Qiang Li (李强)

Recent Projects

Actually, a diversity of influential machine learning approaches have already been applied across all the major sub-domains of Earth system science and are increasingly being integrated into operational schemes and being used to discover patterns, to improve our understanding and to evaluate comprehensive physical models (Reichstein et al., 2019). I have done several related projects in CUG including rainfall-runoff prediction and water quality prediction with long-short term memory (LSTM) in Haihe River Basin, Huangshui River Basin, and Yangtze River Basin between 2019 to 2022. Currently, I attempt to break through accuracy-interpretability dilemma of deep learning. I am concentrating on Physics-guided Deep Learning (PGDL), which combines physics laws and deep learning to gain superior performance and more interpretability.

Physics-informed Deep Learning



Project Page | Paper (This skatch is revised based on Jiang et al., 2020)

Considering the respective merits of physical approaches and AI models, the synergy of the two paradigms has recently been envisioned as an attractive research topic in the geoscience community (Gil et al., 2019).
Here, we consider three different hybrid strategies: a) pre-process: AI models learn and output the parameters of physical models, then physical models output the simulation; post-process: physical models are calibrated and output the initial simulated runoff, then AI models eliminates the gap between the simulation of physical models and the observation, and output the final simulation; c) structure-coupled: physical laws are fused into the structures of AI models, then the mixed models output simulation.
Different hybrid models are developed and compared to find out the most appropriate hybrid strategy for different application problems, such as hydrology prediction.

Rainfall-runoff Prediction



Project Page | Paper

In this study, both process-based model (EXPHYDRO in this study) and data-driven model (LSTM in this study) are utilized to predict runoff with CAMELS dataset. Results indicate that LSTM has better accuracy (mean NSE of 0.77), while mean NSE of EXPHYDRO is 0.71.
Furthermore, we carried out some interesting experiments: a) LSTM-EXPHYDRO: the LSTM learns and outputs the parameters of the EXPHYDR, then the EXPHYDRO outputs the simulated runoff; b) EXPHYDRO-LSTM: the EXPHYDRO is calibrated and outputs the initial simulated runoff, then the LSTM eliminates the gap between the EXPHYDRO simulation and observation, and outputs the final simulated runoff. Preliminary experiments shows that EXPHYDRO-LSTM has better performance than LSTM-EXPHYDRO.

Water Quality (WQ) Prediction with LSTM

  

Project Page | Paper

This study aims to compare and analyze the LSTM prediction performance of different water quality indicators (such as DO, COD, and TP) in different basins (such as Yangtze River Basin and Haihe River Basin) of China. Results show that LSTM performs better in Haihe River Basin than in Yangtze River Basin for its bigger temporal autocorrelation/cross-correlation of water quality indicators.

Publications

I'm interested in devleoping efficient deep learning models for Earth system sciences (e.g. hydrology, extreme meteorological event, water quality, and disaster prediction) and revealing their insight as well as interpretability.

Journal Papers:

1. Comparative Analysis of Water Quality Prediction Performance based on LSTM in Haihe River Basin, China
   Qiang Li, Yinqun Yang, Ling Yang, Yonggui Wang*
   Environmental Science and Pollution Research (IF=5.190) 2022 | paper | code

2. Comparative Analysis of Water Quality Prediction based on LSTM in Different River Basins, China
   Qiang Li, Yinqun Yang, Ling Yang, Yonggui Wang*
   Journal of Environmental Management (IF=8.910) 2022 (Under review) | code

3. The Architecture and Application of an Automatic Operational Model System for Basin Scale Water Environment Management and Design Making Supporting
   Yonggui Wang, Qiang Li, Wanshun Zhang, Shan Hu, Hong Peng*
   Journal of Environmental Management (IF=8.910) 2021 | paper | pdf

4. Pollutants Removal Efficiency Assessment of Constructed Subsurface Flow Wetlands in Lakes with Numerical Models
   Yonggui Wang, Qiang Li, Wanshun Zhang, Shaofei Wang, Hong Peng*
   Journal of Hydrology (IF=6.708) 2021 |paper | pdf

5. Early Warning of Heavy Metal Pollution after Tailing Pond Failure Accident
   Yonggui Wang, Yinqun Yang*, Qiang Li, Yaxin Zhang, Xiaolong Chen
   Journal of Earth Science (IF=2.433) 2022 | paper | pdf

Invention Patents:

1. Water Quality Prediction Method of River Sudden Water Pollution Accident based on Improved LSTM-SEQ2SEQ Model
   Qiang Li, Yonggui Wang, Yaxin Zhang
   China Invention Patent 2022 (Authorized) | certificate
   Included in Catalogue of Advanced and Applicable Technologies for Ecological and Environmental Protection of Wuhan, 2022

2. Water Quality Prediction Method based on Physical-restrained and Process-Guided Deep Learning Model
   Qiang Li, Yonggui Wang
   China Invention Patent 2022 (Authorized) | certificate

3. A Method and System for Collecting Hydrologic and Water Quality Data based on Web Crawler
   Tianyi Xie, Yonggui Wang, Qiang Li
   China Invention Patent 2020 (Under substantive review) | certificate

Software Copyright:

1. Intelligent Service System for Water Quality Monitoring and Early Warning
   Qiang Li, Yonggui Wang, Tianyi Xie
   China Software Copyright 2020 (Authorized) | certificate

2. Service System for Hydrological and Water Quality Data
   Tianyi Xie, Yonggui Wang, Qiang Li
   China Software Copyright 2020 (Authorized) | certificate

Awards

• 2022, National Scholarship for Graduate Students, Ministry of Education, China

• 2022, The First-class Academic Scholarship for Graduate Students, China University of Geosciences

• 2021, National Scholarship for Graduate Students, Ministry of Education, China

• 2021, The First-class Academic Scholarship for Graduate Students, China University of Geosciences

• 2021, The Second prize in College Students Patent Invention Contest, China University of Geosciences

• 2020, The First-class Academic Scholarship for Graduate Students, China University of Geosciences

• 2017, Scholarship for Geology National Science Base Class, China University of Geosciences

• 2016, Scholarship for Geology National Science Base Class, China University of Geosciences

• 2016, Admission Scholarship for New Undergraduate Students, China University of Geosciences