During an examination of various specimens previously collected from different locations and times, we discovered seven Murina specimens that had been collected in May 2023 from the Guanyinshan Provincial Nature Reserve, Yuanyang, Yunnan, China. Based on the assessment of morphological characteristics and molecular data analysis, these specimens were determined to represent a previously unidentified species, designated Murina yuanyang sp. nov.

During an examination of various specimens previously collected from different locations at different times, we discovered four Murina specimens that had been collected in October and December 2023 from the Huanglianshan National Nature Reserve, Lvchun, Yunnan, China. Morphologically, these specimens can be distinguished from M. pluvialis and other congeneric species based on a combination of body size, hair distribution, fur colour, and skull and teeth characteristics; molecularly, an analysis of Cyt b and COI gene sequences showed that these specimens form a monophyletic group with M. pluvialis with high posterior probability and bootstrap support values. Furthermore, the genetic distance between our specimens and M. pluvialis was greater than the minimum threshold for interspecific differentiation, indicating that they are phylogenetically close but have diverged. Based on the above assessment of morphological characteristics and molecular data analysis, these specimens were determined to represent a previously unidentified species, designated Murina lvchun Xin Mou & Song Li, sp. nov.

The new species, Harpiola isodon Kuo et al., 2006, was described from Taiwan, China. So far, no distribution of this species outside Taiwan has been reported. During two field investigations of small mammals in Guanyin Mountains Provincial Nature Reserve, Yuanyang, Yunnan, China, in April 2022 and May 2023, five individuals of Harpiola were collected in the mid-montane evergreen broad-leaved forest. Our morphological and molecular results reveal that these individuals from the Chinese mainland belong to Harpiola isodon, extending the occurrence of this species well beyond its known distributions in Taiwan, China and Vietnam.

This research documents a new record of Miniopterus magnater in the south-western region of China, a significant discovery given the limited diversity of the Miniopterus genus within the country. Only three species of Miniopterus occur in China: Miniopterus magnater, Miniopterus fuliginosus and Miniopterus pusillus. These species share a high degree of morphological similarity, particularly in their external characteristics. This underscores the necessity for the identification of additional distinguishing traits that can aid in the taxonomic differentiation of these closely-related species. During the 2023 field expedition to various nature reserves in Yunnan Province, China, we encountered specimens of the genus Miniopterus. Utilising a combination of morphological assessments and phylogenetic analyses, we identified six individuals as Miniopterus magnater. A review of the existing geographical distribution data revealed that this species is primarily found in central and southern regions of China, with no previous records from the south-western part of the country. Based on our findings, we present a novel record of Miniopterus magnater's distribution in the south-western region of China.

Mirostrellus joffrei Thomas, 1915 is the sole species within the genus Mirostrellus and its occurrence is notably rare in China. Despite the recent discovery of four previously unreported specimens from western Yunnan, China (Zaoteng River, 25.31°N, 98.80°E, altitude 1451 m) at the National Museum of Prague in the Czech Republic, comprehensive descriptions or detailed accounts of this species within China have yet to be published. During a field expedition within the Guanyinshan Provincial Nature Reserve, Yuanyang, Yunnan, China, a bat was captured using a mist net. The specimen was of moderate size, with its ventral surface presenting a golden-brown hue and its dorsal surface ranging from dark brown to obsidian. The forearm measured 37.31 mm, while the fifth digit was conspicuously diminished in comparison to the others. The snout was robust, featuring prominent buccal glands. It was characterised by a well-developed upper frame and a barely visible low sagittal crest. The upper canines possessed well-developed posterior cusps. The dentition comprised two upper and two lower premolars, with the first upper premolar being notably small and the lower molars being distinctly myotodont. These attributes correspond with the known traits of Mirostrellus joffrei. Phylogenetically, the sequence of this specimen clustered with that of M. joffrei, forming an independent clade. Through an examination of its morphological characteristics and a phylogenetic analysis of the mitochondrial cytochrome b (Cyt b), cytochrome oxidase subunit I (COI) and recombination activating gene 2 (RAG2) sequences, we identified the specimen as M. joffrei, thus confirming the presence of Joffre’s pipistrelle in China.

Fine-grained recognition of biological morphology remains constrained by subtle structural variation, limited anatomical information from single-view imaging, and inefficient integration of heterogeneous visual evidence. This study developed a multi-view deep learning framework for squirrel skull classification. Multi-view images are utilized as inputs to construct a deep learning framework that converts multi-source perceptual information into highly discriminative feature representations. The architecture integrates multi-view collaborative feature extraction, cross-view dynamic fusion, and fine-grained anatomical region enhancement within an end-to-end recognition pipeline. Specifically, multi-view collaborative encoding preserves view-specific cranial variation, dynamic fusion adaptively weights complementary features across perspectives, and regional enhancement strengthens diagnostic responses from key anatomical structures, thereby improving sensitivity to fine-grained morphological differences. Comprehensive evaluation on a newly constructed multi-view squirrel skull dataset demonstrated that the proposed method significantly outperformed multiple mainstream deep learning models, including ResNet, ViT, and Swin Transformer. The proposed model achieved an Accuracy of 92.37%, Precision of 92.05%, Recall of 91.62%, an F1-score of 91.83%, and a Macro-F1 of 91.24%. Ablation analyses confirmed the contribution of each module, with the full framework increasing accuracy by more than 8 percentage points relative to the single-view baseline. Analysis of input-view number showed a progressive improvement in classification performance as the number of views increased from one to five, confirming that multi-view anatomical information substantially strengthened structural representation. These results indicate that coordinated multi-view feature learning provides a robust and generalizable framework for fine-grained biological classification and morphology-oriented intelligent sensing.