FABDEM全球数据是基于2011-2015年间的哥白尼DEM，通过使用随机森林回归模型的机器学习从中去除建筑物和森林高度偏差。该数据是第一个同时去除建筑物和森林高度的全球数字高程模型数据集。该套算法方法将建筑区的平均绝对垂直误差从1.61米减少到1.12米，将森林区的平均绝对垂直误差从5.15米减少到2.88米。新的数字高程模型比现有的全球数字高程模型更准确，网格间距为1弧秒（赤道地区约为30米）。
\n在此我们提供中国新疆范围和中亚五国范围的1°×1°瓦片数据，不提供分国家、分省的镶嵌数据，瓦片数据沿用全球数据的组织结构和命名方式，大家可以按提供的1°×1°网格检索图，检索资金感兴趣的区域。
\n说明：覆盖新疆范围的共224幅，覆盖中亚五国的共549幅。\n中亚五国包含哈萨克斯坦、塔吉克斯坦、吉尔吉斯斯坦、乌兹别克斯坦和土库曼斯坦
", "ds_source": "FABDEM全球DEM地形数据，是在COPDEM30的基础上，使用随机森林回归模型的机器学习技术，从中移除了建筑物和树的高度偏差得到的。全球数据开放于布里斯托大学研究数据存储库。网址为：https://data.bris.ac.uk/data/dataset/s5hqmjcdj8yo2ibzi9b4ew3sn
", "ds_process_way": "制作FABDEM全球数据的工作流程包括三个阶段：\n(1)数据准备，包括处理预测数据和参考DEMs；\n(2)随机森林校正，分别对森林和建筑物进行移除；\n(3)后期处理，合并校正后的DEMs，填补不真实的坑洞，并应用平滑滤波器。
", "ds_quality": "建筑区的平均绝对垂直误差1.12米，森林区的平均绝对垂直误差2.88米。网格间距为1弧秒（赤道地区约为30米）。
", "ds_acq_start_time": null, "ds_acq_end_time": null, "ds_acq_place": "中国新疆，哈萨克斯坦、塔吉克斯坦、吉尔吉斯斯坦、乌兹别克斯坦、土库曼斯坦", "ds_acq_lon_east": null, "ds_acq_lat_south": null, "ds_acq_lon_west": null, "ds_acq_lat_north": null, "ds_acq_alt_low": null, "ds_acq_alt_high": null, "ds_share_type": "login-access", "ds_total_size": 21884229149, "ds_files_count": 2042, "ds_format": ".tiff格式", "ds_space_res": "30", "ds_time_res": "无", "ds_coordinate": "WGS84", "ds_projection": "WGS84", "ds_thumbnail": "611eb40a-a12a-4a79-a5fb-5b3ccddb6731.png", "ds_thumb_from": 0, "ds_ref_way": "", "paper_ref_way": "", "ds_ref_instruction": "Jeffrey Neal, Laurence Hawker (2023): FABDEM V1-2. https://doi.org/10.5523/bris.s5hqmjcdj8yo2ibzi9b4ew3sn", "ds_from_station": null, "organization_id": "a5877b42-96ea-4f13-af7e-246f355413d6", "doi_value": "", "subject_codes": [ "170.40" ], "quality_level": 1, "publish_time": "2025-07-11 14:32:08", "first_publish_time": null, "last_updated": "2025-07-27 15:06:04", "protected": false, "protected_to": null, "lang": "zh", "cstr": null, "license": null, "extra": null, "files_shape": [ { "name": "FABDEM_1x1_tiles_xj", "size": null, "is_dir": true }, { "name": "FABDEM_1×1_tiles_CA", "size": null, "is_dir": true }, { "name": "FABDEM_data_CA", "size": null, "is_dir": true }, { "name": "FABDEM_data_XJ", "size": null, "is_dir": true } ], "features": null, "data_level": 0, "i18n": { "en": { "title": "FABDEM 30-meter resolution terrain data from Xinjiang and Central Asia in China", "ds_abstract": "The FABDEM global data is based on Copernicus DEMs from 2011-2015, from which building and forest height deviations are removed through machine learning using random forest regression models. This data is the first global digital elevation model dataset to remove both building and forest heights. This algorithm reduces the average absolute vertical error in building areas from 1.61 meters to 1.12 meters, and reduces the average absolute vertical error in forest areas from 5.15 meters to 2.88 meters. The new digital elevation model is more accurate than existing global digital elevation models, with a grid spacing of 1 arcsecond (approximately 30 meters in the equatorial region).
\nHere, we provide 1°×1° tile data across Xinjiang in China and five Central Asian countries. We do not provide mosaic data by country or province. The tile data follows the organizational structure and naming method of global data. You can search for areas of financial interest according to the provided 1°×1° grid search map.
\nExplanation: A total of 224 pieces cover Xinjiang and a total of 549 pieces cover five Central Asian countries.\nThe five Central Asian countries include Kazakhstan, Tajikistan, Kyrgyzstan, Uzbekistan and Turkmenistan
", "ds_source": "FABDEM global DEM terrain data is obtained based on COPDEM30 and using machine learning technology of random forest regression models to remove height deviations of buildings and trees. Global data is open to the University of Bristol Research Data Repository. The website is: data.bris.ac.uk/data/dataset/s5hqmjcdj8yo2ibzi9b4ew3sn
", "ds_process_way": "The workflow for producing FABDEM global data includes three phases:\n(1)Data preparation, including processing forecast data and reference DEMs;\n(2)Random forest correction to remove forests and buildings separately;\n(3)Post-processing combines the corrected DEMs, fills in unrealistic potholes, and applies a smoothing filter.
", "ds_quality": "The average absolute vertical error in building areas is 1.12 meters, and the average absolute vertical error in forest areas is 2.88 meters. The grid spacing is 1 arcsecond (approximately 30 meters in the equatorial region).
", "ds_acq_place": "China Xinjiang, Kazakhstan, Tajikistan, Kyrgyzstan, Uzbekistan, Turkmenistan", "ds_ref_instruction": "Jeffrey Neal, Laurence Hawker (2023): FABDEM V1-2. https://doi.org/10.5523/bris.s5hqmjcdj8yo2ibzi9b4ew3sn", "ds_format": ".tiff format", "ds_projection": "WGS84", "ds_space_res": "30", "ds_time_res": "no" } }, "license_type": null, "doi_reg_from": "reg_local", "cstr_reg_from": "reg_local", "doi_not_reg_reason": null, "cstr_not_reg_reason": null, "is_paper_in_submitting": false, "ds_topic_tags": [ "数字高程模型", "DEM" ], "ds_subject_tags": [ "地图学" ], "ds_class_tags": [], "ds_locus_tags": [ "中国新疆", "吉尔吉斯斯坦", "哈萨克斯坦", "土库曼斯坦", "乌兹别克斯坦", "塔吉克斯坦" ], "ds_time_tags": [ 2011, 2012, 2013, 2014, 2015 ], "ds_contributors": [ "李锦" ], "ds_meta_authors": [ "李锦" ], "ds_managers": [ "李锦" ], "category": "DEM数字高程" }