Poaching sources and trade routes in Peru and Ecuador warn of the unsustainable rural demand for preferred parrot species

Illegal wildlife trade remains highly active in the Neotropics, as indicated by the thousands of parrots annually sold in illicit city markets. However, little is known about where parrots are poached, whether certain parrot species are selected among those available in the wild, their trade routes, and potential conservation impacts. We conducted a large‐scale survey in Peru and southern Ecuador to simultaneously estimate the relative abundance of parrots in the wild and as household pets in rural areas, determine their origin (locally poached or bought at city markets), and measured the shortest distances to their native ranges and markets through the existing grid of roads. Household‐poached native parrots were found in 96% of the rural localities surveyed. Most pets were locally poached, with only 14% of them bought at markets. Parrot poaching was highly selective, with preferred species (mainly Amazon parrots and large macaws) being collected much more than expected given their abundances in the wild and attaining higher prices than the other species. Individuals that were moved away from their native ranges or bought at distant markets were of those species most preferred by people, and covered large distances (up to 1010 km), even crossing country boundaries. Our results differ from those previously obtained from city markets and seizures of illegally traded parrots in Peru, where preferred species were underrepresented. Local poaching and rural trade activities act at very large spatial scales and negatively affect the population trends of preferred parrot species. This unsustainable scenario is a challenge to the application of effective conservation actions aimed at halting poaching and illegal trade. These actions should focus on very extensive and remote rural areas throughout the Neotropics rather than just on well‐known markets located in large cities.


K E Y W O R D S
harvesting, illegal trade, parrots, pets, trade routes

| INTRODUCTION
Wildlife trade is considered a global conservation threat (Scheffers et al., 2019). The Convention on International Trade in Endangered Species of Wild Fauna and Flora (CITES) greatly contributed to the regulation of international wildlife trade to protect species from overexploitation since 1973 (Challender & MacMillan, 2014). However, many species continue to be illegally traded, both globally and domestically (Ribeiro et al., 2019;Rosen & Smith, 2010).
Parrots (Psittaciformes) are one of the most threatened avian orders (Olah et al., 2016), with approximately 30% of the 402 extant species currently threatened with extinction (IUCN, 2020). Moreover, parrots are among the groups of vertebrates with the largest proportion of species involved in the wildlife trade (Scheffers et al., 2019), as they are greatly sought after as pets and cage birds (Bush et al., 2014). CITES regulates the international trade of almost all parrot species, and a large proportion of individuals legally moved worldwide belong to rather abundant and nonthreatened species (Chan et al., 2021). The wild bird trade bans of the United States (1992) and European Union (2005), combined with factors, such as economic development, human population growth, expanding international travel routes, or cultural shifts, led to geographic changes in trade and substitution of wild-sourced by captive-bred parrots internationally traded Chan et al., 2021). However, international illegal trade persists among neighboring (e.g., Ortiz-von Halle, 2018) and distant countries in the case of some highly demanded species such as the gray parrot Psittacus erithacus (Martin et al., 2018). More concerning than illegal international trade appears to be illegal domestic trade, especially in the Neotropics (Ortiz-von Halle, 2018). A recent assessment by parrot experts suggested that approximately 70% of the Neotropical parrot populations were threatened by illegal domestic trade (Berkunsky et al., 2017). The fact that tens of thousands of parrots are annually sold at illicit markets in countries such as Bolivia, Peru, or Brazil is of great concern (Gastañaga et al., 2011;Herrera & Hennessey, 2007;N obrega-Alves et al., 2013). Moreover, these figures surely underestimate the actual extraction rates, as they do not account for the high mortality during capture, transport, and captivity before and after selling wildcaught birds (Baños-Villalba et al., 2021;Gonz alez, 2003) neither for the high percentage of poached parrots that remains undetected during surveys (Gastañaga et al., 2011). Furthermore, these surveys were mostly restricted to markets located in large cities and thus field studies aimed at estimating the number of parrots removed for self-supply or sold by poachers on a more local scale in rural areas, which could largely outnumber parrots sold in city markets (Romero-Vidal et al., 2020), are needed (S anchez-Mercado et al., 2021).
Despite its conservation implications (Berkunsky et al., 2017), little is known about several key aspects of the illegal domestic trade of parrots (S anchez-Mercado et al., 2021). Pires (2015a) and Pires et al. (2016) found that most parrots available for sale in city markets of Bolivia and Peru were poached within a radius of approximately 250 km, suggesting that traded individuals can be moved over highly variable distances. Most of these parrots were from species considered common in the wild (Pires, 2015b), although information on their actual abundances was not available. A more detailed study conducted in Peru showed that the majority of individuals illegally traded in city markets belonged to the few species for which harvest quotas for legal exportation were established, assuming that their international trade is sustainable because they are common species. Although approximately 10% of the traded species were threatened (Daut, Brightsmith, Mendoza, et al., 2015), these authors concluded that, overall, parrot poaching mostly focus on common species, with little impact on wild populations and threatened species (Pires, 2015b). However, estimates of parrot abundances in the wild, which are needed for assessing the true impact of poaching (Tella & Hiraldo, 2014), are rarely available (Marsden & Royle, 2015). Recently, Romero-Vidal et al. (2020) simultaneously evaluated the relative abundance of parrot species in the wild and as household pets in Colombia, finding that amazons and macaws, the most attractive species (assessed as a composite of body size, coloration, and ability to imitate human speech) were poached much more frequently than expected when considering their availability. These results confirmed parrot poaching as a selective pressure that might affect the population trends and conservation status of the species most preferred as pets (Tella & Hiraldo, 2014).
Here, we conducted a large-scale survey across Peru and southern Ecuador to assess the role of local poaching in providing parrot pets in rural areas. By estimating the relative abundances of parrot species in the wild and as pets, we assessed whether these species were poached in proportion to their abundance in the wild or not. Moreover, we investigated if pets were poached locally, transported from distant areas where they were poached, or bought at distant city markets. For transported individuals, we reconstructed trade routes through the available grid of roads. Attending to the selective poaching hypothesis (Romero-Vidal et al., 2020;Tella & Hiraldo, 2014), we predicted that: (1) parrot poaching is not random but directed toward the most preferred species, mainly amazons and large macaws, (2) the most preferred species are transported at greater distances, even far from their distribution ranges and markets, and (3) are more valuable and thus more expensive. Finally, if selective poaching and transport act at large spatial scales, (4) preferred species should have negative population trends that might reflect a more concerning global conservation status (Tella & Hiraldo, 2014). We also compared the proportion of preferred pet species found through our survey of rural areas in Peru with those obtained from previously published surveys conducted in city markets and seizures performed by authorities, to add insight on the difficulties of tackling the illegal parrot trade.

| Study area
We focused this study on Peru but also included southern Ecuador (Figure 1) because of the possibility of detecting long distance, transboundary trade of poached parrots among rural areas of both countries. The Andes separate the Pacific coastal region from the Amazonian region, creating a wide diversity of biomes across Peru and Ecuador. These countries are thus considered as the most biodiverse areas in the world (Myers et al., 2000), and hold approximately 15% of the extant parrot species (IUCN, 2020).

| Relative parrot abundances in the wild
Estimating parrot abundances in the wild is a challenging task, and there are a variety of methods with different pros and cons, the choice of which depends on the research goals of each particular study (Dénes F I G U R E 1 Study area covering Peru and southern Ecuador, showing the routes (black lines), roadside parrot surveys (red lines), and locations where household parrot pets were recorded (blue dots). Rivers and boundaries between countries are shown as blue and white lines, respectively. Localities with illicit pet markets are included. The point in Lima corresponds to a pet that was moved there by a rural pet owner (and poacher), as a present for familiars, from a long distance. Marsden & Royle, 2015;Tella et al., 2021). Following Romero-Vidal et al. (2020), we choose roadside surveys since this methodology allows to sample very large areas to increase the likelihood of recording parrots showing low densities and/or patchy distributions (Dénes et al., 2018;Tella et al., 2021). Using this methodology, the number of individuals recorded strongly correlates (r = .93) with density estimates accounting for differences in detectability among species (i.e., through distance-sampling modeling), and allows the calculation of relative species-specific abundances (individuals/km) for whole parrot communities, including the scarcer species that yield insufficient encounters for modeling detectability (see further details, strengths, and weakness of this method in Tella et al., 2021).
Road surveys were designed a priori, using recent satellite images (Romero-Vidal et al., 2020), in a stratified way to cover proportionally the biomes inhabited by parrots and habitats with different degrees of transformation (from pristine forests to agriculture areas and villages), to avoid biases in the recording of parrot species, as some avoid habitat transformations while others are more abundant in anthropogenic landscapes . We selected unpaved and lightly traveled roads, often accessible only by 4 Â 4 vehicles, to avoid potential traffic disturbance on parrot detection. Using this approach, Tella et al. (2021) found that the relative abundances of species were not correlated with their distance-dependent detection probabilities, showing that less commonly encountered species are actually uncommon and not less detected because they might avoid the proximity of roads. The stratification carried out a priori proved to be a difficult task, as it was limited by the distribution of the roads, and required about 1 month of work with satellite images before each expedition. Moreover, once in the field, we found that some roads were not passable even with a 4 Â 4 vehicle, as they were cut off by landslides or flooding. Despite these logistical constraints, we were able to survey 3725.21 km between 28 November and December 18, 2014 and between 7 and October 31, 2018, covering the study area ( Figure 1) and the surveyed biomes reasonably well (see Table S1 for effort distribution among biomes and ecoregions).
Car surveys were conducted by 4 persons during early mornings and afternoons, driving at a low speed (20-40 km/h) and stopping, when needed, to identify the parrot species and count the number of individuals (flock size). When parrots were aurally detected but not visually recorded, we used the average flock size recorded for the species for analyses (Romero-Vidal et al., 2020;Tella et al., 2021). As in previous works (Romero-Vidal et al., 2020;Tella et al., 2021), we found a strong correlation (r = .90, n = 23, p < .001) between our estimates of relative parrot abundances (individuals/km) and densities (individuals/km 2 ) obtained through distancesampling modeling (see details and Figure S1), thus supporting that our parrot counts without correction for detectability are good proxies of their actual abundances.

| Relative abundances of poached parrots
We surveyed the rural human settlements (excluding isolated houses and clusters of less than 10 houses because they housed very few people) that were encountered through the road surveys to record the presence of parrot pets and thus have estimates of the relative abundances of wild and poached parrot species in the same areas. This methodology proved to be useful for estimating poaching pressure in rural areas where this activity is not actively prosecuted and thus poached pets are not hidden (Romero-Vidal et al., 2020). The time employed in each rural human settlement (0.5-3 h) was positively related to its size, more sampling time spent in larger populations. The streets surveyed varied according to the spatial structure of the settlements. In settlements with houses distributed along the road, rectilinear sampling followed the main road, while in those with more conventional distributions, the surveys were made using parallel streets to increase the area sampled. Most of the localities encountered (98%) were nonindigenous.
A high percentage of the parrot pets we recorded were displayed on streets or in public establishments (e.g., stores, hostels, gas stations), while others were hidden inside the houses and detected aurally. In both cases, we met the owners and informed them of our interest in wildlife and its use as pets, and asked for their consent to see their pets and learn more about them. In some cases, owners of the pets on display also had other pets not exposed (e.g., inside their homes) and showed them to us. Pet owners often introduced us other people, friends or familiars, who also kept parrot pets. The fact that we speak their native language (Spanish) made it easier for them to engage in friendly conversations, voluntarily telling us details about their pets. While the time spent in these unstructured conversations largely varied among owners (usually between 5 and 20 min), we ensured to ask them the following information: (1) whether pets were poached locally by them or bought from others, (2) how parrots were poached (caught as adults or taken as chicks from their nests), and (3) for the parrots that were bought, whether they were bought from other poachers or at markets, and how much they paid for them. In no case did we show people the intention of buying parrots or any other type of poached fauna, to avoid encouraging this illegal activity. We did not record any information about the people and all remained anonymous. Therefore, no sensitive data or personally identifiable information (PII) was collected. This procedure was approved by the Ethical Committee of our research institution (Comisi on Etica para Investigaci on con Seres Humanos, CEIH, Universidad Pablo de Olavide, code 21/8-3).

| Distances to distribution ranges and markets
Using the existing grid of paved and unpaved roads (available at: https://portal.mtc.gob.pe/estadisticas/descarga. html, https://geonode.wfp.org/layers/ogcserver.gis.wfp.org %3Ageonode%3Aecu_trs_roads_osm), we measured the shortest distances from the georeferenced locality where we recorded each poached parrot pet to the closest point of its native distribution range (available at: http:// datazone.birdlife.org/species/requestdis). When parrots were bought in markets, we also obtained the shortest distance to the market of origin. The main Amazonian and coastal markets were identified from previous studies (Daut et al., 2015a;Gastañaga et al., 2011;Shanee et al., 2017) (Figure 1). When the identity of the market was not provided, we conservatively measured the distance to the closest one. Also conservatively, the distance to the distribution range was set as 0 km for parrots poached within or at less than 10 km from the edge of its distribution range. QGIS 3.4.12 and ArcGIS 10.3.1 were used to calculate distances.

| Conservation status and legal trade
We used the most recent IUCN Red List (IUCN, 2020) to extract and score the global conservation status (0: Least Concern, 1: Near Threatened, 2: Vulnerable, 3: Endangered, 4: Critically Endangered) and population trend (1: decreasing, 2: stable, 3: increasing) of each parrot species. We did not use the national conservation status of the species for analyses as the Red Book of Peruvian wildlife (SERFOR, 2018) only assessed the conservation status of 9 out of the 40 parrot species studied. We obtained the species with harvest quotas for their legal international trade according to CITES from Salinas (2014), actualizing taxonomic changes (IUCN, 2020) and accepting the split of the mitred parakeet Psittacara mitratus into 3 different species (Arndt, 2006) as supported by our own field observations.

| Statistical analyses
To achieve our objectives, which require a very broad spatial scale, analyses were performed at the level of the entire study area. Analyses at smaller scales (e.g., considering Pacific coastal, Andean, and Amazonian regions as distinct units) would bias results, as (a) poached parrots transported at long-distances would be excluded (especially affecting positively selected species, see Results), and (b) pets found far from their distribution ranges should be removed from selection analyses as their availability in the wild could not be estimated. However, despite these important caveats, we also performed separate selection analyses per region (excluding species not present there in the wild), showing the consistency of results (see Supplementary Materials).
The Savage selectivity index (hereafter SI) allows for statistical testing of resources selection (Manly et al., 2007), and has recently been used to statistically test the selection of poached species for the pet trade Romero-Vidal et al., 2020). We calculated SI for each species, using the number of individuals recorded in the wild as the available resource and the number of individuals recorded as pets as the used resource (Romero-Vidal et al., 2020). SI values close to 1 indicate nonselection (i.e., the use of the resource is proportional to its availability), while statistically significant values below and above 1 indicate negative and positive selection, respectively (Manly et al., 2007). Bonferroni correction for multiple tests was applied to calculate the statistical significance of SI (see Romero-Vidal et al., 2020 for details). We tested differences in SI between amazons (genus Amazona), large macaws (genus Ara), and the remaining species using a Kruskal-Wallis test and post hoc pairwise Wilcoxon rank-sum exact tests (applying Bonferroni correction). We used Mann-Whitney tests to assess differences in SI between species with and without legal harvest quotas, as well as between prices of selected and non-selected species, and differences between countries. We evaluated whether or not the pets were moved (response variable: 0.1) in relation to their SI using generalized linear mixed models (GLMM) (binomial error distribution and logistic link function; random term: species). We used generalized linear models (GLM) to assess whether the SI of a species was related to its price (transformed to US$; normal error distribution and identity link function), conservation status (rank: 0-4), and population trend (rank: 1-3) (multinomial error distribution, cumprobit link function). The fit of models were checked using DhARMA (Hartig, 2018). Statistical analyses were performed using R 3.5.2 (R Core Team, 2018). Differences between proportions were tested using chi-square tests.
Most parrots (86.1%) were locally poached by people and kept as their own pets (71.6%) or bought from other local poachers (28.4%), without entering markets. Only 13.9% of the parrots were bought in markets, with just 0.3% corresponding to coastal markets (2 parrots in Chiclayo and one in Lima) and the rest to Amazonian markets ( Figure 1). Poachers mainly extracted chicks from their nests (62.4% of pets), but also captured adults in crops with traps and nets (37.6%).
F I G U R E 2 Poaching selection (Savage index, SI; black dots) and relative abundances (%) in the wild (red bars) and as household pets (blue bars) for each parrot species. Species are grouped as positively selected (statistically significant SI values > 1, indicated by the vertical dashed line), negatively selected (statistically significant SI values < 1), and nonselected (SI values not statistically significant). *Species recorded in the wild but not as pets.

| Parrot poaching selection
We recorded 7475 wild parrots from 40 species through the roadside surveys. Relative abundances of parrots in the wild greatly varied among species, and just 8 species summed 82% of all recorded individuals (Table S1). The SI showed a nonrandom poaching pressure. While the most abundant species were negatively selected, several of the uncommon species were poached more than expected (i.e., positively selected) attending to their relative abundances in the wild (Figure 2, Table S2). SI did not differ between species with legal export quotas and the rest of the species (Mann-Whitney, U = 99.00, p = .58), with 2 of the former species being positively selected (Brotogeris sanctithomae and Psittacara frontatus, Table S1). As expected, amazons (genus Amazona) and large macaws (genus Ara) were the most positively selected species (Kruskal-Wallis, H = 19.12, df = 2, p < .0001; Figure 3; see Tables S3, S4 and S5 for analyses performed separately for the Pacific coastal, Andean, and Amazonian regions). Pets of these most preferred species were more often hidden (i.e., not displayed on streets or public establishments) than those of the other species (61.1 vs. 37.8%; χ 2 = 9.95, df = 1, p < .01, n = 979). While all amazons and large macaws were nestlings poached in their nests, 24% of the pets of other species were trapped as adults (χ 2 = 25.05, df = 1, p < .0001).
In the case of Peru, the proportion of amazons and large macaws we found poached in rural areas was higher than previously found by other studies in city markets (n = 31,197 parrots, years 2007-2011) and among parrots seized by authorities (n = 22,213 parrots, years 2007-2021) (Figure 4a, χ 2 = 19.57, df = 4, p < .001). These differences could derive-at least partially-from the temporal mismatch between the 3 sources of information. However, the results are similar when comparing our data obtained in 2014 and 2018 with seizures made in the same years (n = 3222 parrots) (Figure 4b, χ 2 = 14.16, df = 2, p < 0.001).

| Parrot trade in rural areas
Parrot pets not obtained from markets were mostly individuals poached locally (i.e., those kept as pets within their native distribution range; n = 612), although we also detected some parrots poached in distant areas (n = 145; range of distances to their distribution range: 19.38-1009.79 km). The probability of moving parrots out of their native ranges increased with the selectivity index: household pets that were poached locally had a lower SI (mean = 7.65; SD = 19.62) than those moved from distant areas (mean = 35.20; SD = 40.48; Figure 5a). On the other hand, parrots bought in markets were obtained in local or distant markets (distances to markets: 0-1048 km; Table S1), the largest distances corresponding to 2 macaws (Ara ararauna and A. macao) moved from Peru to Ecuador. Most of these bought parrots (99.7%) came from the Amazonian markets (Figure 1), and were mainly from species with high SI (mostly large macaws and some amazons). Parrots bought at local markets had F I G U R E 3 Boxplot of median selectivity indexes for amazons (genus Amazona), large macaws (genus Ara), and the other parrot species recorded as pets. The red line indicates nonselection, with values above or below 1 indicating positive or negative poaching selection. Post hoc pairwise Wilcoxon tests with Bonferroni correction show no differences between amazons and large macaws ( p = 1), but significant differences between these and other species (amazons and other species: p = .0009; large macaws and other species: p = .0122).

F I G U R E 4 (a) Proportion of preferred (Amazona parrots and
Ara macaws) and nonpreferred (other species) pets found in our survey of rural areas (black bars), in previous surveys of city markets (orange bars), and in seizures made by authorities (blue bars) in Peru. (b) Comparison of data from rural areas and seizures restricted to the years in which we conducted the study (2014 and 2018). No information was published for city markets in these years. lower SI (mean = 39.64; SD = 59.46) than those obtained at distant markets (mean = 62.58; SD = 50.77), although differences were not significant (z = 0.43, p = .670). Although not statistically significant (p = .06), the trend observed suggests that parrots bought from the most distant markets had the highest SI ( Figure 5b).
As predicted, when pets were bought, the prices of positively selected species were higher than those of the other species (Mann-Whitney, U = 5.0, p = .013). Moreover, there is a trend (not statistically significant; p = .06) suggesting that species prices increase with their SI (Figure 6a).

| Conservation status of poached species
Although SI was unconnected to the global conservation status of the species (z = À0.46, p = .643), it was related to their population trends, species with population declines showing higher SI than those with stable or increasing populations (Figure 6b).

| Unraveling domestic illegal trade
Together with habitat loss, the illegal pet trade is considered one of the major threats for parrots and other birds in the Neotropics (Berkunsky et al., 2017;Dayer et al., 2020), and much of the information on species poached, their origin, and the actors involved in this activity has come from surveys of illicit city markets in Peru (Daut, Brightsmith, Mendoza, et al., 2015;Gastañaga et al., 2011;Leberatto, 2016;Pires, 2015aPires, , 2015b. Although extremely F I G U R E 5 (a) Locally poached parrots locally kept as pets showed a lower selectivity index (SI) than those moved from their native ranges. (b) When parrots were bought at distant markets, individuals brought from afar had higher SI than those of closer markets (see Figure S2 and S3 for model fits).
F I G U R E 6 (a) The prices (in US$) of parrot species tend to increase with their SI. (b) Relationship between poaching selection (SI) and the global population trends of the species. The red line indicates nonselection (SI = 1), with values above or below 1 indicating positive or negative poaching selection respectively (see Figure S4 and Supplementary Material for model fit). valuable, these studies lack a wider approach, which we overcome here by conducting a large-scale survey of household pets in rural areas across Peru and southern Ecuador, shedding light on a new scenario with different types of trade. Our results show that markets provide only a small fraction of pets (approximately 14%) to rural areas, whereas most of them are directly poached by pet owners or by neighbors who sold the poached parrots in the same or nearby localities. This poaching and trade system predominates in the Andean region and the coastal regions of Tumbes (Peru) and Guayas (Ecuador), where most poached parrots remain in the same areas (see also Biddle et al., 2021). However, parrots poached in Amazonian areas inhabited by people often remain as local pets but are also traded to other rural areas and regions. We also confirmed that some parrots poached in remote Amazonian areas by indigenous people (Gonz alez, 2003) are transported by middlemen through rivers to small Amazonian markets. These parrots are then sold locally, distributed to other rural areas, or transported to markets of coastal cities such as Lima, mostly to supply the urban demand. This domestic parrot trade seems to operate since pre-Columbian times, following routes from Amazonian areas to Andean mountains or even coastal areas (Capriles et al., 2021;Guaman Poma de Ayala, 1980) similar to those recorded now for parrots and other traded taxa such as primates (Shanee et al., 2017).

| Differences in poaching pressure and transport among species
Previous surveys and analyses obtained from illicit markets lead to the idea that parrot poaching is an opportunistic activity mainly focused on the most common and available species (Daut et al., 2015a;Dayer et al., 2020;Leberatto, 2016;Pires, 2015b;Pires et al., 2021), despite of the lack of adequate information on the abundance of these species. When the relative abundance of traded species was measured in the wild, however, the scenario changed dramatically toward the selective poaching of certain species both in the Neotropics (Romero-Vidal et al., 2020;present results) and Southern Asia . In the Neotropics, Amazons and large macaws are the most preferred species due to their coloration, size, and mimicry ability (Romero-Vidal et al., 2020;Tella & Hiraldo, 2014), and were poached much more than expected and reached higher prices than the rest of the species in Peru and Ecuador. On the other hand, the scarcity of preferred species often provokes their replacement with less attractive ones (Luna et al., 2018), such as small parakeets, which are more abundant and affordable.
These species appear in larger numbers than preferred species both in our survey of household pets and in previous market surveys (Daut et al., 2015a;Pires, 2015a), but much less than expected given their abundances in the wild. Moreover, some common species are persecuted as crop pests  and thus are killed but also live-trapped for selling (Herrera & Hennessey, 2007). We found that in the Andean valleys, where agriculture is the main subsistence economic activity (Quintana et al., 2019), some unattractive but common species (e.g. Psittacara) are trapped by farmers both to protect their crops and obtain additional income by locally selling them as pets. The fact that these species are easier to catch in large numbers with traps and nets than the more preferred species, which are always collected from nests, may also contribute to their lower prices.
Poached parrots were not randomly transported among rural areas. As we expected attending to their attractiveness (Romero-Vidal et al., 2020;Tella & Hiraldo, 2014), preferred species were more often transported over larger distances, even crossing country boundaries. Moreover, the actual relationships between poaching selection and distances moved are probably stronger than what we reflect in our analyses. On the one hand, we measured the shortest distances through the main road network, while traffickers could travel much longer distances through unpaved tertiary roads to avoid cities and thus minimizing seizure risks of valuable species. On the other hand, we measured the shortest road distances to the supposed distribution ranges of species. However, we learned from our field surveys and local knowledge that the most preferred species are already absent from large areas due to overharvesting (Merkord et al., 2009). In fact, when we had informal conversations with local people, they often explained that certain species of amazons and macaws had disappeared from certain areas for years, and that they had to travel further to find them. Therefore, their current distributions may be farther (i.e., constrained to more remote areas) than those depicted in the available distribution maps. It is worth mentioning that the logistic constraints impeding us to survey the most remote Amazonian areas might have affected our poaching selection results. Although one could expect that the most remote and inaccessible areas to people could hold densest populations of both selected and non-selected parrot species, we know from this study that selected species tend to be moved over larger distances. Therefore, we cannot discard that the selective transport of some preferred species from remote Amazonian areas could have inflated their SI. More research is needed to assess the poaching pressure, abundances and conservation status of parrots in these and other Neotropical remote areas.

| Comparison of rural areas with city markets and seizures
We found that preferred species are more prevalent as pets in rural areas than previously reported from surveys at city markets and seizures in Peru. These differences do not seem to be the result of a lower demand for these species in cities (where social networks also helps to acquire wild pets unlike in rural areas, where most pets are poached locally), but of a lower detectability in city markets (2.5%-3%, Gastañaga et al., 2011). Indeed, the most valuable species are often hidden to keep them out of sight of the authority (Ortiz-von Halle, 2018), with sellers less concerned about the loss of inexpensive species, such as Brotogeris versicolorus, which are often displayed in large numbers in markets without hiding them from the authorities (Daut et al., 2015a).
Seizures reflect the proportion of species exposed in urban markets, as most enforcement and confiscations are focused in cities (Gastañaga et al., 2011;Ortiz-von Halle, 2018). Prosecution contributed to stopping or reducing trade in some markets (Gastañaga et al., 2011), but not in rural areas, as confirmed by our conversations with pet owners in these areas. Although they kept the most valued species most often inside their homes, it was not out of fear that these pets might be confiscated but rather that they might be lost or stolen by others. In fact, most people were confident and seemed unaware that they had illegal pets, and none of them mentioned a single case of household pet confiscation in rural areas.

| Addressing the uncertain sustainability of parrot poaching
Our survey emphasizes previous concerns about the impacts of the illegal domestic parrot trade in the Neotropics (Berkunsky et al., 2017;S anchez-Mercado et al., 2021;Wright et al., 2001). When surveying rural areas, we found poached pets in 96% of the surveyed localities, a figure similar to that found (93%) in rural areas of Colombia (Romero-Vidal et al., 2020). Moreover, Biddle et al. (2021) found in one of our surveyed rural Ecuadorian areas, through well-designed questionnaire interviews, that 66% of the communities had parrots as pets and 66% of people in these communities had owned them. It is important to note that most of the pets we recorded were locally poached and only small percentages (14%) were purchased from markets. Therefore, the estimated minimum number of 80,000-90,000 parrots sold annually in Peruvian city markets (Gastañaga et al., 2011) represents only a small fraction of all parrots poached annually, even more so taking into account the high mortality rates caused by the harvesting activities (Gonz alez, 2003).
Our results show that local poaching is widespread across rural areas of Peru and Ecuador (see also Romero-Vidal et al., 2020 for Colombia) and that preferred species are demanded from areas far from their distribution ranges, being transported over distances much larger than those previously estimated using straight-line buffers around markets (Pires, 2015a). Selective poaching could therefore affect not only population trends of restricted-range endemics (e.g., Ara ambiguus guayaquilensis, Amazona lilacina; Biddle et al., 2020Biddle et al., , 2021 but also of widely distributed species, given the large spatial scales of poaching and demand from distant areas. Furthermore, these effects may be difficult to track, as several species may be limited by the availability of nesting sites (Hesse & Duffield, 2000;Olah et al., 2014;White Jr et al., 2005) and, if large trees with nesting holes are not removed, these long-lived species may be resilient to the removal of nestlings, despite having compromised the recruitment of individuals. The potential impacts of selective poaching on population trends of large macaws and amazons are likely concomitant with habitat loss (Berkunsky et al., 2017), leading to a scarcity of large trees needed for nesting, facilitating poaching by making the location of their nests easier to access by poachers. Moreover, their lifehistory traits (i.e., greater longevity and low productivity associated with large body sizes and long generation times) make them particularly vulnerable to extinction (Olah et al., 2016), as shown for mammals hunted for bushmeat in the Peruvian Amazon (Bodmer et al., 1997). However, we did not find a relationship between poaching pressure, measured through the SI of each species, and their global conservation status. This may be due to the large estimated distribution ranges of many species, which reduce their threat status when evaluated by the IUCN. Information on actual fine-scale ranges is urgently needed for proper IUCN Red List assessments, taking into account the inadvertent range contractions of many parrot species due to wildlife trade and the relationship between population declines and selective poaching reported here. The last relationship does not imply causality, as multiple causes may contribute to the decline of Neotropical parrot populations, but it suggests that poaching may play a key role, as noted by expert opinion (Berkunsky et al., 2017). However, more research efforts are needed to better understand the actual impact of poaching on population declines, as little is known about the sustainability of this illegal activity and its effects on population dynamics and viability (Beissinger & Bucher, 1992;Carvalho-Roel & Júnior, 2021;Valle et al., 2018).
Concern should be also extended to the sustainability of the trade of nonpreferred and still common species in the long term. Peru established harvest quotas for the legal international trade of some species, which varied with years attending to sustainability criteria (Salinas, 2014). However, the numbers of individuals from these species illegally sold in city markets largely exceed the quotas for legal international trade (Daut, Brightsmith, Mendoza, et al., 2015) and are even higher when considering the large local poaching and trade in rural areas. Indeed, our results add to previous concerns about the potential unsustainability of quotas allowed for legal international trade of two species, Brotogeris sanctithomae and Psittacara frontatus (Ortiz-von Halle, 2018), as we found that they are poached for illegal rural pet demand more than expected given their abundance in the wild.

| Challenging conservation actions
Although domestic parrot trade was banned in most Neotropical countries for decades (Ortiz-von Halle, 2018;Romero-Vidal et al., 2022), parrot poaching continues to flourish in Peru and, to a lesser extent, Ecuador. Law enforcement appears to be stronger in Ecuador (Mestanza-Ram on et al., 2020), which could explain a lower number or pet parrots kept per house there than in Peru. SERFOR formulated in 2016 the National Strategy to Reduce Illegal Wildlife Trade in Peru (2016-2021) for law enforcement and actions against wildlife crime (Ortiz-von Halle, 2018). Enforcement is key to halt wildlife trade (Holden et al., 2019) and can be more effective when it is geographically predictable, such as trade through city markets. In fact, authorities annually seize thousands of parrots there (Daut, Brightsmith, Mendoza, et al., 2015, this study). However, our results show that illegal pet keeping is widespread in rural areas, where only a very small fraction of pet parrots come from markets. Pet keeping seems to be a rural tradition (Ortiz-von Halle, 2018) similar to bushmeat consumption, another harvesting activity that translates to cities (Chaves et al., 2020). This deep-rooted tradition, coupled with low investment and high corruption in governmental agencies (Daut et al., 2015b;Shanee et al., 2017), decentralizing governance of territories (Ortiz-von Halle, 2018), and limited knowledge and high turnover of wildlife officials (Daut et al., 2015a), greatly hinder effective prosecution of illegal wildlife trade in very large rural areas. Moreover, the few available rescue centers are insufficient to house all seized parrots (Leberatto, 2016). Therefore, they are often released into the wild by authorities without considering their condition and origins, even causing introductions into non-native ranges and potential diseases transmissions to wild populations (Daut et al., 2016). Although we found two conservation NGOs working on the conservation of highly endangered parrot species in rural Ecuador, there is no evidence for such activities in rural Peru despite the existence of several Peruvian NGOs concerned about wildlife trade (Daut et al., 2015b). Therefore, greater involvement of both authorities and NGOs in rural areas should be promoted.
Several complementary actions are needed to reduce the parrot poaching in Peru and Ecuador, which could be extended to other Neotropical countries that are experiencing similar situations (S anchez-Mercado et al., 2021). As for subsistence hunting (Bodmer & Lozano, 2001), which is often interconnected with domestic trade (S anchez-Mercado et al., 2016), some rural development actions could compensate for the economic benefits of parrot poaching. Ecotourism has proven to be an effective conservation tool in Peru (Brightsmith et al., 2008). In fact, most of the preferred parrot species we recorded in the wild were within or close to ecotourist and protected areas, according to the fact that parrot poaching is lower in areas under protection (Wright et al., 2001). Moreover, programs aimed to compensate for economic losses due to crop damage (Barbosa & Tella, 2019) could reduce trapping for trade as a by-product of parrot persecution. However, pet keeping is a trans-cultural rooted tradition (Romero-Vidal et al., 2020;Tella & Hiraldo, 2014) difficult to manage, so urgent actions directed at changing people's behaviors to reduce the demand for wild pets are needed (S anchez-Mercado et al., 2020(S anchez-Mercado et al., , 2021. Captive breeding and propagation have been suggested to counter illegal trade (Fukushima et al., 2021), and exotic pet parrots have replaced poached natives in rural areas of Colombia where awareness campaigns and law enforcement acted (Romero-Vidal et al., 2020). A few exotic parrot species are widely available in large cities (Daut et al., 2015a; authors pers. obs.), although not in rural areas of Peru and Ecuador, where exotic captive-bred parrots only reached 2% of all household pets we recorded. However, there are risks associated with these species. Although the risk of invasion of captive-bred parrots is very low compared to wild-caught ones (Abell an et al., 2017;Carrete & Tella, 2008), these exotic species can spread diseases such as the Beak and Feathers Disease Virus, the Avian Polyomavirus, the Avian Bornavirus, and the bacterium Chlamydiophila psittaci, which are among the most common in captive parrots (Friend et al., 2001;Wellehan et al., 2016). Perhaps more efficiently, awareness campaigns should be promoted to reduce the demand of wild parrot pets through cultural change in the medium-long term (S anchez-Mercado et al., 2021). For instance, schooling resulted effective to fight against wildlife hunting in the Bolivian Amazon (Luz et al., 2015). However, this may also be a difficult task, as shown by S anchez-Mercado et al. (2020). These authors found that, after three decades of closely working with Venezuelan communities to halt the illegal harvesting of a threatened amazon,, people perceived negatively the sale but not the poaching of individuals for their personal use as pets. Given the complex scenario of parrot poaching and illegal trade, greater efforts both in law enforcement and from the emerging field of conservation social science (Dayer et al., 2020) are urgently needed to avoid overharvesting and the collapse of species populations. These actions would also help to reduce the illegal trade of many other taxa (Daut et al., 2015a;Shanee, 2012), as Peru and Ecuador are major global hotspots of biodiversity but also of wildlife trade worldwide (Scheffers et al., 2019).

ACKNOWLEDGMENTS
This paper is based on Chapter 4 of the lead author's PhD Dissertation (https://rio.upo.es/xmlui/handle/10433/11606), without infringements of copyright by this fact. We thank I. Af an (LAST-EBD) for helping in the design of road surveys using satellite maps and three anonymous reviewers for their valuable suggestions who contributed greatly to the improvement of the manuscript. Logistic and technical support were provided by Doñana ICTS-RBD. This work was funded by Loro Parque Fundaci on (Project SEJI/2018/024). Funding for open access publishing: Universidad Pablo de Olavide/CBUA. Comments provided by three anonymous reviewers greatly helped to improve the manuscript.

CONFLICT OF INTEREST STATEMENT
The authors declare no conflict of interest.

DATA AVAILABILITY STATEMENT
All data analyzed in this article are provided in the Appendix. All roadside surveys and parrot abundances in the wild were recorded using the ObsMapp (Observation International, Amsterdam, The Netherlands) application for smartphones, which uploads the observations to the citizen science platform Observation (www.observation. org, Observation International, Amsterdam, The Netherlands). Therefore, all records, exact location, and associated information can be viewed and downloaded (looking for the observer Pedro Romero-Vidal).