Zuxiang Situ
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Zuxiang Situ has a total of 32 co-authors. Here are the top co-authors by number of shared publications:
- Qianqian Zhou
School of Civil and Transportation Engineering, Guangdong University of Technology, Guangzhou 510006, China
6 shared publications Shuai Teng
6 shared publicationsGongfa Chen
School of Civil and Transportation Engineering, Guangdong University of Technology, Guangzhou, China
4 shared publicationsJianliang Zhang
3 shared publicationsQianqian Zhou
3 shared publicationsWanen Feng
3 shared publications
Latest Publications
- Open Access
- 19 November 2026
- Published by IWA Publishing in Journal of Hydroinformatics
Evaluation and quantile mapping-based correction of TIGGE multi-model precipitation forecasts in mountainous watersheds
Accurate precipitation forecasts are essential for characterizing and predicting extreme rainfall events, particularly in mountainous watersheds where precipitation exhibits strong spatial and temporal variability. This study evaluated 6 h accumulated precipitation forecasts from three TIGGE (THORPEX Interactive Grand Global Ensemble) systems, CMA, ECMWF, and NCEP, during the April-October seasons of 2017–2022 over the Jinbao River Basin, Guangxi, China. Forecast performance was assessed using categorical and continuous verification metrics. The ECMWF ensemble maximum precipitation forecasts showed the highest skill for heavy rainfall and rainstorm events but exhibited systematic overestimation. To mitigate this bias, a segmented quantile mapping (QM) method was applied to post-process the ECMWF ensemble maximum forecasts. After correction, overprediction and false alarms were substantially reduced, while the TS (Threat Score) for rainstorm events remained stable at approximately 0.2 for lead times of 1–4 days. These results indicate that QM improves the reliability of deterministic extreme precipitation forecasts from the ECMWF ensemble maximum. Because only six years of observations were available, however, the study cannot fully characterize precipitation return periods of 50 years or longer; results beyond this level should be interpreted with caution and evaluated against longer observational records.
- 19 November 2026
- Published by IWA Publishing in Journal of Hydroinformatics
Open AccessEvaluation and quantile mapping-based correction of TIGGE multi-model precipitation forecasts in mountainous watersheds
Accurate precipitation forecasts are essential for characterizing and predicting extreme rainfall events, particularly in mountainous watersheds where precipitation exhibits strong spatial and temporal variability. This study evaluated 6 h accumulated precipitation forecasts from three TIGGE (THORPEX Interactive Grand Global Ensemble) systems, CMA, ECMWF, and NCEP, during the April-October seasons of 2017–2022 over the Jinbao River Basin, Guangxi, China. Forecast performance was assessed using categorical and continuous verification metrics. The ECMWF ensemble maximum precipitation forecasts showed the highest skill for heavy rainfall and rainstorm events but exhibited systematic overestimation. To mitigate this bias, a segmented quantile mapping (QM) method was applied to post-process the ECMWF ensemble maximum forecasts. After correction, overprediction and false alarms were substantially reduced, while the TS (Threat Score) for rainstorm events remained stable at approximately 0.2 for lead times of 1–4 days. These results indicate that QM improves the reliability of deterministic extreme precipitation forecasts from the ECMWF ensemble maximum. Because only six years of observations were available, however, the study cannot fully characterize precipitation return periods of 50 years or longer; results beyond this level should be interpreted with caution and evaluated against longer observational records.
- Top Cited Papers
- 26 January 2024
- Published by Elsevier in Journal of Hydrology
- Vol. 630
Improving urban flood prediction using LSTM-DeepLabv3+ and Bayesian optimization with spatiotemporal feature fusion
- 26 January 2024
- Published by Elsevier in Journal of Hydrology
- Vol. 630
Top Cited PapersImproving urban flood prediction using LSTM-DeepLabv3+ and Bayesian optimization with spatiotemporal feature fusion
- Open Access
- 5 May 2023
- Published by Copernicus GmbH in Hydrology and Earth System Sciences
- Vol. 27 (9) , 1791-1808
A deep-learning-technique-based data-driven model for accurate and rapid flood predictions in temporal and spatial dimensions
An accurate and rapid urban flood prediction model is essential to support decision-making for flood management. This study developed a deep-learning-technique-based data-driven model for flood predictions in both temporal and spatial dimensions, based on an integration of long short-term memory (LSTM) network, Bayesian optimization, and transfer learning techniques. A case study in northern China was applied to test the model performance, and the results clearly showed that the model can accurately predict the maximum water depths and flood time series for various hyetograph inputs, with substantial improvements in the computation time. The model predicted flood maps 19 585 times faster than the physically based hydrodynamic model and achieved a mean relative error of 9.5 %. For retrieving the spatial patterns of water depths, the degree of similarity of the flood maps was very high. In a best case scenario, the difference between the ground truth and model prediction was only 0.76 %, and the spatial distributions of inundated paths and areas were almost identical. With the adoption of transfer learning, the proposed model was well applied to a new case study and showed robust compatibility and generalization ability. Our model was further compared with two baseline prediction algorithms (artificial neural network and convolutional neural network) to validate the model superiority. The proposed model can potentially replace and/or complement the conventional hydrodynamic model for urban flood assessment and management, particularly in applications of real-time control, optimization, and emergency design and planning.
- 5 May 2023
- Published by Copernicus GmbH in Hydrology and Earth System Sciences
- Vol. 27 (9) , 1791-1808
Open AccessA deep-learning-technique-based data-driven model for accurate and rapid flood predictions in temporal and spatial dimensions
An accurate and rapid urban flood prediction model is essential to support decision-making for flood management. This study developed a deep-learning-technique-based data-driven model for flood predictions in both temporal and spatial dimensions, based on an integration of long short-term memory (LSTM) network, Bayesian optimization, and transfer learning techniques. A case study in northern China was applied to test the model performance, and the results clearly showed that the model can accurately predict the maximum water depths and flood time series for various hyetograph inputs, with substantial improvements in the computation time. The model predicted flood maps 19 585 times faster than the physically based hydrodynamic model and achieved a mean relative error of 9.5 %. For retrieving the spatial patterns of water depths, the degree of similarity of the flood maps was very high. In a best case scenario, the difference between the ground truth and model prediction was only 0.76 %, and the spatial distributions of inundated paths and areas were almost identical. With the adoption of transfer learning, the proposed model was well applied to a new case study and showed robust compatibility and generalization ability. Our model was further compared with two baseline prediction algorithms (artificial neural network and convolutional neural network) to validate the model superiority. The proposed model can potentially replace and/or complement the conventional hydrodynamic model for urban flood assessment and management, particularly in applications of real-time control, optimization, and emergency design and planning.
- Open Access
- 1 March 2022
- Published by IWA Publishing in Journal of Hydroinformatics
- Vol. 24 (2) , 406-419
Comparison of classic object-detection techniques for automated sewer defect detection
Sewer systems play a key role in cities to ensure public assets and safety. Timely detection of defects can effectively alleviate system deterioration. Conventional manual inspection is labor-intensive, error-prone and expensive. Object detection is a powerful deep learning technique that can complement and/or replace conventional inspection, especially in complex environments. This study compares two classic object-detection methods, namely faster region-based convolutional neural network (R-CNN) and You Only Look Once (YOLO), for the detection and localization of five types of sewer defects. Model performances are evaluated based on their detection accuracy and processing speed under parameterization impacts of dataset size and training parameters. Results show that faster R-CNN achieved higher prediction accuracy. Training dataset size and maximum number of epochs (MaxE) had dominant impacts on model performances of faster R-CNN and YOLO, respectively. The processing speed increased along with the increasing training data for faster R-CNN, but did not vary significantly for YOLO. The models' abilities to detect disjoint and residential wall were highest, whereas crack and tree root were more difficult to detect. The results help to better understand the strengths and weaknesses of the classic methods and provide a useful user guidance for practical applications in automated sewer defect detection.
- 1 March 2022
- Published by IWA Publishing in Journal of Hydroinformatics
- Vol. 24 (2) , 406-419
Open AccessComparison of classic object-detection techniques for automated sewer defect detection
Sewer systems play a key role in cities to ensure public assets and safety. Timely detection of defects can effectively alleviate system deterioration. Conventional manual inspection is labor-intensive, error-prone and expensive. Object detection is a powerful deep learning technique that can complement and/or replace conventional inspection, especially in complex environments. This study compares two classic object-detection methods, namely faster region-based convolutional neural network (R-CNN) and You Only Look Once (YOLO), for the detection and localization of five types of sewer defects. Model performances are evaluated based on their detection accuracy and processing speed under parameterization impacts of dataset size and training parameters. Results show that faster R-CNN achieved higher prediction accuracy. Training dataset size and maximum number of epochs (MaxE) had dominant impacts on model performances of faster R-CNN and YOLO, respectively. The processing speed increased along with the increasing training data for faster R-CNN, but did not vary significantly for YOLO. The models' abilities to detect disjoint and residential wall were highest, whereas crack and tree root were more difficult to detect. The results help to better understand the strengths and weaknesses of the classic methods and provide a useful user guidance for practical applications in automated sewer defect detection.
- 1 July 2021
- Published by American Society of Civil Engineers (ASCE) in Journal of Water Resources Planning and Management
- Vol. 147 (7) , 04021036
Convolutional Neural Networks–Based Model for Automated Sewer Defects Detection and Classification
Automated detection and classification of sewer defects can complement the conventional labor-intensive sewer inspection process by providing an essential tool to classify sewer defects in a more efficient, accurate, and consistent way. This paper presents a convolutional neural networks (CNNs)–based model to automatically detect and classify six most commonly observed sewer defects (i.e., cracks, disjoints, obstacles, residential walls, tree roots, and normal categories) obtained from multisource CCTV images under various circumstances. Data augmentation techniques (including geometric and color transformations) are applied to enhance the model performance. The proposed CNN model is further compared with a state-of-the-art solution (retraining the SqueezeNet using defect images) by adopting transfer learning technique. An average prediction accuracy of 90% is achieved, indicating that the investigated defects can be well recognized by the model without any expert knowledge of sewer detection. There is a higher degree of confidence in predicting tree roots and disjoints, followed by residential walls and cracks. Results show that the prediction accuracy has increased by 15% thanks to data augmentation. Despite the transferred SqueezeNet model achieved a higher accuracy (95%), it cost almost 13 times the computation time of the CNN model. The study demonstrates the feasibility of the deep learning technology in the automated classification of sewer defects and advances the knowledge in the research field.
- 1 July 2021
- Published by American Society of Civil Engineers (ASCE) in Journal of Water Resources Planning and Management
- Vol. 147 (7) , 04021036
Convolutional Neural Networks–Based Model for Automated Sewer Defects Detection and Classification
Automated detection and classification of sewer defects can complement the conventional labor-intensive sewer inspection process by providing an essential tool to classify sewer defects in a more efficient, accurate, and consistent way. This paper presents a convolutional neural networks (CNNs)–based model to automatically detect and classify six most commonly observed sewer defects (i.e., cracks, disjoints, obstacles, residential walls, tree roots, and normal categories) obtained from multisource CCTV images under various circumstances. Data augmentation techniques (including geometric and color transformations) are applied to enhance the model performance. The proposed CNN model is further compared with a state-of-the-art solution (retraining the SqueezeNet using defect images) by adopting transfer learning technique. An average prediction accuracy of 90% is achieved, indicating that the investigated defects can be well recognized by the model without any expert knowledge of sewer detection. There is a higher degree of confidence in predicting tree roots and disjoints, followed by residential walls and cracks. Results show that the prediction accuracy has increased by 15% thanks to data augmentation. Despite the transferred SqueezeNet model achieved a higher accuracy (95%), it cost almost 13 times the computation time of the CNN model. The study demonstrates the feasibility of the deep learning technology in the automated classification of sewer defects and advances the knowledge in the research field.