Effective implementation of silvopastoralism, a key Nature-Based Solution for Europe's climate goals, is hindered by a lack of decision-support tools clarifying trade-offs between efficiency and extent of carbon sequestration. To address this, we developed a multi-objective scenario analysis (4064 scenarios) to identify optimal strategies for silvopastoral expansion across the EU27 Mediterranean bioregion. We found an inverse relationship defining a clear trade-off: scenarios achieving the highest mean sequestration (up to 2.5 Mg CO2 ha-1 year-1) are spatially limited, whereas those maximising total gains (approaching 107 Mg CO2 year-1 in total) do so by incorporating vast areas, lowering mean rates. This trade-off is formalised by a Pareto front, from which we defined a best-balanced optimal scenario and three policy regimes (conservative, balanced, expansive). Progressing across the front involved shifting from converting primarily shrubby and sparsely vegetated lands to incorporating grasslands and mixed agro-systems. At the NUTS2 level, Spain and Greece emerged as hotspots. Notably, converting arable land was not a primary contributor to carbon gains, as the marginal carbon benefit on these productive soils is lower than on marginal lands due to their higher baseline soil carbon levels, indicating that large-scale implementation can focus on marginal lands to avoid conflicts with food security. While subject to uncertainties of the underlying land-use and carbon models, this analysis demonstrates that our framework enables policymakers to select spatially explicit strategies aligned with specific budget or sequestration goals. These insights can inform CAP eco-schemes and national LULUCF strategies. The resulting maps and code are freely available.

Wildfires are major modifiers of Mediterranean forest ecosystems, with long-term impacts on biodiversity and ecological resilience. However, how fire-history contrasts and seasonal dynamics jointly relate to avian community structure remain insufficiently understood in Mediterranean pine forests. To address this gap, we assessed how avian communities vary across sites with different time since fire using passive acoustic monitoring (PAM) across four Pinus halepensis forests in northern Greece representing a post-fire chronosequence: unburned for at least 40 years and burned in 2001, 2009, and 2018. At each site, autonomous AudioMoth recorders captured 10-min soundscape samples every 30 min for ten consecutive days per season. Bird vocalizations within each sample were identified to species level using BirdNET (v2.4) followed by expert validation of a subset of detections to reduce misclassification. We used Non-metric Multidimensional Scaling (NMDS) to visualize how bird community composition varied among sites and seasons, and Bray–Curtis dissimilarities to describe differences in species composition. We detected 43 species and found that both seasonality and fire history significantly structured avian assemblages, with seasonal variation explaining the largest portion of community differences. Burned sites were dominated by generalist and shrubland species, while unburned forests supported mature-forest specialists. Indicator species analysis identified 13 taxa associated with different successional stages. Overall bird communities showed variation across sites with contrasting fire histories, reflecting potential changes in habitat structure across successional stages, although these patterns should be interpreted cautiously given the limited number of fire-history categories. Our findings highlight the importance of integrating fire history and seasonal dynamics in biodiversity monitoring, and support PAM as a scalable tool for evaluating ecological recovery in fire-affected Mediterranean landscapes.

Greater botanical diversity in grazing systems can enhance animal nutrition through improved forage quality and complementary resource use. However, biodiversity effects in managed grasslands cannot be inferred from species richness alone. Understanding how grazing herbivores respond behaviourally to structurally and chemically heterogeneous swards is essential for predicting biodiversity–function relationships in pastoral systems. This study evaluated sheep foraging behaviour in response to pasture diversity and spatial configuration. Two pasture systems (P1 and P2) were established using five grasses and four legumes. In P1, spatially homogenous grass–legume mixtures were arranged across three pasture units differing in species richness: (1) Lolium perenne and Trifolium pratense, (2) Lolium perenne, Festuca arundinacea, Trifolium pratense and Onobrychis viciifolia, and (3) Lolium perenne, Lolium multiflorum, Festuca arundinacea, Trifolium pratense, Trifolium repens and Onobrychis viciifolia. In P2, a homogenous grass–legume mixture consisted of Dactylis glomerata, Phleum pratense, Medicago sativa and Onobrychis viciifolia was compared with spatially segregated monocultures of the same species. Grazing trials were conducted over seven consecutive days in June 2025, and foraging behaviour was recorded using systematic scan sampling at 1-min intervals during daily 30-min morning sessions. In P1, grazing activity increased with species richness, with sheep grazing more in the most diverse unit (83.9%) than in intermediate (80.6%) and low-diversity units (75.1%) (p < 0.05). In P2, grazing activity was greater in spatially segregated monocultures (86.3%) than in the aggregated mixture (80.2%) (p < 0.05), with Onobrychis viciifolia and Medicago sativa most preferred. These findings indicate that both botanical diversity and spatial arrangement influence grazing efficiency. Sheep adjusted selective behaviour to balance nutrients and secondary compounds.

Recent advances in machine learning have accelerated automated species detection across diverse ecological domains, enabling large-scale, non-invasive monitoring of biodiversity. In ornithological research, the combination of passive acoustic monitoring (PAM) and rapidly-developing novel identification tools such as BirdNET—a deep learning–based sound recognition algorithm—offers new opportunities for surveying vocally active bird communities. Here, we present the first worldwide evaluation of BirdNET using 4224 one-minute recordings from 67 sites across all continents annotated by local experts. More specifically, we assessed the capacity of BirdNET to accurately identify individual vocalizations and characterize bird communities based on the automated analysis of passively collected soundscapes. We further analyzed how its performance varies across continents, biomes, species, and minimum confidence thresholds. The proportion of correct BirdNET predictions (precision) was generally high and consistent across continents (range: 0.57–0.71) and biomes (range: 0.55–0.76). In contrast, the proportion of vocalizations successfully detected (recall) was generally lower and more heterogeneous across continents (range: 0.24–0.52) and biomes (range: 0.34–0.72), reflecting differences in species coverage and local ecological context. BirdNET predictive power, as measured by the Precision-Recall Area Under the Curve (PR AUC; higher values indicating better performance), was highest in North America, Oceania, and Europe (range: 0.16–0.23), moderate in Central/South America (0.13), and lowest in Africa and Asia (range: 0.03–0.04). Species-specific analyses revealed substantial heterogeneity in detection accuracy, with optimal confidence thresholds varying widely by species and analytical goal. Our results establish a global reference point for BirdNET reliability and highlight where algorithmic refinement and expanded acoustic sampling are most needed.

Foraging behaviour of grazing animals is influenced by the diversity of available forage, its spatial arrangement, and the presence of plant secondary compounds (PSCs), which can affect intake patterns and dietary choices. This study aimed to evaluate how different pasture compositions and spatial configurations, along with variations in PSC concentrations, influence the foraging behaviour of sheep in grass–legume systems. Six pasture types were established in northern Greece, including monocultures, adjacent monocultures, strip plantings, and homogeneous grass–legume mixtures with varying ratios. Twenty-four ewes grazed these pastures for 30 minutes daily over 10-day trials. Grazing behaviour was recorded using scan sampling, and herbage mass and PSC concentrations were measured. Herbage mass was higher in mixed pastures than in monocultures or strip systems. Grasses had higher alkaloid concentrations, while legumes had higher condensed tannins. Grazing activity was greatest in homogeneous mixtures (77.0–79.7%) and lowest in tall fescue monoculture (40.3%) but increased when legumes were present. Prior consumption of legumes enhanced subsequent grazing on tall fescue (43.9%), indicating sequence-dependent behaviour. Sheep showed clear species preferences, favouring sainfoin (40.0%) over birdsfoot trefoil (33.8%), and both over tall fescue (26.2%) in mixed systems. Combining grasses and legumes with complementary nutritional and phytochemical traits, especially in homogeneous mixtures, improves grazing activity and supports adaptive foraging strategies in sheep. Designing pastures with diverse species and appropriate spatial arrangements can enhance pasture utilization, improve animal performance, and support more sustainable grazing systems.

The potential Evapotranspiration (PET) is a crucial hydrological variable for understanding the water balance and ecosystem dynamics, particularly in different altitudinal Mediterranean forest environments. Traditional empirical methods for estimating PET have frequently failed to generalize accurately to heterogeneous environments, including Mediterranean forests, due to climatic and topographic diversity. This study develops and evaluates bi-directional transferable Deep Learning (DL) models for PET estimation using meteorological and environmental data collected from Mediterranean forest areas, namely Pertouli and Taxiarchis, covering multiple altitudinal levels. The models were trained and evaluated in both directions: first trained on Pertouli and transferred to Taxiarchis, and second trained on Taxiarchis and transferred to Pertouli. These models were deployed and retrained using Transfer Learning (TL) for PET estimation to address the limited data availability at the target site. Several model architectures were implemented and evaluated in different Scenarios. The experimental findings show that the transferable DL models achieve R2 values between 0.7 and 0.87 and RMSE values from 0.55 to 0.75, according to the data availability of the target site and the meteorological data used for training. Among the evaluated model architectures, the RNN model achieved the highest performance with R2 = 0.767, MAE = 0.521, RMSE = 0.668, and sRPIQ = 0.755 when the training data were limited, while the LSTM obtained the highest accuracy under higher data availability conditions, namely, R2 = 0.873, MAE = 0.384, RMSE = 0.55, and sRPIQ = 0.831. This work demonstrates the feasibility of bi-directional model transfer across altitudinal forest gradients under Mediterranean conditions.

Under the current global biodiversity crisis, there is a need for automated and noninvasive monitoring techniques that can gather large amounts of data cost-effectively at various ecological scales, from local to large spatial scales. These data can then be analyzed to inform stakeholders and decision-makers. One such technique is passive acoustic monitoring, which is commonly coupled with automatic identification of animal species based on their sound. Automated sound analyses usually require the training of sound detection and identification algorithms. These algorithms are based on annotated acoustic datasets which mark the occurrence of sounds of species inside sound recordings. However, compiling large annotated acoustic datasets is time-consuming and requires experts, and therefore, they normally cover reduced spatial, temporal, and taxonomic scales. This data paper presents WABAD, the World Annotated Bird Acoustic Dataset for passive acoustic monitoring. WABAD is designed to provide the public, the research community, and conservation managers with a novel and globally representative annotated acoustic dataset. This database includes 5047 min of audio files annotated to species-level by local experts with the start and end time and the upper and lower frequencies of each identified bird vocalization in the recordings. The database has a wide taxonomic and spatial coverage, including information on 91,931 vocalizations from 1192 bird species recorded at 72 recording sites in 29 recording locations (mainly countries) and distributed across 13 biomes. WABAD can be used, for example, for developing and/or validating automatic species detection algorithms, answering ecological questions, such as assessing geographical variations on bird vocalizations, or comparing acoustic diversity indices with species-based diversity indices. The dataset is published under a Creative Commons Attribution 4.0 International license that permits redistribution and reuse on the condition that the original work is properly credited.

Accurate estimation of potential evapotranspiration (PET) is of paramount importance for water resource management, especially in Mediterranean mountainous environments that are often data-scarce and highly sensitive to climate variability. This study evaluates the performance of four machine learning (ML) regression algorithms—Support Vector Regression (SVR), Random Forest Regression (RFR), Gradient Boosting Regression (GBR), and K-Nearest Neighbors (KNN)—in predicting daily PET using limited meteorological data from a high-altitude in Central Greece. The ML models were trained and tested using easily available meteorological inputs—temperature, relative humidity, and extraterrestrial solar radiation—on a dataset covering 11 years (2012–2023). Among the tested configurations, RFR showed the best performance (R2 = 0.917, RMSE = 0.468 mm/d, MAPE = 0.119 mm/d) when all the above-mentioned input variables were included, closely approximating FAO56–PM outputs. Results bring to light the potential of machine learning models to reliably estimate PET in data-scarce conditions, with RFR outperforming others, whereas the inclusion of the easily estimated extraterrestrial radiation parameter in the ML models training enhances PET prediction accuracy.

Under the current global biodiversity crisis, there is a need for automated and non-invasive monitoring techniques that are able to gather large amounts of information cost-effectively at large scales. One such technique is passive acoustic monitoring, which is commonly coupled with automatic identification of animal species based on their sound. Automated sound analyses usually require the training of sound detection and identification algorithms. These algorithms are based on annotated acoustic datasets which mark the occurrence of sounds of particular species. However, compiling large annotated acoustic datasets is time consuming and requires experts, and therefore they normally cover a reduced spatial and taxonomic scale. This data paper presents WABAD, the World Annotated Bird Acoustic Dataset for passive acoustic monitoring. WABAD is designed to provide the public, the research community, and conservation managers with a novel and globally representative annotated acoustic dataset. This database includes 5,044 minutes of audio files annotated to species-level by local experts with the start and end time, and the upper and lower frequencies of each identified bird vocalisation in the recordings. The database has a wide taxonomic and spatial coverage, including information on 90,662 vocalisations from 1,147 bird species recorded at 70 recording sites in 27 countries and distributed across 13 biomes. WABAD can be used, for example, for developing and/or validating automatic species detection algorithms, answering ecological questions, such as assessing geographical variations on bird vocalisations, or comparing acoustic diversity indices with species-based diversity indices. The dataset is published under a Creative Commons Attribution Non Commercial 4.0 International copyright.

BirdNET is a popular machine learning tool for automated recognition of bird sounds. However, evidence on how to optimize its settings for accurate bird monitoring remains limited. Here, we evaluate how BirdNET settings influence model performance in identifying bird vocalizations and characterizing bird communities, using 4224 1-min recordings from 67 recording locations worldwide. Giving equal importance to recall and precision, a low confidence score threshold (0.1-0.3) appears optimal for detecting bird vocalizations, whereas higher thresholds (around 0.5) are more suitable for characterizing bird communities. Based on our findings, we recommend increasing the Overlap parameter from its default value of 0 to 2 s, as this consistently improves BirdNET performance in detecting both bird vocalizations and species presence. The effect of the Sensitivity parameter varied across regions. However, a value of 0.5 maximizes global performance for community-level analyses across all confidence thresholds, and a value of 1.5 generally yields better results for vocalization-level studies, particularly at low confidence thresholds. Our findings offer practical guidance for selecting BirdNET settings in passive acoustic bird surveys, enhancing both the identification of bird vocalizations and the characterization of bird communities.

Agroforestry has a long history of evolution in Europe and has been especially selected under the unfavorable socioeconomic and environmental conditions of the Mediterranean region. The recent changes in social-ecological conditions have increased the interest in the contribution of agroforestry to the mitigation of forthcoming challenges. Thus, the present study aimed to analyze the socioeconomic and ecological suitability of agricultural lands for preserving, restoring, and establishing agroforestry practices in Europe. We classified different agroforestry systems based on the LUCAS database, finding that most agroforestry in Europe is in areas associated with older human populations of varying densities and employment levels at lower altitudes, gentler slopes, moderate annual mean temperature and precipitation, and in medium textured soils with limited organic carbon content. Focusing on the prevalent agroforestry system of silvopasture, the majority of which is found in three Mediterranean ecoregions of mainly sclerophyllous forests, the most important factors for the occurrence of this system were subsoil available water content (Aegean), land cover (Adriatic), and topsoil available water content (Iberian). The suitable area for silvopasture according to MaxEnt was 32%, 30%, and 22% of the Aegean, Adriatic, and Iberian ecoregion’s area, respectively. Such mapping of agroforestry suitability can help policymakers to undertake adaptive management for the implementation of agroforestry-based solutions to address ecosystem restoration, food insecurity, and rapid environmental changes and threats.

In recent decades, climate change has significantly influenced the frequency and intensity of wildfires across Mediterranean pine forests. The loss of forest cover can bring long-term ecological changes that impact the overall biodiversity and alter species composition. Understanding the long-term impact of wildfires requires effective and cost-efficient methods for monitoring the postfire ecosystem dynamics. Passive acoustic monitoring (PAM) has been increasingly used to monitor the biodiversity of vocal species at large spatial and temporal scales. Using acoustic indices, where the biodiversity of an area is inferred from the overall structure of the soundscape, rather than the more labor-intensive identification of individual species, has yielded mixed results, emphasizing the importance of testing their efficacy at the regional level. In this study, we examined whether widely used acoustic indicators were effective at capturing changes in the avifauna diversity in Pinus halepensis forest stands with different fire burning histories (burnt in 2001, 2009, and 2018 and unburnt for >20 years) on the Sithonia Peninsula, Greece. We recorded the soundscape of each stand using two–three sensors across 11 days of each season from March 2022 to January 2023. We calculated for each site and season the following five acoustic indices: the Acoustic Complexity Index (ACI), Acoustic Diversity Index (ADI), Acoustic Evenness Index (AEI), Normalized Difference Soundscape Index (NDSI), and Bioacoustic Index (BI). Each acoustic index was then assessed in terms of its efficacy at predicting the local avifauna diversity, as estimated via two proxies—the species richness (SR) and the Shannon Diversity Index (SDI) of vocal bird calls. Both the SR and SDI were calculated by having an expert review the species identification of calls detected within the same acoustic dataset by the BirdNET convolutional neural network algorithm. A total of 53 bird species were identified. Our analysis shows that the BI and NDSI have the highest potential for monitoring the postfire biodiversity dynamics in Mediterranean pine forests. We propose the development of regional-scale acoustic observatories at pine and other fire-prone Mediterranean habitats, which will further improve our understanding of how to make the best use of acoustic indices as a tool for rapid biodiversity assessments.

Coastal wetlands are considered as systems of high avian diversity and are usually used for livestock production throughout the world. In this study, the diversity and seasonal abundance of avian species were monitored for two years on a monthly basis in a coastal grazing land in Evros Delta (Greece). The effects of livestock (cattle) presence and different classes of vegetation cover on the number of bird species were also investigated. A total of 96 bird species belonging to 29 families were recorded. The most commonly encountered species was the Eurasian skylark Alauda arvensis. The cattle presence was not significantly correlated (p>0.05) with the abundance of recorded bird species. On the contrary, patches with vegetation cover 25.1 - 50.0% and 50.1 - 75.0 % were used by more bird species in relation to patches with cover ≤25.0% or >75.0%. We concluded that the use of livestock grazing to preserve the desired vegetation cover (25 – 75%) is a promising management tool.

Silvopasture, a traditional agroforestry practice, combines the presence of trees, shrubs, herbage, and livestock in time and space to provide multiple ecosystem services that contribute to human well-being. However, the abandonment of traditional agroforestry practices across Europe has led to substantial changes in vegetation characteristics, mainly due to woody plant expansion and, as a consequence, changes in wildlife that rely on open habitats. This study examines the effects of a 20-year abandonment of silvopastoral practices (i.e., livestock grazing and fuelwood harvesting) in a typical agroforestry Mediterranean landscape (kermes oak shrubland, natural grassland, and olive groves) on European hare (Lepus europaeus) habitat use. We estimated tree, shrub, and herb cover using a densitometer and hare habitat use using pellet counts within 2004-m(2) rectangular plots in 2002, 2011, and 2021. Hare pellet density in olive groves was significantly lower in 2021 compared to 2002, while the opposite trend was found in grassland for the same period. Woody plant cover expanded from 2002 to 2021. We suggest that the woody plant encroachment that followed the abandonment of traditional silvopastoral practices in the area is the main driver behind the reported decline in hare use of the habitat, as it became less open and therefore less favorable for the species. Maintaining a mosaic of open and closed habitats at the landscape level, which was once provided by silvopastures, is vital for the conservation of this species.

There has been little research on the distribution and ecology of the four dormouse species occurring in Greece; the Edible Dormouse (Glis glis), Forest Dormouse (Dryomys nitedula), Hazel dormouse (Muscardinus avellanarius) and Mouse-tailed Dormouse (Myomimus roachi). As a result, the latter three species are listed as data deficient (DD) in the National Red Data Book. Recently, the government has tried to address this knowledge gap, funding dormouse surveys within the Natura 2000 network. In this context, we used a combination of nest-tubes (n=442) and track-tunnels (n=238) to study dormouse distribution and habitat use across 37 sites representing different habitat types (with varying levels of grazing) of two mountainous N2K sites (GR1270001, GR1270005) in central Macedonia. We detected G. glis at 28 sites, D. nitedula at 32 sites, and M. avellanarius at seven sites. Positive identification of the different species was twice as likely in track-tunnels (unbaited; metal sheets covered in soot) than nest-tubes. We estimated relative abundance across sites using Royle-Nichols occupancy models, except for M. avellanarius due to data limitations. For all species, we examined habitat use using MaxEnt ecological-niche models. Our findings show that D. nitedula has the widest distribution, occurring even in sparse forests and maqui with moderate or high livestock grazing intensity. G. glis is common, but restricted to medium-high elevation forests. M. avellanarius appears to have a discontinuous distribution. If this study is representative of its status across the country, that species requires conservation efforts.

For large herbivores inhabiting arid/semi-arid environments, water can be a limiting resource affecting their distribution and abundance for periods when water requirements are not met via forage. The Cyprus mouflon (Ovis gmelini ophion) is such a species, which is endemic to the mountain habitats of Cyprus. Recognizing water scarcity to be a major pressure to the mouflon, and with global warming projected to intensify hot and dry periods in the region, the Game and Fauna Service has been maintaining a network of locally designed watering troughs in Pafos Forest—the mouflon’s stronghold—since 1997. This study describes the mouflon’s use of the water troughs and examines whether visitation rates differed at the daily or weekly scale in response to environmental, climatic or anthropogenic parameters. Using camera traps, ten troughs were monitored from September 2017 to March 2018 (1,065 days; range 29–164 days per trough). Mouflon were detected at seven troughs (mean herd size 1.5 ± 1.2) during 373 independent detections (≥30 min interval between photographs), with visits peaking during late morning and midday hours. Generalized mixed-effect models showed mouflon visiting water troughs more frequently during hotter days, regardless of recent precipitation. Visits were also more frequent at water troughs located close to tar roads. Moreover, there was no evidence of mouflon avoiding water troughs used by predators (red foxes, feral dogs) at either daily or weekly scale, or during hunting days. The study supports the value of artificial water troughs for mediating, partially at least, the effects of climate change on mountain ungulates such as the Cyprus mouflon. Additional studies are proposed that will examine both mouflon drinking patterns across all seasons and ways of improving the effectiveness of the current water trough grid.

Automatic Weather Stations (AWS) are extensively used for gathering meteorological and climatic data. The World Meteorological Organization (WMO) provides publications with guidelines for the implementation, installation, and usages of these stations. Nowadays, in the new era of the Internet of Things, there is an ever-increasing necessity for the implementation of automatic observing systems that will provide scientists with the real-time data needed to design and apply proper environmental policy. In this paper, an extended review is performed regarding the technologies currently used for the implementation of Automatic Weather Stations. Furthermore, we also present the usage of new emerging technologies such as the Internet of Things, Edge Computing, Deep Learning, LPWAN, etc. in the implementation of future AWS-based observation systems. Finally, we present a case study and results from a testbed AWS (project AgroComp) developed by our research team. The results include test measurements from low-cost sensors installed on the unit and predictions provided by Deep Learning algorithms running locally.

Grazing has long been recognized as an effective means of modifying natural habitats and, by extension, as a wildlife and protected area management tool, in addition to the obvious economic value it has for pastoral communities. A holistic approach to grazing management requires the estimation of grazing timing, frequency, and season length, as well as the overall grazing intensity. However, traditional grazing monitoring methods require frequent field visits, which can be labor intensive and logistically demanding to implement, especially in remote areas. Questionnaire surveys of farmers are also widely used to collect information on grazing parameters, however there can be concerns regarding the reliability of the data collected. To improve the reliability of grazing data collected and decrease the required labor, we tested for the first time whether a novel combination of autonomous recording units and the semi-automated detection algorithms of livestock vocalizations could provide insight on grazing activity at the selected areas of the Greek Rhodope mountain range. Our results confirm the potential of passive acoustic monitoring (PAM) techniques as a cost-efficient method for acquiring high resolution spatiotemporal data on grazing patterns. Additionally, we evaluate the three algorithms that we developed for detecting cattle, sheep/goat, and livestock bell sounds, and make them available to the broader scientific community. We conclude with suggestions on ways that acoustic monitoring can further contribute to managing legal and illegal grazing, and offer a list of priorities for related future research.

Summary The Fennoscandian population of the Lesser White-fronted Goose Anser erythropus (LWfG) is on the verge of extinction and migrates from northern Fennoscandia to Greece on a regular seasonal basis. For the first time, diet selection was investigated during two years at Kerkini Lake, a wintering site in Greece. The relative use of LWfG’s feeding habitats was systematically recorded by visual observations of the LWfG flocks. Food availability was measured by the relative cover of available vegetation types while the diet composition was determined by the microhistological analysis of droppings. In addition, we determined crude protein, neutral detergent fibre, acid detergent fibre and acid detergent lignin content of the most preferred plant species by LWfG and all vegetation categories that contributed to LWfG diet in the middle of the duration of their stay at Kerkini Lake and after their departure from the lake. LWfG feeding habitat was exclusively marshy grassland in water less than 5 cm deep up to 300–400 m away from the shore. LWfG selected a diverse number of plant species (33), however, grass made up the 58% of their diets. The most preferred plant species were Echinochloa crus-galli , Cyperus esculentus , Scirpus lacustris and Ranunculus sceleratus . LWfG departed from Kerkini Lake in mid-December to the Evros Delta (Thrace, eastern Greece), when either food availability falls in very low levels or flooding occurred in their main feeding habitat. Consequently, as long as food and habitat resources are available for LWfG, it is very likely that the birds will winter mainly at Kerkini Lake and not at the Evros Delta, which will contribute to further minimisation of the theoretical risk of accidental shooting of LWfG at the latter wintering habitat. Thus, future conservation actions should primarily focus on the grassland improvement at Kerkini Lake enhancing the availability of food resources for LWfG (mainly grasses) and the protection of the feeding habitat from flooding.