The Black Stork (Ciconia nigra) is a new winter resident in Beijing due to temperature changes.To understand the wintering ecology of this species better, a field survey covering the number of birds of this population, habitat selection, feeding activity and grouping behavior was conducted at the Shidu Nature Reserve from January 2004 to March 2009. The results show that the Black Stork selected the Juma River at this nature reserve as their new winter habitat. The number of birds in this population decreased from 28 in the 2004/2005 winter to 17 in the 2007/2008 winter with a subsequent recovery to 23 the following year. The wintering flock was formed in mid-November and dispersed in mid-March, but the date changed with seasonal temperature fluctuations. The storks exhibited feeding habitat fidelity and the main food type was fish (＞ 92.4%). There was no significant variation in food composition between adults and sub-adults (Mann-Whitney U test, U = 1.00, p = 0.44). Feeding activity occurred in the morning and at noon during early winter, but concentrated in the afternoon during mid winter, divided into dawn and dusk in late winter. Daily fish intake was 538 g for adults and 449 g for sub-adults if the period of foraging reached six hours in the wild, which was similar to the level under artificial feeding. Agonistic behavior among feeding birds was observed among group members in late winter. The main negative factor for wintering Black Stork was a reduced feeding habitat resulting from increased water depth due to damming of the river to benefit tourism and to wetland exploitation.
We present a preliminary examination of the home range and habitat use of male Reeves's Pheasants (Syrmaticus reevesii) in an agricultural-forest plantation landscape on the Xianjuding Forest Farm, Hubei Province, central China. Fieldwork was carried out from March to August in 2003. The home range of males averaged 33.17 ± 12.55 ha by MCP (minimum convex polygon) and 21.05 ± 5.61 ha by a 95% fixed kernel estimator. The core area by a 60% fixed kernel estimator was 3.92 ± 0.27 ha. We did not detect significant seasonal variations in home range, core area and movement in this farm. Chinese fir (Cunninghamia lanceolata) plantations were the dominant habitat type within the home ranges and core areas. The males used their habitat non-randomly in spring and summer, preferred the fir plantations and avoided broadleaf forests in both seasons. In addition, the males used shrubs less than were available in the spring. The vegetation structure of different habitats may be the leading factor affecting the use of the habitats.
北京地区珍稀鸟类的保护对维护当地生物多样性具有重要意义。随着人工智能技术的发展,利用深度学习技术自动识别鸟类成为鸟类调查保护的重要手段。实际鸟类图像存在背景复杂以及相近科属鸟类具有外观相似等特点,导致模型识别精度不佳。针对以上问题,本文提出一种基于TC-YOLO模型的鸟类识别方法。首先,为解决鸟类识别中复杂背景导致的漏检问题,本文方法结合CARAFE(content-aware reassembly of features)机制,自适应生成不同特征点所对应的上采样核,在更大的感受野内聚合上下文语义信息,有效聚焦鸟类前景区域。其次,为解决鸟类识别中相似外观导致的误检问题,本文方法引入TSCODE(task-specificcontextdecoupling)解耦定位和分类任务,通过获取多层级特征图的信息以回归目标边界,并利用包含底层纹理和高层语义的特征进行物种分类,进而提高模型的鸟类识别精度。最后,本文开展对比实验以验证模型的性能。实验结果表明,TC-YOLO模型的平均精度均值在包含北京地区28种国家一级保护鸟类的自建数据集Beijing-28和鸟类公开数据集CUB200-2011上分别达到78.7%和75.3%,均优于已有方法,而且在公开数据集MS COCO上验证了TC-YOLO模型拥有较强的泛化性。本文提出的TC-YOLO模型对背景复杂或外观相似的鸟类图像都能有效识别,漏检率和误检率较低,能够为鸟类保护提供重要技术支撑。
