An improved GM(1,3) model combining terrain factors and neural network error correction for urban land subsidence prediction
Urban land subsidence is a slow-density geology disaster caused by the withdrawal of groundwater or the application of water at the land surface. Developing a method of effectively monitoring, predicting and preventing land subsidence has become an urgent urban disaster issue and a great challenge....
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Taylor & Francis Group
2020-01-01
|
| Series: | Geomatics, Natural Hazards & Risk |
| Subjects: | |
| Online Access: | http://dx.doi.org/10.1080/19475705.2020.1716860 |
| _version_ | 1818822084219895808 |
|---|---|
| author | Qihang Zhou Qingwu Hu Mingyao Ai Chengli Xiong Hongfang Jin |
| author_facet | Qihang Zhou Qingwu Hu Mingyao Ai Chengli Xiong Hongfang Jin |
| author_sort | Qihang Zhou |
| collection | DOAJ |
| description | Urban land subsidence is a slow-density geology disaster caused by the withdrawal of groundwater or the application of water at the land surface. Developing a method of effectively monitoring, predicting and preventing land subsidence has become an urgent urban disaster issue and a great challenge. Traditional mathematical statistics prediction models lack physical meaning. Thus, in this paper, an improved third-order gray prediction GM(1, 3) model that combines terrain factors and neural network error corrections is proposed for urban land subsidence prediction. First, the correlation between land subsidence and terrain factors is verified through a correlation analysis. Second, a geospatial weight matrix based on the terrain factors is presented for auxiliary variable selection. Finally, a neural network is used to model the prediction errors to improve the model accuracy. Precision levelling data for Haiyan County, Zhejiang Province, China from 1999 to 2016 are used as the experimental data. The results show that the proposed method can achieve the same accuracy level as precision levelling. The average absolute error of three-phrase prediction is less than 10 mm, and the relative error can reach 0.4%. The proposed method can be expected to replace precision levelling for the prediction of long-term land subsidence and provide decision support for urban settlement disaster prevention. |
| first_indexed | 2024-12-18T23:18:27Z |
| format | Article |
| id | doaj.art-0f29d7a3127c4347a35dace0c1edd858 |
| institution | Directory Open Access Journal |
| issn | 1947-5705 1947-5713 |
| language | English |
| last_indexed | 2024-12-18T23:18:27Z |
| publishDate | 2020-01-01 |
| publisher | Taylor & Francis Group |
| record_format | Article |
| series | Geomatics, Natural Hazards & Risk |
| spelling | doaj.art-0f29d7a3127c4347a35dace0c1edd8582022-12-21T20:48:04ZengTaylor & Francis GroupGeomatics, Natural Hazards & Risk1947-57051947-57132020-01-0111121222910.1080/19475705.2020.17168601716860An improved GM(1,3) model combining terrain factors and neural network error correction for urban land subsidence predictionQihang Zhou0Qingwu Hu1Mingyao Ai2Chengli Xiong3Hongfang Jin4School of Remote Sensing and Information Engineering, Wuhan UniversitySchool of Remote Sensing and Information Engineering, Wuhan UniversitySchool of Remote Sensing and Information Engineering, Wuhan UniversityDepartment of Surveying and Mapping, The Second Surveying and Mapping Institute of Zhejiang ProvinceDepartment of Surveying and Mapping, The Second Surveying and Mapping Institute of Zhejiang ProvinceUrban land subsidence is a slow-density geology disaster caused by the withdrawal of groundwater or the application of water at the land surface. Developing a method of effectively monitoring, predicting and preventing land subsidence has become an urgent urban disaster issue and a great challenge. Traditional mathematical statistics prediction models lack physical meaning. Thus, in this paper, an improved third-order gray prediction GM(1, 3) model that combines terrain factors and neural network error corrections is proposed for urban land subsidence prediction. First, the correlation between land subsidence and terrain factors is verified through a correlation analysis. Second, a geospatial weight matrix based on the terrain factors is presented for auxiliary variable selection. Finally, a neural network is used to model the prediction errors to improve the model accuracy. Precision levelling data for Haiyan County, Zhejiang Province, China from 1999 to 2016 are used as the experimental data. The results show that the proposed method can achieve the same accuracy level as precision levelling. The average absolute error of three-phrase prediction is less than 10 mm, and the relative error can reach 0.4%. The proposed method can be expected to replace precision levelling for the prediction of long-term land subsidence and provide decision support for urban settlement disaster prevention.http://dx.doi.org/10.1080/19475705.2020.1716860urban land subsidence; prediction; gm(13); neural network; geospatial weight matrix |
| spellingShingle | Qihang Zhou Qingwu Hu Mingyao Ai Chengli Xiong Hongfang Jin An improved GM(1,3) model combining terrain factors and neural network error correction for urban land subsidence prediction Geomatics, Natural Hazards & Risk urban land subsidence; prediction; gm(1 3); neural network; geospatial weight matrix |
| title | An improved GM(1,3) model combining terrain factors and neural network error correction for urban land subsidence prediction |
| title_full | An improved GM(1,3) model combining terrain factors and neural network error correction for urban land subsidence prediction |
| title_fullStr | An improved GM(1,3) model combining terrain factors and neural network error correction for urban land subsidence prediction |
| title_full_unstemmed | An improved GM(1,3) model combining terrain factors and neural network error correction for urban land subsidence prediction |
| title_short | An improved GM(1,3) model combining terrain factors and neural network error correction for urban land subsidence prediction |
| title_sort | improved gm 1 3 model combining terrain factors and neural network error correction for urban land subsidence prediction |
| topic | urban land subsidence; prediction; gm(1 3); neural network; geospatial weight matrix |
| url | http://dx.doi.org/10.1080/19475705.2020.1716860 |
| work_keys_str_mv | AT qihangzhou animprovedgm13modelcombiningterrainfactorsandneuralnetworkerrorcorrectionforurbanlandsubsidenceprediction AT qingwuhu animprovedgm13modelcombiningterrainfactorsandneuralnetworkerrorcorrectionforurbanlandsubsidenceprediction AT mingyaoai animprovedgm13modelcombiningterrainfactorsandneuralnetworkerrorcorrectionforurbanlandsubsidenceprediction AT chenglixiong animprovedgm13modelcombiningterrainfactorsandneuralnetworkerrorcorrectionforurbanlandsubsidenceprediction AT hongfangjin animprovedgm13modelcombiningterrainfactorsandneuralnetworkerrorcorrectionforurbanlandsubsidenceprediction AT qihangzhou improvedgm13modelcombiningterrainfactorsandneuralnetworkerrorcorrectionforurbanlandsubsidenceprediction AT qingwuhu improvedgm13modelcombiningterrainfactorsandneuralnetworkerrorcorrectionforurbanlandsubsidenceprediction AT mingyaoai improvedgm13modelcombiningterrainfactorsandneuralnetworkerrorcorrectionforurbanlandsubsidenceprediction AT chenglixiong improvedgm13modelcombiningterrainfactorsandneuralnetworkerrorcorrectionforurbanlandsubsidenceprediction AT hongfangjin improvedgm13modelcombiningterrainfactorsandneuralnetworkerrorcorrectionforurbanlandsubsidenceprediction |