YDYunyuan Deng

Yunyuan Deng

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Yunyuan Deng has a total of 9 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Multiple Spatial and Temporal Scales Evaluation of Eight Satellite Precipitation Products in a Mountainous Catchment of South China

    by Binbin Guo, Tingbao Xu, Qin Yang, Jing Zhang, Zhong Dai, Yunyuan Deng, Jun Zou

    Satellite precipitation products (SPPs) have emerged as an important information source of precipitation with high spatio-temporal resolutions, with great potential to improve catchment water resource management and hydrologic modelling, especially in data-sparse regions. As an indirect precipitation measurement, satellite-derived precipitation accuracy is of major concern. There have been numerous evaluation/validation studies worldwide. However, a convincing systematic evaluation/validation of satellite precipitation remains unrealized. In particular, there are still only a limited number of hydrologic evaluations/validations with a long temporal period. Here we present a systematic evaluation of eight popular SPPs (CHIRPS, CMORPH, GPCP, GPM, GSMaP, MSWEP, PERSIANN, and SM2RAIN). The evaluation area used, using daily data from 2007 to 2020, is the Xiangjiang River basin, a mountainous catchment with a humid sub-tropical monsoon climate situated in south China. The evaluation was conducted at various spatial scales (both grid-gauge scale and watershed scale) and temporal scales (annual and seasonal scales). The evaluation paid particular attention to precipitation intensity and especially its impact on hydrologic modelling. In the evaluation of the results, the overall statistical metrics show that GSMaP and MSWEP rank as the two best-performing SPPs, with KGEGrid ≥ 0.48 and KGEWatershed ≥ 0.67, while CHIRPS and SM2RAIN were the two worst-performing SPPs with KGEGrid ≤ 0.25 and KGEWatershed ≤ 0.42. GSMaP gave the closest agreement with the observations. The GSMaP-driven model also was superior in depicting the rainfall-runoff relationship compared to the hydrologic models driven by other SPPs. This study further demonstrated that satellite remote sensing still has difficulty accurately estimating precipitation over a mountainous region. This study provides helpful information to optimize the generation of algorithms for satellite precipitation products, and valuable guidance for local communities to select suitable alternative precipitation datasets.


  • Open Access

    Towards Classification of Architectural Styles of Chinese Traditional Settlements Using Deep Learning: A Dataset, a New Framework, and Its Interpretability

    by Qing Han, Chao Yin, Yunyuan Deng, Peilin Liu

    The classification of architectural style for Chinese traditional settlements (CTSs) has become a crucial task for developing and preserving settlements. Traditionally, the classification of CTSs primarily relies on manual work, which is inefficient and time consuming. Inspired by the tremendous success of deep learning (DL), some recent studies attempted to apply DL networks such as convolution neural networks (CNNs) to achieve automated classification of the architecture styles. However, these studies suffer overfitting problems of the CNNs, leading to inferior classification performance. Moreover, most of the studies apply the CNNs as a black box providing limited interpretability. To address these limitations, a new DL classification framework is proposed in this study to overcome the overfitting problem by transfer learning and learning-based data augmentation technique (i.e., AutoAugment). Furthermore, we also employ class activation map (CAM) visualization technique to help understand how the CNN classifiers work to abstract patterns from the input. Specifically, due to a lack of architectural style datasets for the CTSs, a new annotated dataset is first established with six representative classes. Second, several representative CNNs are leveraged to benchmark the new dataset. Third, to address the overfitting problem of the CNNs, a new DL framework is proposed which combines transfer learning and AutoAugment to improve the classification performance. Extensive experiments are conducted on the new dataset to demonstrate the effectiveness of our framework. The proposed framework achieves much better performance than baselines, greatly mitigating the overfitting problem. Additionally, the CAM visualization technique is harnessed to explain what and how the CNN classifiers implicitly learn for recognizing a specified architectural style.