Forest Degradation and Weather Jointly Affect Early‐Life Development in a Tropical Understory Bird

TL;DR

Forest degradation and hot weather together affect the growth and development of young birds in tropical forests, with intact canopies offering protection against climate stress.

Contribution

This study reveals how forest degradation and climate interact to influence nestling development in tropical birds, using a multi-metric approach.

Findings

Nestlings in smaller or degraded forest patches had lower body condition, especially during drought.

Corticosterone levels in larger forest patches reflect developmental stage rather than stress.

Intact high-canopy forests buffer nestlings against heat and drought, supporting better development.

Abstract

Tropical forest birds face mounting pressures from habitat loss, degradation, and climate warming, yet their combined effects on early‐life development remain unclear. Using over a decade of morphological and behavioural observations from Kenya's Taita Hills and 4 years of nestling corticosterone measurements across eight forest patches differing in size and degradation, we examined how forest quality and climate shape nestling condition in the understorey insectivore placid greenbul (
Phyllastrephus cabanisi placidus). Nestlings in smaller or more degraded patches showed lower body condition. Provisioning rates did not vary with forest quality, suggesting that poor condition in degraded habitats may result from lower prey quality rather than reduced parental effort. Unexpectedly, corticosterone levels were higher in larger forest patches, and nestlings with elevated corticosterone also showed more advanced wing development, indicating that corticosterone here may reflect developmental stage (readiness to fledge) rather than condition per se. In addition, nestling condition declined in hot weather. Furthermore, although body extremities were shorter in small or degraded patches at low temperatures, in hot weather nestlings developed extremities similar in length to those in larger, less degraded forest patches, consistent with short‐term thermoregulatory plasticity. By contrast, in high‐canopy areas tarsus growth decreased with increasing temperature, highlighting that multiple developmental mechanisms may operate simultaneously. High canopy cover also buffered body condition under dry weather, whereas nestlings in low‐canopy habitats were particularly vulnerable during drought. Our findings demonstrate how forest degradation and climate interact to shape nestling growth, physiology, and potential thermal plasticity. Maintaining intact forest canopies and large forest patches is thus critical for preserving the microclimatic buffering needed to support offspring development in a warming climate. Because multiple, and sometimes opposing, developmental mechanisms may act simultaneously, integrating morphological, behavioural, and physiological indicators provides a powerful multi‐metric approach to disentangle how habitat and climate jointly shape early‐life development.

More than a decade of field data from Kenya's Taita Hills show that forest degradation and warming jointly shape nestling development in a tropical forest bird. Nestlings in small or degraded forest patches had poorer condition, especially during low rainfall, while intact high‐canopy forests buffered them against drought and heat. Wings and tarsi were generally shorter in degraded forest but reached similar length as in near‐natural forest during high temperatures. Corticosterone levels, higher in larger patches, reflected developmental stage rather than nutritional stress. These results highlight the key role of intact canopies in supporting offspring development under climate change.

Figures

Location of the Taita Hills within the Eastern Arc Mountain chain. Map (a) displays the main mountain blocks of the Eastern Arc, including their position within Africa (inset) and highlights the location of the Taita Hills among these mountain chains. Map (b) zooms in on the Taita Hills, emphasizing the isolation of the main Taita Hills blocks (Mbololo and Dabida Massif) from Sagala Hills, also part of the Eastern Arc Mountains. Map (c) shows the forest fragments within the Taita Hills. Map lines delineate study areas and do not necessarily depict accepted national boundaries. Shapefiles used to reconstruct the geographical extent of the Eastern Arc Mountains were sourced from Platts et al. (2011), while natural forest boundaries within the Taita Hills are based on Pellikka et al. (2009).

Weather and forest degradation jointly influence nestling condition. Maximum likelihood model predictions (lines and 95% confidence intervals (grey shading)) and partial residuals (dots) for the relationships between forest quality (forest patch size, canopy height or canopy cover) and (a) wing length, (b) tarsus length and (c) Scaled Mass Index (SMI) across temperature (temp) and precipitation (prec) quartiles (Q1, Q2 and Q3). (a) Wing length increased with increasing forest patch sizes under low temperatures (Q1 = 27°C). Conversely, wing length varied less with forest patch size under medium (Q2 = 28°C) and high temperatures (Q3 = 30°C), with nestlings in small forest patches exhibiting longer wings at high temperatures than those experiencing low or medium temperatures in the same patches. Nestlings born in large forest patches consistently exhibited long wings under all temperature conditions. (b) Tarsus length increased with canopy height at low temperatures (Q1 = 27°C) but declined as canopy height increased at high temperatures (Q3 = 30°C). In low‐canopy areas, nestlings had longer tarsi under medium (Q2 = 28°C) and high temperatures (Q3 = 30°C) than under low temperatures. (c) Nestling SMI was positively related to canopy cover under all precipitation scenarios. However, in forest areas with low canopy cover nestlings exhibited low SMI when average precipitation was below 1 mm (Q = 1 mm), while under high precipitation (Q3 = 4 mm), nestlings in these habitats were able to maintain higher SMI values. In forests areas with high canopy cover, nestlings were able to maintain high SMI values under all precipitation levels.

Nestlings in larger forest patches were heavier and had higher corticosterone levels. Maximum likelihood model predictions (lines and 95% confidence intervals (grey shading)) and partial residuals (dots) for the relationship between forest patch sizes and (a) nestling weight or (b) corticosterone levels. Forest patch size showed a positive association with both nestling weight and corticosterone levels. Corticosterone values in (b) are shown on a back‐transformed scale for visualization, although the models were fitted using a log‐transformed response variable (see methods).

At high temperatures parents fed less and nestlings were in worse condition. Maximum likelihood model selection (lines and 95% confidence intervals (grey shading)) and partial residuals (dots) for the relationship between daily maximum temperature averaged over a month before sampling and (a) Scaled Mass Index (SMI) or (b) SMI/wing length ratio and (c) daily maximum temperature and food provisioning rates. Temperature showed a negative association with body condition metrics: Hotter conditions were linked to reduced SMI (a), while wing length increased relative to SMI (b), likely indicating a trade‐off between somatic tissue gain and structural (wing) growth. Provisioning rates similarly declined with increasing temperatures (c). Provisioning rates in (c) are shown on a back‐transformed scale for visualization, although the models were fitted using a log‐transformed response variable (see methods).

Summary of the main results. Predictions concerned the effects of forest quality and weather (temperature and precipitation) on nestling condition, corticosterone levels (Cort), and parental provisioning behaviour, and morphological plasticity in response to thermal stress in the placid greenbul. Environmental drivers are depicted as pictograms; baseline conditions are near‐natural forest, average rainfall, and temperature. Decreased forest quality and rainfall are depicted by black downward arrows, increased temperatures by black upward arrows, and interactions by two adjacent pictograms. Additional columns show morphological (wing length, tarsus length, weight, scaled mass index (SMI)), SMI:Wing (SMI‐to‐Wing length ratio), physiological, and behavioural responses of placid greenbuls to these environmental drivers, with red downward arrows indicating negative relationships and blue upward arrows indicating positive relationships. Wing length and tarsus length were generally shorter in degraded forest but reached similar length as in near‐natural forest during high temperatures.

Tables

TABLE 1. Parsimonious models (models within ΔAICc < 2) derived from AIC model ranking of linear mixed models for variation in nestling condition and parental provisioning rate. Estimates for control predictors included in the models as well as all models considered in the model selection process are provided in Tables S4–S10.

Model rank	Intercept	Forest patch size	Canopy height	Canopy cover (%)	Temperature	Precipitation	Forest patch size * Temperature	Canopy height * Temperature	Canopy cover (%) * Temperature	Canopy cover (%) * Precipitation	K	ΔAICc	weight
(a) Wing length
1	40.03 (39.33; 40.72)	0.30 (−0.09;0.70)			0.13 (−0.24; 0.49)		−0.51 (−0.93; −0.08)				12	0.00	0.33
2	39.99 (39.27; 40.69)	0.48 (0.1; 0.88)									10	1.29	0.17
(b) Tarsus length
1	23.05 (22.78; 23.29)		0.02 (−0.11; 0.14)		0.03 (−0.10; 0.15)			−0.18 (−0.28; −0.08)			13	0.00	0.53
(c) Weight
1	18.03 (17.63; 18.43)	0.49 (0.26; 0.72)				0.16 (−0.03; 0.36)					12	0.00	0.19
2	18.03 (17.62; 18.43)	0.49 (0.26; 0.72)									11	0.42	0.15
(d) Corticosterone levels
1	1.49 (1.31;1.67)	0.13 (0.00; 026)									10	0.00	0.22
2	1.47 (1.20; 1.66)										9	1.45	0.10
(e) SMI
1	18.88 (18.35; 19.21)			0.31 (0.14; 0.48)	−0.18 (−0.35; −0.01)	0.17 (0.01; 0.33)			0.10 (−0.02; 0.23)	−0.16 (−0.31; −0.00)	15	0.00	0.31
(f) SMI‐to‐wing length ratio
1	0.48 (0.47; 0.49)				−0.01 (−0.01; −0.00)						10	0.00	0.11
(g) Parental provisioning
1	1.24 (1.10; 1.37)				−0.15 (−0.20; −0.08)						9	0.00	0.84

Full text

Introduction

1

Habitat loss and degradation, together with anthropogenic climate warming, are widely recognised as major drivers of biodiversity loss in terrestrial ecosystems (Haddad et al. 2015; Opdam and Wascher 2004; Verboom et al. 2010). Forested habitats, which support the majority of the world's terrestrial species, are particularly vulnerable to these interacting threats (Haddad et al. 2015; Newbold et al. 2015). Reduced forest area and associated edge effects result in smaller and more isolated forest patches (Suorsa et al. 2003), while degradation alters vegetation structure and the availability of food and other critical resources (Boulton et al. 2008; Vergara et al. 2021; Zanette et al. 2000). These changes alone pose significant challenges to forest‐dependent species (Gibson et al. 2013; Miranda et al. 2021). However, when combined with rising temperatures and more frequent hot and dry conditions, negative impacts may be exacerbated as smaller and lower quality forests provide reduced microclimatic buffering (Conenna et al. 2017; Hardwick et al. 2015; Laurance 2004). As a result, species restricted to poor quality and smaller forests face both reduced resource availability (Zanette et al. 2000) and exposure to harsher environmental conditions that threaten their survival (Conenna et al. 2017). While research on the combined effects of habitat and climate change on species demography is gaining traction (Neate‐Clegg et al. 2023; Stouffer et al. 2021; Wolfe et al. 2025), the mechanisms which drive the demographic processes underlying population dynamics, such as reproductive performance and offspring condition, remain underexplored, especially in tropical forest systems that may be particularly vulnerable to such changes.

Although habitat and climate change may affect many taxa (Mantyka‐Pringle et al. 2012; Penjor et al. 2021), our study focuses on birds, using a tropical understorey passerine as a model system. In birds, these combined stressors can have particularly severe consequences during reproduction, as offspring are vulnerable to environmental conditions, which can impact their survival and development (Miranda et al. 2021; Rodríguez and Bustamante 2003). Forest loss and degradation can reduce the abundance and quality of food resources (Boulton et al. 2008; Vergara et al. 2021; Zanette et al. 2000), leading to reduced parental provisioning (Miranda et al. 2021; Zanette et al. 2000). When habitat quality is low, parents may be forced to travel further in search of prey, limiting the quantity or frequency of food deliveries to the offspring (Hakkarainen et al. 1997; Staggenborg et al. 2017). Extreme weather exacerbates these effects: for example, hotter, drier conditions can reduce arthropod availability for insectivorous animals and increase the energetic demands of both parents and offspring (Aggarwal et al. 2023; Alexander 2005; Barras et al. 2021). Such constraints can ultimately impair offspring growth and development, reflected in reduced body condition or elevated corticosterone levels, a glucocorticoid hormone widely used as an indicator of physiological stress and nutritional state in birds (Leshyk et al. 2012; Messina et al. 2021, 2020; Miranda et al. 2021; Suorsa et al. 2003), thereby increasing offspring mortality rates and putting populations at risk.

Furthermore, in edge‐dominated forest patches where vegetation structure is degraded and microclimate buffering is weakened, offspring may be exposed to periods of high temperature and low humidity during critical developmental stages, potentially influencing their development and physiology. Nestlings are particularly sensitive to thermal stress due to their lack of insulating plumage (Bourne et al. 2020). Indeed, experimental studies show that nestlings reared in hot conditions often exhibit reduced body size and mass (Andrew et al. 2017; Rodriguez and Barba 2016), which can persist into adulthood (Teplitsky et al. 2008) and reflect ecogeographical rules such as Bergmann's rule (larger body sizes in cooler climates; Bergmann 1848) and Allen's rule (longer appendages in warmer climates; Allen 1877). Therefore, understanding how microclimatic effects due to forest loss and degradation affect reproductive success is critical for predicting species responses to future climate scenarios.

Tropical forest understory insectivores are of particular concern because they often inhabit stable microclimates, require high structural complexity for foraging, and are sensitive to changes in arthropod communities (Aggarwal et al. 2023; Newmark and Stanley 2016; Powell et al. 2015; Şekercioğlu 2002; Sekercioglu et al. 2002). Although the population‐level effects of habitat change on these birds are relatively well documented (Şekercioğlu 2002; Stouffer et al. 2021; Vergara et al. 2021), sublethal effects that could contribute to long‐term declines, such as reduced body condition or physiological stress, have only recently received attention. Furthermore, most studies linking microclimate change to bird condition have been conducted in temperate regions (e.g., Conenna et al. 2017), where many species are short‐lived (Martin 1996, 2015) and have historically experienced a wider range of climatic conditions (Huey et al. 2012; Shah et al. 2017). In contrast, long‐lived tropical understorey birds adapted to relatively stable environments may be less physiologically flexible to sudden climatic changes (Huey et al. 2012; Martin 2015; Shah et al. 2017; Wiersma et al. 2007), highlighting the need for research in tropical systems.

Here, we uniquely combine morphological, physiological and behavioral observations to examine the effects of habitat loss, degradation and climate warming on early‐life development of tropical forest birds, an area that remains largely unexplored. We integrate a long‐term (> 10 breeding seasons) dataset from the cloud forests of the Taita Hills biodiversity hotspot in Kenya. We focus on the placid greenbul (* Phyllastrephus cabanisi placidus*), a cooperative breeding understorey insectivore, to examine how forest quality (forest patch size and canopy structure based on terrestrial laser scanning LiDAR metrics), weather (temperature and precipitation), and their interaction shape parental provisioning and nestling condition. Previous work in this system shows that placid greenbuls avoid gaps in the forest canopy and favour areas where vegetation structure is intact (Kung'u et al. 2025), probably because arthropod prey is more abundant in these microhabitats (Kung'u et al. 2023). In addition, intact vegetation structure in these forests has been shown to buffer understory temperatures (Abera et al. 2023, 2024; Terschanski et al. 2024), potentially mitigating thermal stress and its downstream effects on offspring. We predict that (i) nestlings reared in small forest patches and degraded areas will receive less food, be leaner, and have higher corticosterone levels, (ii) hotter and drier conditions will exacerbate these effects, but less so in large forest patches or high quality areas with better microclimate buffering, and (iii) offspring may exhibit morphological plasticity to thermal stress (e.g., reduced body mass, elongated wings and tarsus), particularly in degraded, hotter forests. By examining how habitat quality and weather together influence the condition of offspring, our study highlights the importance of maintaining and restoring the structural integrity of forests to support viable populations of tropical understorey birds under ongoing climate change.

Materials and Methods

2

Study Area

2.1

We conducted our study in the fragmented cloud forest of the Taita Hills (Figure 1; south‐eastern Kenya; 30°25′ S, 38°20′ E). The Taita Hills represent the northern‐most section of the Eastern Arc Mountains, a chain of mountain blocks with a common geological past that extends from south‐eastern Kenya to southern Tanzania (Figure 1; Lovett and Wasser 2008). Despite the severe loss and fragmentation of the cloud forests covering the Eastern Arc mountain ranges, they still host many endemic species of flora and fauna, making them a global biodiversity hotspot (Bennun and Njoroge 1999; BirdLife International 2025; Burgess et al. 2007). Large‐scale deforestation within the Taita Hills ecosystem between 1955 and 2004, driven primarily by small‐scale agriculture and exotic tree plantations, resulted in the loss of half of the native cloud forest (Pellikka et al. 2009). As a result, the ecosystem now consists of a mosaic of natural cloud forest patches of various sizes (from < 1 ha to ~250 ha), separated by human settlements, small‐scale agriculture, and exotic plantations. The Taita landscape comprises three isolated mountain blocks: Mount Sagalla, Mbololo and the Dabida Massif (Figure 1). The Dabida massif (1400–2200 m above sea level) hosts three large natural forest patches (70–120 ha) along with about ten smaller patches (< 1–20 ha; Aerts et al. 2011; Pellikka et al. 2009). Thus, while there is pronounced variation in forest patch sizes, large continuous forest is absent in the Taita Hills. Even the largest remaining patches (ca. 250 ha) are not comparable to truly extensive forest blocks and are likely still influenced by edge and area‐related processes. In addition to deforestation and fragmentation, past selective logging of large trees for commercial or domestic purposes and the continued extraction of pole‐sized trees and firewood collection have altered the vegetation structure thus degrading the quality of the remaining natural forests (Aerts et al. 2011; Kung'u et al. 2023; Thijs 2015). Although degradation is especially prevalent in small‐sized forest patches, deterioration in forest quality is also evident in larger patches that lack the protection status of national forest reserves (Kung'u et al. 2023; Wekesa et al. 2020).

Location of the Taita Hills within the Eastern Arc Mountain chain. Map (a) displays the main mountain blocks of the Eastern Arc, including their position within Africa (inset) and highlights the location of the Taita Hills among these mountain chains. Map (b) zooms in on the Taita Hills, emphasizing the isolation of the main Taita Hills blocks (Mbololo and Dabida Massif) from Sagala Hills, also part of the Eastern Arc Mountains. Map (c) shows the forest fragments within the Taita Hills. Map lines delineate study areas and do not necessarily depict accepted national boundaries. Shapefiles used to reconstruct the geographical extent of the Eastern Arc Mountains were sourced from Platts et al. (2011), while natural forest boundaries within the Taita Hills are based on Pellikka et al. (2009).

Study Species

2.2

The placid greenbul is a medium‐sized (mean weight: 30.6 g) insectivorous passerine that inhabits cloud forests in East Africa (Bennun and Njoroge 1999). Sexes look alike, but males are larger than females. In the Taita Hills, the placid greenbul has been recorded as a facultative cooperative breeder, with approximately 70% breeding in groups (containing a dominant breeding pair and one to five subordinates) and the rest breeding in pairs (Cousseau 2020). Their breeding season extends from the onset of the short rains in East Africa (ca. September/October) until April. Dominant breeding pairs are territorial, as they have been recorded re‐nesting in the same area in consecutive years (Cousseau 2020). Territory size depends on habitat quality: groups occupy larger home ranges in areas with reduced canopy cover, while territories are smaller in areas with high canopy cover (Kung'u et al. 2025). Females lay and incubate two to three eggs. When breeding within a cooperative group, a subset of subordinates of both sexes assists in feeding the nestlings (hereafter referred to as helpers; Cousseau et al. 2022; Van de Loock et al. 2023), allowing the females to reduce their investment in feeding when helpers are present (load lightening theory; Van de Loock et al. 2023). Cooperative breeding in the placid greenbul has been shown to increase fledging success (Van de Loock et al. 2023; Van de Loock et al. 2017), and nestlings born in groups with more helpers have been recorded to fledge with longer wings (Van de Loock et al. 2023).

Sampling Nestlings

2.3

Nest Searching

2.3.1

Placid greenbul nests were systematically searched and geolocated using Garmin handheld GPS receivers (accuracy ≤ 6 m) by four experienced local field assistants during the breeding season between 2007 and 2022 in eight forest fragments within the Dabida Massif of the Taita Hills. To ensure systematic coverage, the forest patches were distributed among the four field assistants, and every 2 weeks, assistants were rotated to different patches. Each day, field assistants focused on a specific section of the forest patch (for larger patches) or an entire small forest patch and searched thoroughly by walking throughout the section. Searches were concentrated in the understory below 4 m, where open cup nests are typically built in forks of understory shrubs, climbers, and small trees at an average height of 1.3 m (Fishpool and Tobias 2020; Van de Loock et al. 2019). In addition, field assistants used behavioral cues, such as observation of adults carrying nest materials or food, and also drew on prior experience and knowledge of local flocks (see field methods also in Spanhove et al. 2014).

Nest Monitoring

2.3.2

Once located, each nest was monitored every 4 days until the nestlings fledged or the nesting attempt failed. If a nest containing hatched eggs was found, the age of the nestlings was determined using a pictorial guide to their daily development. Between the ages of 6 and 12 days (average 9 days) and while still in the nest, nestlings were tagged with a metal ring, individually colour‐banded, and blood samples were taken. Body morphometric measurements, including weight, wing length, and tarsus length, were also taken. The number of nestlings was determined by the number of individuals ringed per nest.

Corticosterone Sampling

2.3.3

During four breeding seasons (2017, 2018, 2019 and 2020), we also measured corticosterone levels when nestlings were on average 9 days old (range 7–12 days). Nestlings were taken out of the nest individually, that is, in the case of two nestlings one remained in the nest until being handled, and the order of removal was noted down. In most cases nestlings were bled within 180 s of removal from the nest (mean ± sd: 104 ± 40 s, range 37–259 s, six samples above 180 s were not included in the data analysis; Romero et al. 2005). No playback was used before or during nestling handling. Blood samples were taken through venipuncture from the wing vein, collected into heparinized capillaries, and stored on ice until return to the field station. Blood samples were centrifuged the same day, the amount of plasma measured with a Hamilton syringe and stored in 500 μL pure ethanol (Goymann et al. 2007). Blood cells were stored separately in ethanol. The sex of each nestling was determined molecularly using a set of sex‐linked primers P2/P8 (Griffiths et al. 1998).

Corticosterone Assay

2.4

Plasma corticosterone concentrations were quantified using a commercially available enzyme immunoassay (EIA; ENZO Life Science Corticosterone EIA Kit; Cat. No. ADI‐901‐097). Following the protocol described in Apfelbeck et al. 2024, (see also, Goymann et al. 2007), samples were centrifuged for 10 min at 3900 rpm, the supernatant was pipetted into new extraction tubes and dried under a N2 stream at 37°C. Then 4 mL dichloromethane and 500 μL ddH2 were added, vortexed and stored overnight in a refrigerator. The following day, the samples were placed on a shaker for 30 min, centrifuged at 3900 rpm for 10 min, and then freeze‐decanted twice to separate the aqueous and organic phases. Both organic phases of each sample were collected in a new glass tube, dried under a stream of N2 at 37°C and resuspended in an assay buffer provided by the manufacturer. Samples were corrected for dilution. The distribution of samples between plates was balanced for the breeding seasons. The sensitivity of the assay was 27 pg/mL. Spiked chicken plasma (CP; extracted and non‐extracted) was used as a control to calculate the coefficient of variation (CV). The intra‐assay CV of non‐extracted CP replicates was 7% ± 0.014 (mean ± SDev) and of extracted CP was 3% ± 0.012 (mean ± SDev). Inter‐assay CV of non‐extracted CP was 11% and of extracted CP was 8%. The mean % CV of duplicates was ≤ 10%.

Determining Group Sizes and Composition

2.5

Group size and composition (i.e., the breeding role of each group member: dominant breeding male, dominant breeding female, or subordinate) for each nest was determined during nest visits through focal observations, targeted mist‐netting, and video recordings. Focal observations were made during the incubation period (2007–2015) or immediately after the nestlings were ringed (2012–2022). Typically, we positioned ourselves 10 m away from the nest and lured group members to the nest by intermittently playing a distress call for about 10 min. Adults were captured with mist nets when nestlings were ca. 5 days old (range 3–8 days old). An individual was identified as a dominant breeding female if a brood patch was present, or as a dominant breeding male if cloacal swelling was observed. Other individuals caught alongside the dominant breeding pair during the mist‐netting sessions were identified as subordinates. Not all subordinates help to feed the nestlings, so we identified helpers from video recordings at the nest. Filming was done when nestlings were on average 9 days old (range 6–12 days, see below). At nest sites where estimates of group size or number of helpers were ambiguous due to the presence of unringed individuals, or where colour band combinations could not be accurately determined, we considered such estimates unsatisfactory and excluded them from the analysis when controlling for the influence of cooperative breeding on nestling condition in our models.

Food Provisioning

2.6

Continuous video recordings of nestling feeding were made for 4–6 h (between ca. 0700 and 1400 h) using a camouflaged and waterproofed high‐definition camera (Sony Corp), mounted on a tripod and positioned approximately 1.5 m away from the nest. We extracted all feeding events and, where possible, identified the feeding individual based on the colour ring combination. To obtain the feeding rate for each nest, we extracted all feeding events from all individuals and divided them by the video length. We discarded videos that lasted < 4 h or > 6 h to minimize bias in feeding rates due to variation in feeding behaviour at different times of the day. We also discarded videos in which there was a fledging or predation event.

Forest Quality

2.7

We characterized forest quality based on canopy cover structure using airborne laser Light Detection And Ranging (hereafter LiDAR) data. LiDAR data were collected using a Leica ALS60 sensor mounted on an aircraft during three flight campaigns in January–February 2014, February 2015, and February–March 2022. During the first two flight campaigns, the aircraft flew at an average altitude of approximately 1450 m above ground, and the collected data had an average return density of 3.4 points per square meter. LiDAR data collected in February–March 2022 were obtained at an average altitude of 800 m, resulting in an average return density of 4.5 points per m^2^. LiDAR returns were classified into ground and non‐ground returns, followed by the calculation of a 1‐m resolution digital terrain model (DTM) using LAStools software (rapidlasso GmbH). Based on the DTM, we normalized the return elevations to heights from the ground level and interpolated the maximum heights to create a 1‐m resolution canopy height model (CHM).

Three canopy structure proxies (i.e., mean canopy height (m), canopy height variation, and percent canopy cover) were calculated from normalised point clouds for a 50‐m radius (approximately 0.79 ha) area around each sampled nest location using the lidR R package (Roussel et al. 2020). We estimated the mean canopy height based on the 25th percentile of return heights. The variation in canopy height was assessed using the coefficient of variation of the return heights. All returns below 3 m were considered to be ground or understorey returns and were excluded. Canopy cover was calculated as the ratio of first returns from the canopy to all first returns at a threshold of 3 m (Heiskanen et al. 2015). Nest sites sampled between breeding seasons 2007 and 2017 were matched to LiDAR data scanned during the first two field campaigns (January–February 2014 and February 2015), while the remainder (sampled during breeding seasons 2018–2022) were matched to the most recent LiDAR data (i.e., field campaigns conducted in February–March 2022).

We also considered patch size as a proxy for forest quality. The size of each forest patch was determined by calculating the area within natural forest boundary maps generated from airborne imagery (Pellikka et al. 2009).

Weather Data

2.8

Weather data in the form of daily precipitation and temperature for the Taita Hills ecosystem were obtained from the Kenya Meteorological Department. This dataset is a composite of ground‐based observations (local weather stations) and meteorological satellite data, harmonized into a gridded format with a spatial resolution of 4 km by 4 km. From these data, we also calculated mean precipitation and temperature for 1 month prior to the date when each nestling's biometrics were collected as an indication of the abiotic environmental conditions during the nesting and nestling development period. We chose to average over a month preceding nestling sampling to capture environmental effects on arthropod availability during both pre‐ and post‐hatching phases, as well as thermoregulatory challenges experienced during incubation and nestling development, which last approximately 17 and 12–14 days respectively. Our approach follows Wiley and Ridley (2016), who used a one‐month weather window to represent the lag between rainfall and insect emergence in the Kalahari.

Across the sampling years of this study (2007–2022) the mean maximum and minimum daily temperature across the entire study area during the placid greenbuls' breeding season were 29°C (range 19°C—37°C) and 16°C (range −1°C—23°C), respectively. Daily precipitation over the same period averaged 4 mm (range 0 mm–108 mm). In the month preceding sampling of each nestling, mean maximum temperature and minimum temperature were 28°C (range 24°C—34°C) and 16°C (range 9°C—21°C), respectively, while mean precipitation of the same period averaged 3 mm (range 0 mm–14 mm).

Derived Body Condition Indices

2.9

In addition to direct morphometric measures (weight, wing length, and tarsus length), we derived two additional body condition indices. The Scaled Mass Index (SMI) was calculated from nestling weight and tarsus length measurements as: SMI = weight × (average tarsus length/tarsus length)*bSMA, where bSMA is the scaling exponent (slope) obtained from a standardized major axis (SMA) regression of log body weight on log tarsus length, following Peig and Green (2009). SMI, where the weight of individuals has been adjusted by their body length, is a reliable measure of the gross nutrient reserve of animals (Labocha et al. 2014; Peig and Green 2009). As suggested by Peig and Green (2009), we excluded 24 out of 750 individuals that were apparent outliers based on log body weight on log tarsus length (standardized residual ≥ 3 or ≤ −3; Figure S1; Peig and Green 2009) using the Smart package in R (Warton et al. 2006). The average tarsus length used to calculate SMI was 23.61 mm and a scaling exponent of 1.78 was obtained (n = 726).

We also calculated the ratio of SMI (numerator) to wing length (denominator) to examine how nestlings allocate energetic resources between size‐adjusted body condition (somatic reserves) and flight‐related structural growth, two key developmental components that can be influenced by habitat and weather conditions (Ashton and Armstrong 2002; Wright et al. 2006). Higher values suggest relatively greater investment in somatic reserves (herein SMI), whereas lower values indicate relatively greater investment in wing development. Thus, while SMI informs about the nutritional status of individuals, the ratio between SMI and wing length indicates whether individuals adjust investment into different body parts depending on environmental conditions.

Sample Sizes: Morphometrics, Physiology and Food Provisioning Rates

2.10

Data on nestling morphometrics (SMI, weight, and tarsus length) were collected over 13 breeding seasons (2007–2022), while food provisioning recordings were conducted across 11 breeding seasons within the same period (i.e., 2008–2010 and 2013–2022). Wing length and SMI‐wing length ratio were measured during ten of these seasons (2012–2022) and corticosterone levels were measured during four seasons (2017–2020). Four nestlings were excluded from the SMI, weight, and SMI‐to‐wing length ratio analyses because their recorded weights were disproportionately high relative to their tarsus and wing lengths, likely indicating erroneous measurements. However, these individuals were retained in the tarsus and wing length analyses. Some nests had missing predictor variables (e.g., LiDAR data, sex, nestling age or breeding female), which further reduced sample sizes. Final sample sizes were wing length (n = 470), weight (n = 534), tarsus length (n = 537), SMI (n = 531), SMI/wing length ratio (n = 463), corticosterone levels (n = 198), and food provisioning rates (n = 171).

Statistical Modelling

2.11

Effects of Forest Quality and Weather on Nestling Condition

2.11.1

All statistical analyses were performed in the R environment, version 4.1.2 (R Core Team 2022). We used linear mixed models (LMM) to investigate whether variation in nestling body condition (modelled separately for weight, tarsus length, wing length, SMI, SMI‐to‐wing length ratio, and corticosterone levels) was explained by forest quality, weather, or their interaction. Weather predictors included the average maximum temperature and average precipitation in the month before the sampling date (referred to as temperature and precipitation in the text) in all the models. In addition, the following predictors known to influence nestling condition were included in all models: (1) the number of nestlings in each nest, which may influence how much food nestlings receive and thus their growth rate, (2) sex, as adult male placid greenbuls are generally larger than females, (3) nestling age, as nestlings were sampled at different ages (6–12 days). For nests where nestling age could not be determined with certainty in the field, particularly in cases of multiple nestlings (2–3), we assigned the younger age within the estimated range to the visibly smaller nestling or adjusted its age downwards if a fixed number had been given. Conversely, we assigned the older age range to the larger nestling or maintained the given fixed age. This assignment was made when the difference in their tarsus length was > 0.5 mm and wing length was > 1 mm (Eck et al. 2011). This approach was based on field observations showing that eggs of placid greenbuls, like those of many bird species, do not hatch synchronously (Węgrzyn et al. 2023), making it likely that siblings within the same nest are of different ages. For nests with a single nestling and where an estimated age range was provided, we assigned the highest age in the range.

For a subset of nests (see Table S1), we had data on cooperative group size and the number of individuals feeding nestlings (both of which have been shown to influence nestling condition and breeding success in our study area; Van de Loock et al. 2023; Van de Loock et al. 2017), which allowed us to run an additional set of models. As group size and number of feeders were strongly correlated (r = 0.63), we built separate models for each covariate where such information was available. However, neither group sizes nor number of feeding individuals was significantly associated with any of the nestling condition indexes (Tables S2 and S3). Therefore, to reduce model overparameterization when evaluating variation in nestling body condition indexes, cooperative breeding covariates were excluded from the final models, which were fitted using a larger sample size than the subset for which such data were available.

When analysing the relationship between forest quality, weather, and corticosterone levels, we also included sampling time (i.e., time between nestling removal and blood sampling), the order in which nestlings were sampled (first or second, Honarmand et al. 2010), and wing length. Nestlings with a sampling time longer than 180 s, three extreme outliers (corticosterone > 40 ng/mL) and five nests with three nestlings were excluded from the analysis.

In all models, we included breeding female ID as a random factor, as females mothered multiple nestlings in different breeding seasons within the same territory. We also included nest ID, nested within the breeding female ID, to account for non‐independence of nestlings from the same nests and forest patch ID to account for non‐independence of nestlings sampled within the same forest patch.

Effects of Forest Quality and Weather on Food Provisioning Rates

2.11.2

We used linear mixed models (LMMs) to investigate the effects of forest quality, weather, and their interaction (i.e., forest quality*weather) on feeding frequency (i.e., number of feeds per hour). Rather than using the average maximum temperature before the sampling date as one of the weather predictors, as was done in the nestling condition models, we used daily maximum temperature since daily temperature has been shown to influence food provisioning rates in other studies (Bourne et al. 2021; Low et al. 2008; Wiley and Ridley 2016). We did not include maximum temperature averaged over a month prior to sampling date and maximum temperature on the day of the food provisioning video recordings in the same models, as they were highly correlated (r = 0.70). We included nestling age, the number of nestlings, and the number of feeding individuals as predictor variables in the models as these have been shown to influence feeding frequency in our study species (Van de Loock et al. 2023) and in others (Wiley and Ridley 2016). Two random factors were included in the models: (1) forest patch, to account for the non‐independence of feeding behaviour of individuals within the same patch, and (2) breeding female, as the same groups nest in the same territory in different breeding seasons.

General Modelling Procedure

2.11.3

Prior to analysis, we checked for collinearity between the explanatory variables. Given that different LiDAR‐based canopy structure metrics served as measures of within‐forest quality, while forest patch size served as an indicator of landscape‐level forest quality, these predictors were evaluated in separate model sets. Continuous predictor variables were scaled and centred to a mean of zero and a standard deviation of one to facilitate interpretation of the model coefficients. Model diagnostic plots were visually inspected to ensure that the assumptions of normality and homoscedasticity of the residuals were not violated. Two response variables (i.e., food provisioning rate, and corticosterone levels) were log‐transformed to meet model assumptions.

We constructed an a priori candidate model set to evaluate the independent and interactive effects of forest quality and weather using the Akaike's information criterion corrected for small sample size (AICc, Burnham and Anderson 2002). The following models were compared and ranked: (1) a null model (response variable ~ random effects); (2) a baseline model including the relevant control predictors for each response (i.e., nestling age, sex, and number of nestlings in each sampled nest for nestling condition models, and nestling age, number of nestlings and number of feeding individuals for food provisioning models); (3) four models including each of the forest quality proxies; (4) three models based on weather predictors only (precipitation and temperature separately or combined); (5) 12 models including each of the forest quality proxies and one or both of the weather predictors; and (6) 12 models including the interaction of each of the forest quality proxies and weather predictors (i.e., forest quality proxies*weather predictors). Models within an AIC difference of < 2 (hereafter ΔAICc < 2) were included in the results. However, we did not interpret complex models that included an additional predictor variable and were ranked below a simpler, nested model with a lower AICc value, as this suggests that the added parameter did not improve model fit and was likely uninformative (i.e., a “pretending” variable; Arnold 2010; Leroux 2019). LMMs were fitted using the R package lme4 with the maximum likelihood option to allow comparison of models, while model comparison and selection of parsimonious models was done using the MuMln R package (Barton 2023). We calculated 95% confidence intervals (95% CI) of the slope estimate for predictors in all models to determine the extent of their ecological influence on nestling condition. We did not interpret the effect of ‘control’ predictor variables.

Results

3

Nestling Condition

3.1

When examining variation in wing length, two models within an AIC difference of < 2 were retained (Table 1; Table S4). These models included either the influence of forest quality (patch size) alone or its interaction with temperature. The most parsimonious model, which incorporated the interaction, had a model weight of 0.33, almost twice that of the second‐best model (weight = 0.17), indicating stronger support for the model including this interaction. This model suggested that wing length increased with forest patch size under low temperatures (Figure 2a). This positive relationship, however, weakened under medium and high temperatures, in that nestlings born in small patches exhibited longer wings than when in the same habitats under low temperature regimes (Figure 2a). In large forest patches, nestlings consistently exhibited long wings across all temperature conditions (Figure 2a). The second‐best model suggested that wing length increased with increasing forest patch sizes (Table 1).

**TABLE 1: Parsimonious models (models within ΔAICc < 2) derived from AIC model ranking of linear mixed models for variation in nestling condition and parental provisioning rate. Estimates for control predictors included in the models as well as all models considered in the model selection process are provided in

Weather and forest degradation jointly influence nestling condition. Maximum likelihood model predictions (lines and 95% confidence intervals (grey shading)) and partial residuals (dots) for the relationships between forest quality (forest patch size, canopy height or canopy cover) and (a) wing length, (b) tarsus length and (c) Scaled Mass Index (SMI) across temperature (temp) and precipitation (prec) quartiles (Q1, Q2 and Q3). (a) Wing length increased with increasing forest patch sizes under low temperatures (Q1 = 27°C). Conversely, wing length varied less with forest patch size under medium (Q2 = 28°C) and high temperatures (Q3 = 30°C), with nestlings in small forest patches exhibiting longer wings at high temperatures than those experiencing low or medium temperatures in the same patches. Nestlings born in large forest patches consistently exhibited long wings under all temperature conditions. (b) Tarsus length increased with canopy height at low temperatures (Q1 = 27°C) but declined as canopy height increased at high temperatures (Q3 = 30°C). In low‐canopy areas, nestlings had longer tarsi under medium (Q2 = 28°C) and high temperatures (Q3 = 30°C) than under low temperatures. (c) Nestling SMI was positively related to canopy cover under all precipitation scenarios. However, in forest areas with low canopy cover nestlings exhibited low SMI when average precipitation was below 1 mm (Q = 1 mm), while under high precipitation (Q3 = 4 mm), nestlings in these habitats were able to maintain higher SMI values. In forests areas with high canopy cover, nestlings were able to maintain high SMI values under all precipitation levels.

When examining variation in tarsus length, only one parsimonious model was retained (Table 1; Table S5). This model included the interaction between forest quality (canopy height) and temperature, and carried a model weight of 0.53, indicating that it accounted for over half of the support among the candidate models. This model suggested that tarsus length was positively associated with canopy height at low temperatures, but showed the opposite pattern under high temperatures, where tarsus length decreased with increasing canopy height (Figure 2b). In low‐canopy height areas, nestlings exhibited longer tarsi under high temperatures compared to nestlings in the same habitat under medium or low temperatures (Figure 2b).

When examining variation in weight, six models within an AIC difference of < 2 were retained (Table S6). Only two of these six models were parsimonious (weight = 0.19 and 0.15 respectively) and included forest patch size and precipitation or forest patch size only (Tables 1 and S6). Both models suggested that nestling weight increased with increasing forest patch size (Table 1, Figure 3a). The positive association of precipitation and weight was weak, as the 95% CI of its slope estimate included zero (Table 1). The remaining four models were complex versions of the second‐best model, as they retained a forest quality proxy as well as two or more additional predictors (Table S6). The inclusion of precipitation in the most parsimonious model suggests a modest additional contribution to model fit, while the consistent retention of patch size across all models highlights its strong influence on nestling weight.

Nestlings in larger forest patches were heavier and had higher corticosterone levels. Maximum likelihood model predictions (lines and 95% confidence intervals (grey shading)) and partial residuals (dots) for the relationship between forest patch sizes and (a) nestling weight or (b) corticosterone levels. Forest patch size showed a positive association with both nestling weight and corticosterone levels. Corticosterone values in (b) are shown on a back‐transformed scale for visualization, although the models were fitted using a log‐transformed response variable (see methods).

When examining variation in nestling corticosterone three models were retained within an AIC difference of < 2 (Tables 1 and S7). The most parsimonious model (model weight = 0.22) only included forest patch size as the forest quality proxy. This model suggested that patch size was positively related to nestling corticosterone levels (Table 1; Figure 3b). The second best model (model weight = 0.10) included only the control predictor variables, i.e., both the forest quality and weather parameters were not retained in this model (Tables 1 and S7). The third model (model weight = 0.08) was a complex version of the second‐best model, as it retained forest patch size and temperature as predictors (Table S7).

For SMI one model within an AIC difference of < 2 was retained (Tables 1 and S8). This model accounted for a model weight of 0.31 and included forest quality (canopy cover (%)) and weather (temperature and precipitation) as well as their interactions. The interaction of canopy cover and temperature on SMI was not well supported because the 95% CI of the slope included zero (Table 1). However, temperature was negatively related to SMI (Table 1; Figure 4a). Moreover, the 95% CI of the slope for the interaction of canopy cover and precipitation did not cross zero, indicating strong support for this interaction (Table 1). This interaction suggested that canopy cover had an overall positive association with SMI across all precipitation regimes. However, nestlings born in low canopy cover areas exhibited lower SMI values when average precipitation was below 1 mm, whereas under high precipitation (> 4 mm), these same habitats produced nestlings with higher SMI values (Figure 2c). In contrast, nestlings born in high‐canopy cover areas maintained consistently high SMI across all precipitation levels, with SMI values always exceeding those of nestlings from low‐canopy cover areas (Figure 2c).

At high temperatures parents fed less and nestlings were in worse condition. Maximum likelihood model selection (lines and 95% confidence intervals (grey shading)) and partial residuals (dots) for the relationship between daily maximum temperature averaged over a month before sampling and (a) Scaled Mass Index (SMI) or (b) SMI/wing length ratio and (c) daily maximum temperature and food provisioning rates. Temperature showed a negative association with body condition metrics: Hotter conditions were linked to reduced SMI (a), while wing length increased relative to SMI (b), likely indicating a trade‐off between somatic tissue gain and structural (wing) growth. Provisioning rates similarly declined with increasing temperatures (c). Provisioning rates in (c) are shown on a back‐transformed scale for visualization, although the models were fitted using a log‐transformed response variable (see methods).

When examining variation in SMI‐to‐wing length ratio, ten models were retained within ΔAICc < 2. These models included forest quality, weather parameters, and their interactions (Table S9). The most parsimonious model included only temperature, showing a decline in the SMI‐to‐wing length ratio with increasing temperature (Table 1; Figure 4b). This indicates that nestlings exposed to hotter conditions allocated less into the accumulation of somatic reserves (herein calculated as SMI) relative to wing growth. The other nine models within ΔAICc < 2 were complex versions of the most parsimonious model as they included not only temperature but also other additional predictors (Table S9). Although the most parsimonious model had the highest support (weight = 0.11), its relatively low model weight reflects substantial model uncertainty arising from the presence of several competing models within ΔAICc < 2. This indicates that no single model overwhelmingly accounted for variation in the SMI‐to‐wing length ratio. However, the consistent inclusion of temperature across these models suggests strong support for its negative association with SMI‐to‐wing length ratio.

Food Provisioning Rate

3.2

One model (model weight = 0.84) explaining variation in food provisioning rates was retained within ΔAICc < 2 (Tables 1 and S10). This model included only daily maximum temperature and suggested that carers reduced their provisioning rates to nestlings as temperature increased (Table 1; Figure 4c).

Summary of the Findings

3.3

In summary, our main predictions were largely supported (Figure 5). Nestlings in smaller and lower‐quality forest areas (i.e., areas with low canopy cover or low canopy height) had poorer condition (Prediction 1), and this was exacerbated when precipitation was low (Prediction 2). Furthermore, nestlings showed morphological plasticity in small forest patches and low‐quality forest areas at high temperatures, consistent with plastic responses to heat stress (Prediction 3). By contrast, corticosterone was elevated in larger forest patches rather than smaller ones, opposite to Prediction 1.

Summary of the main results. Predictions concerned the effects of forest quality and weather (temperature and precipitation) on nestling condition, corticosterone levels (Cort), and parental provisioning behaviour, and morphological plasticity in response to thermal stress in the placid greenbul. Environmental drivers are depicted as pictograms; baseline conditions are near‐natural forest, average rainfall, and temperature. Decreased forest quality and rainfall are depicted by black downward arrows, increased temperatures by black upward arrows, and interactions by two adjacent pictograms. Additional columns show morphological (wing length, tarsus length, weight, scaled mass index (SMI)), SMI:Wing (SMI‐to‐Wing length ratio), physiological, and behavioural responses of placid greenbuls to these environmental drivers, with red downward arrows indicating negative relationships and blue upward arrows indicating positive relationships. Wing length and tarsus length were generally shorter in degraded forest but reached similar length as in near‐natural forest during high temperatures.

Discussion

4

The Role of Intact Canopy Structures in Buffering Nestling Growth

4.1

Our study shows that forest quality (patch size and degree of degradation) and weather (temperature and precipitation) jointly influence nestling condition and parental provisioning behaviour in the placid greenbul, a tropical forest understory insectivore. This is consistent with the growing evidence that the detrimental effects of habitat loss and climate warming may be exacerbated in degraded forest patches, where reduced canopy structure can compromise microclimatic buffering and resource availability (Conenna et al. 2017; Richards and Windsor 2007; Senior et al. 2017). Consistent with our first prediction, we observed that nestlings in smaller, more degraded forest patches were generally in poorer condition (i.e., shorter wing length at low temperatures, lower weight; Figures 2a and 3a). In addition, nestlings in areas with lower canopy cover showed reduced SMI during dry weather conditions (Figure 2c). These results suggest that resource limitations associated with smaller and more degraded patches reduce nestling weight and wing growth, whereas SMI was more strongly influenced by canopy buffering and weather conditions. Previous studies in temperate forests have reported that nestling condition often declines with forest fragmentation or lower tree height (Hinsley et al. 2008; Suorsa et al. 2003), and our results show a similar pattern in a tropical system.

We also found that higher ambient temperatures had a consistently negative effect on key measures of nestling body condition (Figure 4a,b). During hot periods, nestling SMI and SMI/wing length ratio were particularly low, suggesting that high temperatures are an additional stressor during chick development. Studies of multiple bird species in both temperate and tropical regions have shown that rising temperatures can reduce insect abundance or alter foraging behaviour (Barras et al. 2021; Catry et al. 2015; Cunningham et al. 2013; Lister and Garcia 2018), ultimately hindering the ability of nestlings to acquire sufficient resources. Further, consistent with the idea that intact forest canopies moderate microclimates (Hardwick et al. 2015; Terschanski et al. 2024; Thom et al. 2020), we found that nestlings in areas with high canopy cover maintained high SMI under low precipitation, whereas those in low canopy cover areas exhibited low SMI under the same conditions. A possible mechanism that may explain these results is the ability of dense canopy cover to maintain high humidity and low thermal levels during dry and hot conditions, creating a refuge for arthropods (Richards and Windsor 2007) and consequently providing a stable food supply for understory insectivorous animals and their offspring. Most arthropods are ectothermic and become less active or retreat into cooler, moister refugia when conditions are hot and dry, making them less available to foraging parents (Mellanby 1939; Terlau et al. 2023). Mortality of certain prey groups may also increase under prolonged desiccation (Rai et al. 2018). Although placid greenbuls, similar to other bulbul species, occasionally consume fruit, especially during the dry, non‐breeding season, nestling provisioning in our study population relies almost exclusively on arthropods (Kung'u et al. in prep; Dinesen et al. 2022; Fishpool and Tobias 2020). Thus, fluctuations in insect availability due to weather variability are likely to have direct consequences for nestling condition and should be a focus of future studies to better understand the mechanistic consequences of habitat degradation and weather variation on resource availability and thus bird populations.

Rainfall Variability Differentially Affects Nestling Growth Across Habitats

4.2

However, while canopy cover had a buffering effect on nestling SMI under drought conditions, we also observed a tendency under high precipitation for nestlings in areas with low canopy cover to reach SMI values higher than those in the same habitats under lower rainfall (Figure 2c). This pattern suggests that nestlings in poorer quality habitats are particularly vulnerable during dry conditions, probably because of reduced arthropod prey, but they may benefit disproportionately from increased rainfall. It is indeed possible that short‐term pulses of rainfall in low‐canopy habitats lead to abrupt increases in arthropod abundance (Richards and Windsor 2007), effectively ‘rescuing’ nestling condition in a way that is not seen in habitats that already enjoy stable microclimatic conditions. This finding highlights the importance of understanding not only average differences in habitat quality, but also how organisms in different habitat types respond to variations in precipitation. If climate change leads to more pronounced wet‐dry cycles, nestlings in degraded areas could experience greater swings in condition, both very poor during periods of low rainfall and relatively better during periods of high rainfall.

High Temperatures Are Associated With Reduced Parental Food Provisioning Across Habitats

4.3

Although nestlings in degraded forest areas weighed less, had lower SMI and shorter wings, forest size or overall quality did not explain variation in total provisioning rates. One explanation lies in the type or quality of food provided rather than the frequency of visits (Jarrett et al. 2020; Seress et al. 2018; Sinkovics et al. 2021). Parents foraging in degraded areas may provide prey that is less nutritious or suboptimal for nestling growth and development (Jarrett et al. 2020; Sinkovics et al. 2021). Some studies have suggested that arthropod communities in small or degraded tropical forest fragments, while potentially rich in total numbers, may be composed of lower‐quality prey types for certain specialist insectivores (Gibb and Hochuli 2002; Gladalski et al. 2019). Nevertheless, we found a strong negative effect of daily temperature on overall provisioning rates, independent of forest quality (Figure 4c). Two complementary mechanisms may be at work. First, arthropods, which are ectothermic, may become less accessible to avian predators when temperatures exceed their thermal tolerance, as they retreat to cooler microhabitats, become less active, or die (Mellanby 1939; Terlau et al. 2023). Second, parents themselves may reduce foraging activity to avoid overheating (Kosheleff and Anderson 2009; Lehmann et al. 2012; Tapper et al. 2020; Wiley and Ridley 2016). While this behavioural change may reduce the energetic cost to parents, it limits provisioning visits when altricial offspring are most in need of food (Van de Ven et al. 2020; Wiley and Ridley 2016), which may further explain the observed correlation between high temperature and poorer nestling condition (as in SMI).

Elevated Corticosterone Levels Reflect Developmental Readiness, Not Nutritional Constraints

4.4

We also predicted that nestlings in smaller, more degraded forest patches would have elevated corticosterone levels due to increased physiological stress (Goutte et al. 2010; Herring et al. 2011; Honarmand et al. 2010; Kern et al. 2001). However, contrary to this expectation and to findings from some temperate forest systems (Suorsa et al. 2003), we found that corticosterone levels were higher in larger forest patches, even though nestlings there tended to be in better morphological condition (Figure 3b). One possible explanation is that corticosterone in this context reflects pre‐fledging readiness rather than nutritional stress, as levels often rise shortly before fledging to enhance locomotor activity, muscle development and alertness (Kern et al. 2001; Rivers et al. 2012; Romero et al. 2005; Sprague and Breuner 2010). As nest predation pressure in our study area is high (up to 70%; Cousseau 2020; Spanhove et al. 2014), it seems plausible that nestlings able to accelerate fledging and develop flight‐related muscles or feathers more quickly could gain a survival advantage (Martin 2015; Rivers et al. 2012). High predation rates may thus interact with developmental physiology, amplifying the selective benefits of elevated corticosterone and advanced wing growth in larger patches. In support of this interpretation, wing length was positively correlated with corticosterone levels and both wing length and body mass were significantly greater in larger forest patches. Thus, wings were not proportionally longer but increased together with overall condition, indicating that nestlings in larger patches developed both mass and wing length more fully, which may facilitate earlier fledging. The positive correlation between corticosterone levels and wing length is consistent with other studies showing that this hormone can stimulate locomotor activity, muscle development and generally reflects increased metabolic needs (Dufty et al. 2002; Jimeno et al. 2017; Romero et al. 2005). Conversely, nestlings in degraded patches may be constrained by limited or lower‐quality prey, restricting their ability to invest in both mass and wing growth and keeping corticosterone at relatively lower levels. To further elucidate these relationships food supplementation experiments in combination with morphology and corticosterone measurements are necessary in future studies. Despite this, nestlings from degraded forest areas did not seem to be at starvation point as starvation is usually marked by elevated corticosterone levels (Kitaysky et al. 2001), that is, the opposite pattern of our findings. Thus, elevated corticosterone in large forest patches may not imply nutritional stress but may instead mark an adaptive, pre‐fledging mobilization of energy reserves. Proportionally smaller mass relative to wing length can therefore reflect two non‐exclusive processes: (i) accelerated pre‐fledging development, or (ii) nutritional stress and developmental compromises imposed by environmental constraints (Ashton and Armstrong 2002; Starck and Ricklefs 1998). Longer wings, higher weight, and SMI with elevated corticosterone levels in larger forest patches are consistent with accelerated fledging, whereas longer wings with reduced SMI and provisioning rates point to nutritional or thermal stress during hot weather especially in small forest patches.

Morphological Plasticity in Response to Heat Stress

4.5

Beyond the role of food limitation, our results suggest that nestlings show morphological adjustments in response to heat stress, especially in degraded areas, in line with our third prediction (Figure 2a,b). Specifically, we found that at low temperatures, nestlings in small forest patches and areas with low canopy height (less‐buffered habitats) grew shorter wings and tarsi than those in larger patches and in areas with high canopy height (better‐buffered habitats). At high temperatures, however, nestlings in less‐buffered habitats tended to grow longer tarsi and wings reaching lengths similar to those observed in better‐buffered habitats, potentially reflecting temperature‐associated developmental plasticity that could facilitate heat dissipation, consistent with expectations under Allen's rule (Allen 1877). While wing length was relatively independent of temperature in large forest patches, we found a negative relationship between temperature and tarsus length in areas with high canopy height. This contrasting pattern could have multiple explanations. Elevated corticosterone in these better‐buffered habitats may inhibit skeletal growth (Durant et al. 2010; Muller et al. 2009), particularly in bone structures such as the tarsus. Alternatively, nestlings in high‐quality patches may simply not need to alter tarsus growth because canopy and microclimate buffering already provide protection from heat (De Frenne et al. 2019; Terschanski et al. 2024). Another possibility is that developmental trade‐offs in locomotory traits play a role, for example if resources are preferentially allocated to wing growth at the expense of tarsus growth, or if reduced time spent on the ground decreases the need for longer tarsi. Our findings that nestlings prioritized wing growth against mass gain, that is, longer wings, but lower SMI, at high temperatures may also indicate thermoregulatory plasticity. In addition to being critical for successful emergence especially under high predation risk (Martin 2015; Rivers et al. 2012), wings may also play a role in passive heat dissipation not only in nestlings, but also in adult birds (Diehl et al. 2023; Pattinson et al. 2020). In the presence of a trade‐off between mass gain and rapid development of flight‐ready wings, nestlings may exhibit a developmental bias towards wing growth, particularly when resources are limited, and temperatures are high. This finding parallels a pattern reported by Jirinec et al. (2021) in adult non‐migratory understory birds of the central Amazon, where mass‐to‐wing length ratios decreased with increasing temperature, suggesting that lighter, longer‐winged morphologies may be advantageous under warming conditions.

Multiple Interacting Drivers of Nestling Development

4.6

While our results broadly support the notion that resource limitation leads to poorer nestling condition in low‐quality habitats and that reduced microclimatic buffering can drive morphological adjustments to facilitate heat loss, some patterns appear contradictory. For example, nestlings in higher quality habitats might be expected to show minimal heat‐stress responses, yet tarsus length actually decreased with temperature in large forest patches and high‐canopy‐height habitats, apparently reflecting a different developmental trajectory. As noted above, corticosterone may underlie this phenomenon, promoting earlier readiness to emerge but inhibiting certain aspects of skeletal growth (Durant et al. 2010; Muller et al. 2009; Rivers et al. 2012; Romero et al. 2005). This contrast highlights how nestling development can be shaped by multiple, sometimes opposing processes, including resource availability, microclimate buffering, thermoregulatory needs, predation risk, and hormonal regulation. Another seeming contradiction was the lacking relationship between forest quality and provisioning rates, despite nestlings being in poorer condition in smaller, degraded forest patches. However, our results are consistent with other studies showing that prey abundance alone does not always reflect nestling condition; prey type, size, and nutritional value are equally important (Barras et al. 2021; Hakkarainen et al. 1997; Sinkovics et al. 2021; Wilkin et al. 2009). If arthropods in degraded areas are not only fewer overall, but also skewed towards less nutritious taxa, such as ants or beetles with harder exoskeletons, or if larger or more protein‐rich items are comparatively rare (Gladalski et al. 2019; Sinkovics et al. 2021), nestlings will fail to gain sufficient mass no matter how often they are fed. This dynamic may be exacerbated during particularly hot or dry periods, when arthropod communities can shift in ways that reduce both abundance and nutritional quality of prey available to nestlings, consistent with seasonal deterioration in food resources observed in temperate systems (Arnold et al. 2010). Crucially, these climate‐driven changes in prey availability can coincide with thermal constraints on adult foraging behaviour, limiting parents' ability to compensate for poor‐quality prey through increased provisioning or selective prey choice. Similar foraging trade‐offs under thermal stress have been reported in other species; for example, pied babbler parents reduce provisioning effort during hot periods and rely on non‐breeding helpers to compensate (Wiley and Ridley 2016). Recent work in our study area also shows that dominant breeders in degraded patches reduce midday activity, likely to minimize heat stress but potentially at the cost of provisioning high‐quality prey (Kung'u et al. 2025). Similarly, understorey insectivores in the Amazon have been found to prefer dark and cool refugia and avoidance of bright and warm conditions, indicating behavioral adjustments to thermal stress that may constrain foraging opportunities (Jirinec et al. 2022). Thus, a better understanding of the relationship between habitat structure, arthropod communities, prey nutritional quality, particularly under hot and dry conditions, is needed to explain why parents may be unable to fully buffer offspring in degraded habitats. Taken together, these complex interactions highlight the importance of understanding how microclimate buffering, vegetation structure, predator pressure, and the timing of nestling development each contribute to fitness outcomes. While simple metrics (e.g., provisioning rate) can be informative, our results highlight the need to integrate broader ecological and behavioural contexts, such as thermoregulatory strategies (Kosheleff and Anderson 2009; Lehmann et al. 2012), prey quality (Barras et al. 2021; Schwagmeyer and Mock 2008; Wilkin et al. 2009), and species life history strategies (long‐lived vs. short‐lived, social or non‐social; Arco et al. 2022; Crick 1992; Meade et al. 2010; Wiley and Ridley 2016), to fully explain patterns in nestling condition. While it is likely that our findings also apply to other forest specialist bird species, there is a need for including further species in future research to assess the generality of our findings.

Conclusion and Conservation Implications

5

Overall, our study provides strong evidence that reduced forest quality and hotter, drier conditions jointly reduce nestling body condition. However, high precipitation can temporarily offset the disadvantage of poor canopy cover, allowing nestlings in degraded areas to reach high SMI levels. This highlights the complexity of climate–habitat interactions, particularly in tropical forest fragments where rainfall variability can be pronounced. Our findings of morphological adjustments, such as reduced mass and prioritization of the length of appendages over mass in response to warming conditions, are consistent with Bergmann's and Allen's rules, suggesting that such plasticity can be expressed early in life. While these shifts may offer short‐term thermal benefits, the long‐term fitness consequences remain uncertain, particularly in the face of increasing climate extremes. Understanding whether reduced mass or relatively longer limbs result in disadvantages or trade‐offs, such as delayed maturity, reduced competitiveness, or lower adult survival, will be essential for predicting population trajectories under ongoing forest fragmentation and climate change (Jirinec et al. 2021; Stouffer et al. 2021). Our results highlight the critical role of vegetation structure, particularly intact canopy structure, in buffering nestlings against thermal stress and low rainfall. Equally, they underscore the importance of considering how physiological and behavioral adaptations, such as corticosterone‐induced readiness to fledge and altered provisioning schedules, may mitigate or amplify the effects of habitat and climate stressors. Multiple, and sometimes opposing, developmental mechanisms (e.g., hormonal, thermoregulatory, locomotory trade‐offs) may act simultaneously, reinforcing the need for integrative approaches. By integrating morphological, behavioral, and physiological indicators, our study illustrates how a multi‐metric approach can resolve ambiguities and provide a more nuanced understanding of nestling responses to interacting habitat and climate stressors. Protecting and restoring tropical forests with intact canopy structure will hence be crucial for sustaining viable populations of tropical understory insectivores in the face of accelerating habitat loss and climate change.

Author Contributions

Gladys Nyakeru Kung'u: writing – review and editing, writing – original draft, visualization, formal analysis, data curation, funding acquisition, conceptualization. Laurence Cousseau: writing – review and editing, data curation, validation, funding acquisition, conceptualization. Virginie Canoine: writing – review and editing, methodology, data curation. Janne Heiskanen: writing – review and editing, data curation. Mwangi Githiru: writing – review and editing, resources, project administration, conceptualization. Peter Njoroge: writing – review and editing, project administration. Petri Pellikka: writing – review and editing, funding acquisition, data curation. Jan Christian Habel: writing – review and editing, Supervision. Luc Lens: writing – review and editing, methodology, supervision, funding acquisition, conceptualization. Beate Apfelbeck: writing – review and editing, validation, supervision, methodology, investigation, funding acquisition, visualization, formal analysis, data curation, conceptualization.

Funding

This study was funded in whole or in part by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) – 392075127, the Austrian Science Fund (FWF) [10.55776/I6837] awarded to BA and the Research Foundation Flanders (FWO‐grant G.0ABI.24N) awarded to LL and LC. For open access purposes, the author has applied a CC BY public copyright license to any author accepted manuscript version arising from this submission. Corticosterone analysis was funded by the German Ornithologist's Society (DOG) and LiDAR data collection by the European Commission DG International Partnerships under DeSIRA programme (FOOD/2020/418‐132), PI PP. G.N.K. is a recipient of a DOC Fellowship of the Austrian Academy of Sciences at the University of Salzburg.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Appendix S1: gcb70771‐sup‐0001‐AppendixS1.pdf.