Causes of the recent decline of a Lesser Kestrel ( <i>Falco naumanni</i> ) population under an enhanced conservation scenario

Lesser Kestrel Falco naumanni (Fleischer, 1818) populations have been one of the best monitored bird in Spain over last 70 years. These suffered a sharp decline between 1950s and 1990s. Since then, periodic censuses showed population increase that lasted until 2010s. In those years, numerous projects for recovery species were also initiated (some EU LIFE Projects, among others), which continued to present day. However, despite conservation efforts, Spanish population, includes about 40% European breeding has declined at rate 6% per year since 2012. this study, we analysed changes habitat size 12 colonies located La Mancha 2003 2021 order identify possible causes current decline. This colonial breeds old buildings, roof area (a proxy house size) was predictor accounted number pairs colony given year. addition, extent herbaceous crops (related prey availability) explained similar amount variance 2021, whereas 2003, availability large Orthoptera itself had significant effect but much lower size. The nestboxes affected positively only 1% variance. largely by decrease density orthopterans, their main prey. model with minimum Akaike Information criterion adjusted small sample (AICc) value included positive association negative rabbit (possibly through hyperpredation phenomenon). Other top models (i.e. ΔAICc ≤ 2 units) effects loss favourable land uses (pasture lands crops) on size, probably because caused reduction Orthoptera. use occurred less often areas protected Natura 2000 network, may thus contributed population. other measures such as installation not efficient expected. Las poblaciones de cernícalo primilla naumanni, han sido una las aves mejor monitorizadas en España durante los últimos años. Estas sufrieron un fuerte declive entre décadas 1950 y 1990. Desde entonces, censos periódicos mostraron incremento poblacional que duró hasta la década 2010. En esos años también se iniciaron numerosos proyectos para recuperación especie (algunos UE, otros), continuado actualidad. Sin embargo, pesar esfuerzos conservación, población española primilla, incluye alrededor del reproductora europea, ha disminuido ritmo anual desde este trabajo, analizamos cambios el hábitat tamaño colonias situadas identificar posibles causas actual declive. Esta cría edificios antiguos superficie tejados (una aproximación al casa) fue explicó número parejas nivel colonia año determinado. Además, extensión cultivos herbáceos (relacionada con disponibilidad presas) cantidad varianza 2021; mientras que, propia grandes ortópteros tuvo efecto significativo, aunque mucho menor anteriores predictores. El cajas nido afectó positivamente reproductoras pero sólo varianza. descenso gran medida por disminución densidad ortópteros, su principal presa. modelo mínimo valor AICc incluía asociación positiva área tejado negativa conejos (posiblemente fenómeno hiperpredación). Otros modelos tener cuenta (es decir, unidades) efectos negativos pérdida usos favorables suelo (pastos herbáceos) sobre colonias, probablemente haber provocado reducción ortópteros. Estos produjeron probabilidad zonas protegidas red y, tanto, esta puede contribuido conservación primilla. otras medidas como instalación nido, no fueron tan eficaces esperaba. is falcon feeds mainly insects cereal fields and, few exceptions, human constructions. fate its intricately linked development agriculture two millennia (Negro et al. 2020), thrived pests, Moroccan locust Dociostaurus maroccanus (Thunberg, 1815) periodically devastated crop Mediterranean basin Middle East (Latchininsky 1998), establishing strengthened partnership humans. Nevertheless, century, coinciding intensive advent widespread chemical inputs (Rhoades 1963), remarkable early 1990s (González & Merino 1990, Biber 1996). following decades, experienced very noticeable increases Southern Europe North Africa there even evidence an genetic diversity local level (Ortego 2007a). line these trends, International Union Conservation Nature Natural Resources (IUCN) assessed ‘Threatened’ 1994, ‘Vulnerable’ 2011 then ‘Least Concern’ (BirdLife 2018). Greece, Sicily Portugal grow steadily decade thanks support management action (Morganti 2019, Gameiro 2020, Tsiopelas 2021). Spain, however, hosts (Iñigo Barov 2010), extensive census revealed severe 2012 (Bustamante 2020). There consensus researchers threats grounds, include: foraging loss, due pesticide nesting sites (Biber 1996, Atienza Tella Pilard 2009), are empirical studies explicitly any hypothesized factors (Sarà 2010, Morganti 2019). Rather, substantiation comes from indirect evidence. For example, selection generally agree preferences (Donázar 1993, Bustamante 1997, Parr 1998, 2004, Franco Sutherland Ursúa 2005, Rodríguez 2006, cf. Catry 2012) or foundation (Calabuig it necessary investigate relationships sizes, examine whether constrained demonstrate threat direct cause With regard pesticides, several Kestrel, seem be mild recent years insufficient explain Vergara 2007, Goutner 2015). More importance use, scarce refer occasional episodes insecticides currently banned 2007b). To our knowledge, can decline, herbicides used kill ruderal vegetation growing along roadsides, indirectly affect insects. chemicals, nitrogen fertilizers, widely used, shown strong impact abundance (Nessel 2021), so need analysed. case nest-site scarcity, contradictory. It seems intuitive could become problem viability medium long term 2007). fact, nest seemed limit rapid growth 2000s (Franco 2005), whereas, neighbouring available nest-sites normally higher than kestrel pairs, resource decreasing time (Forero factors, climate change 2019; see 2003) degradation quality winter quarters Sub-Saharan Africa. If so, would expect differences migrant resident kestrels, (Buchan (EU) network established decades safeguard habitats biodiversity interest under Birds Directive (79/409/EEC) Habitats (92/43/EEC). Many designed preserve agro-ecosystems, maintenance particular, (considered time). great concern latest after establishment reinforcement directly targeting species, projects, well national, regional actions captive breeding, reintroduction, construction ‘primillares’ (small buildings pigeon lofts, provided holes Kestrels), nestboxes, etc. raises question identified adequately addressed programmes. examined Mancha, Spain. both same sites, obtained detailed data around each colony, building conditions Orthoptera, feed nestlings (Rodríguez 2008; P.J. Cordero, A. Gómez J.M. Aparicio, prep.). aim contrast three causes, mutually exclusive, explaining changes: habitat, nest-sites. course found dramatic Oryctolagus cuniculus (Linnaeus, 1758) point being considered emerging pest some districts (Barrio 2011). Their high levels structure, accelerating collapse attracting numbers birds Although needs further explanatory variable analyses. role preserving study area. conducted provinces Ciudad Real Toledo, Central (39°20′N, 3°15′W, 600–800 m asl). maximum temperature averages 32–35 °C summer frequently exceeds 40 °C. Mean temperatures range 24–26 July 4–6 January. Annual precipitation typically 300 400 mm, concentrated spring autumn (Perez González Sanz Donaire 1998). approximately 1000 km2 set plain agriculture, cultivated barley, wheat vineyards. minor include scattered olive groves, almond pistachio trees, pine plantations pasture lands. area, carried out thorough monitoring 1991 2007 (Hernández 2007) revisited 2021. Kestrels form up 60 abandoned farmhouses. tiled roofs (94%) proportion, inside wall (6%) studied colonies. Therefore, measure potential nests particular suitable breeding. colonies, 14% bred artificial nestboxes. migrants arrive January March, depending age 2008), laying six eggs clutch Egg-laying takes place late April May (range 15 April–6 June; Aparicio Bonal 2002). Diet based insects, particularly fed (Bonal 2008). April, before egg-laying surveyed take first period, usually stay near sometimes go into hole, making easy locate count colony. Surveys 11:00 17:00 h detect kestrels they feeding surroundings middle Two observers thoroughly telescope 1 h, activity assess activity. A survey session ceased when new pair appeared 30–40 min. procedure repeated times different days. From 30 onwards, inspected apparently appropriate especially pellets, feathers vestiges indicating presence (Aparicio Land all parcels within km We distance 80% captures radius agricultural parcel determined visually driving walking March noting cadastral maps. information processed using digitized maps ArcMap. SIGPAC, Digital Parcel Identification System (LPIS) existing aids (https://sigpac.mapama.gob.es/fega/visor/) downloaded databases JCCM (accessible http://pagina.jccm.es/agricul/sigpac_datos/localidades.php). satellite images Sentinel-2, provides spatial temporal resolution 10 5 days, respectively, discern arable distinguish fallow land. classified nine uses: ‘Herbaceous crops’ (including cereals leguminous crops), ‘Fallow lands’ (which season ploughed), ‘Vineyards’, ‘Olive groves’, ‘Nut trees’ (pistachio trees), ‘Pine plantations’, ‘Pasture lands’, ‘Abandoned previous 3–10 years) ‘Non-productive (on soils without Orthoptera; roads, solar farms, water salt crust lagoons). sampling Sampling 16 June July, 10:00 18:00 above 25 plots 1-km previously nestling delivery took sampled 10–14 proportional coverage. added before, crops. variant Onsager Henry's (1977) method estimate yields reliable estimates people. consists putting wire rings 0.1 m2 (35.68 cm diameter) spaced interval transects. delimit grasshoppers seen counted person sufficient clarity motion outside circle. plot placed 48 net lines eight rings, 4-m intervals lines. Each patch group people, counting 1-m-wide transects rings. As methods, transects, reported (R2 = 0.91, n 123, P < 0.0001), transect method, time-consuming does require Transects (60 long, measured GPS) four who walked them (10 apart) time. They jumping/moving side path. At transect, recorded sex, whenever possible, specimens body estimated visually. according (< 0.5 g), (0.5 g) (≥ although food (P.J. Rabbit latrine counts certain track length (Virgós 2003). Relative rabbits Any sign (warrens, tracks, faecal scratches observation animals) latrines (Palma 1999) recorded. index latrines, burrows presence/absence signs 100 area). ranged 0 5, where absence plots, five track, 3 more sapling 4 plot, plot. mean indices standardized relative annual maximum. Population sizes polynomial functions increasing highest R2 obtained. paired t-tests comparisons colony) chi-square tests compare frequencies type network. Multiple linear regression analyses associated cover characteristics affecting latter case, evaluated independent variables (see Table complete details) predictors, increment analysing period. Changes logarithm ratio, i.e. log (value 2021/value Some fraction (herbaceous and/or land) strongly correlated cases model. Beyond predict dependent variables, goal determine predictor. made sense separate steps: comparison (Azen 2001, Wang 2013). step, generalized (GLMs) normal distribution errors identity link compared Null models) involved predictors. Akaike's Criterion select units, Burnham Anderson 2002) candidate set, lowest (Burnham account parameters allows us checked overfitting calculating predicted selected leave-one-out cross-validation summarizes model's ability data. Predicted calculated removing dataset, estimating equation determining how predicts removed observation. substantially indication overfitted excessively individual points (Deforest statistical significance R2, tested bootstrap iterations. existence multi-collinearity inflation factor (VIF), varies (when uncorrelated) infinite. VIF values 1.21. falls below 10, commonly cut-off indicative excessive (O'Brien 2007), fairly confident independence allowed apply Pratt (Pratt 1987, Thomas R-square statistic useful contribution collinearity (Liu 2014, Thus, proportion multiplying obtain conservative explained. statistics, module continuous distributed. coefficients GLM performs diagnostics index. All done SPSS v.26 resampling performed Excel (Microsoft) VBA macros. increased throughout decreased fit third-degree (Adjusted 0.86) better second- fourth-degree polynomials 0.68 0.75, respectively), peak (Fig. 1). increased. average 19.8 (± 5.4 se), 12.0 3.1 se) (paired t-test: t −2.71, df 11, 0.02) total 39.4%. rural buildings. Nine managed owners deterioration followed natural process. average, 23% disappeared 20 2). owners, fates. original almost completely demolished cases, walls. acquired nature (A.N. ESPARVEL), repairing poorer condition maintaining Kestrels, 2003. 5.67 4.13 9.83 4.34 8, 15, 24 respectively. already lost replaced. Finally, maintained 50–60 during Arable still occupied most territory 3). changed dramatically. Overall, land, nut tree (almond pistachio), trees non-productive (occupied farms photovoltaic power stations) increased, 3a). total, 142 108.7 lost, 6.7% percentages represent losses 12.1% 34.7% unevenly 3b). Shifts (nut crops, likely if property ecological (10.2% vs. 32.7%; χ 7.63 ; 0.006). ≥ 1.0 hectare once harvested stubble 4). Comparing densities (mean increase: 0.005 ± 0.004, 81, 0.24). dropped significantly −0.0126 22, 0.005) transformed woody Rabbits moderate Unfortunately, did monitor attract attention structure or, indirectly, outbreaks parts uncultivated including established. Farmers put fences prevent invading cleared plants areas, changing (species composition structure). vicinity related (β 0.73, 4.76, 0.001) negatively −0.50, 2.93, 0.017; Regression model: r2 0.79, F2,9 16.9, 0.001). predicting ‘Roof Area’ (‘Large_Orthop_1km’). cannot rule ‘number nestboxes’ predictor, (Table GLMs Area’,