Journal of Geo-information Science

Improving Urban Digital Elevation Models Based on Iinterpretable Random Forest Method Considering Spatial Heterogeneity

LIU Yan, SUN Yanning, CHEN Chuanfa, LIU Panpan, LIU Yating

Journal of Geo-information Science, 2024, 26(4): 978-988. DOI: 10.12082/dqxxkx.2024.230590

Fig. 4 SHAP feature analysis in training area

Other figure/table from this article

Fig. 1 Flowchart of the proposed method
Tab. 1 Texture feature description
Fig. 2 Study area GDEM and remote sensing image
Tab. 2 Data source and data type
Tab. 3 DBI value under different number of clusters
Fig. 3 Spatial partition result
Tab. 4 Statistical information of each subarea
Fig. 5 Relationship between the number of features in the training area and the accuracy of the model
Fig. 6 Precision comparison of GDEM
Tab. 5 Accuracy analysis after COPDEM30 correction in test area (m)
Tab. 6 Correlation coefficient of the characteristic variables of training and test data
Fig. 7 Error histogram of COPDEM30 before and after correction
Fig. 8 Comparison of different GDEM in the test area