. Moreover, the low number of unringed individuals encountered throughout the study period (S. Henshaw in litt. 2026) provides indirect validation of the monitoring coverage: if substantial unrecorded breeding were occurring, we would expect higher resighting rates of unringed birds, which was not observed. # Data from: Demographic responses to population recovery illustrated by 30-years of monitoring a once critically endangered parrot Dataset DOI: [10.5061/dryad.dz08kpsdk](doi.org) ## Description of the data and file structure Echo parakeet (*Alexandrinus eques*) encounter histories (1994 -2022) * N = 2734 individuals - State attribution: Pre-breeder = 'PB' / Breeder = 'B' / Non-breeder = 'NB' / Not seen = '0' ### Files and variables #### File: EP_encounter_history_multistate.csv **Description:**  ##### Variables * id: individual identifier * 1994: year * 1995: year * 1996: year * 1997: year * 1998: year * 1999: year * 2000: year * 2001: year * 2002: year * 2003: year * 2004: year * 2005: year * 2006: year * 2007: year * 2008: year * 2009: year * 2010: year * 2011: year * 2012: year * 2013: year * 2014: year * 2015: year * 2016: year * 2017: year * 2018: year * 2019: year * 2020: year * 2021: year * 2022: year * cohort: year fledged