本研究基于谷歌地球引擎（GEE, Google Earth Engine），调用2020年5-9月10m分辨率的哨兵2号遥感影像（Sentinel-2），将地物光谱特征与随机森林分类方法相结合，加入植被水分指数（Normalized Difference Moisture Index,NDMI），对研究区灌溉和雨养草地进行分类，制作了巴音河中下游灌溉草地数据集。该数据集可为地表过程模拟、水资源规划等提供基础。
", "ds_source": "基于Sentinel-2生产所得
", "ds_process_way": "首先从GEE中调用2020年5月至2020年9月（植被生长期）的所有可用图像，并根据图像相关的“多云像素百分比”进行排序，选择云百分比最低的图像并进行裁剪，得到巴音河中下游这一时期的最佳无云（10 m：红、绿、蓝、NIR）影像。\n
其次，基于Sentinel-2影像农业波段（B11, B8, B2）和短波红外波段（B12, B8A, B4），构建NDVI、GCVI、LSWI、MNDWI、NDBI等植被指数和地形特征指数（坡度），再结合野外采样点进行特征选取，并选用随机森林分类方法提取林地、耕地、草地、建筑、水体、裸地、湿地7种土地利用类型。通过混淆矩阵计算验证分类精度。\n
而后，基于Sentinel-2影像计算水分指数NDMI，结合随机森林分类方法再次对草地进行特征选取，提取出灌溉草地。\n
最后，通过混淆矩阵对结果进行精度验证，并制作灌溉草地分布数据集。", "ds_quality": "
本研究利用混淆矩阵对巴音河中下游林地、耕地、草地、建筑、水体、裸地、湿地7种土地利用类型分类结果进行精度评价，其总体精度达到96%，kappa系数为0.93，分类精度较好。其中，用户精度最高的是裸地和耕地（0.97），最低的是森林（0.87），其余的土地利用类型都在0.9以上，说明该分类结果中裸地和耕地的提取效果最佳，可信度最高，森林的可信度略低。生产者精度最高的是草地（0.98），最低的是城镇用地（0.64）。综合用户精度以及生产者精度来看，裸地、耕地、草地内部纹理特征均匀度高、复杂度低，与周围其他地物相差明显，分类精度较高。总体来看，各地物类型区分性好，分类精度高。\n
基于第一次分类结果，提取出草地，加入植被水分指数NDMI，再次采用随机森林分类方法进行灌溉草地和雨养草地的识别和提取，分类结果总体精度为96%，Kappa系数为0.92。其中，雨养草地的用户精度和生产者精度分别为0.95和0.98，灌溉草地的用户精度为99%，制图精度为0.78，说明了雨养草地和灌溉草地的识别度都比较高。总体来说，分类效果较好，分类结果通过了一致性检验结果，可信度高，表明该方法在灌溉草地和雨养草地分类提取方面具有较好的效果。", "ds_acq_start_time": "2020-05-01 00:00:00", "ds_acq_end_time": "2020-09-30 00:00:00", "ds_acq_place": "青海省德令哈巴音河中下游", "ds_acq_lon_east": 98.14500000000001, "ds_acq_lat_south": 36.99388888888889, "ds_acq_lon_west": 96.97, "ds_acq_lat_north": 37.573888888888895, "ds_acq_alt_low": 2765.0, "ds_acq_alt_high": 5255.0, "ds_share_type": "open-access", "ds_total_size": 1196721, "ds_files_count": 6, "ds_format": "TIF", "ds_space_res": "10m", "ds_time_res": "年", "ds_coordinate": "WGS84", "ds_projection": "UTM Zone_47N", "ds_thumbnail": "e97e9b84-192d-4b30-9a49-282f582c2c7c.jpg", "ds_thumb_from": 0, "ds_ref_way": "", "paper_ref_way": "", "ds_ref_instruction": "", "ds_from_station": null, "organization_id": "52b7b79b-860c-49a5-9083-9a70cf8bed5a", "ds_serv_man": "敏玉芳", "ds_serv_phone": "0931-4967596", "ds_serv_mail": "ncdc@lzb.ac.cn", "doi_value": "", "subject_codes": [ "170.45" ], "quality_level": 3, "publish_time": "2023-05-19 17:26:26", "last_updated": "2025-04-24 12:01:49", "protected": false, "protected_to": null, "lang": "zh", "cstr": "11738.11.NCDC.NIEER.DB2849.2023", "i18n": { "en": { "title": "Spatial distribution dataset of irrigated grasslands in the middle and lower reaches of the Bayin River (2020)", "ds_format": "TIF", "ds_source": "
Based on Sentinel-2 production income
", "ds_quality": "This study used a confusion matrix to evaluate the accuracy of the classification results of 7 land use types in the middle and lower reaches of the Bayin River, including forest land, cultivated land, grassland, buildings, water bodies, bare land, and wetlands. The overall accuracy reached 96%, with a Kappa coefficient of 0.93, indicating good classification accuracy. Among them, bare land and cultivated land have the highest user accuracy (0.97), forest has the lowest accuracy (0.87), and the other land use types are all above 0.9. This indicates that the extraction effect of bare land and cultivated land is the best in this classification result, with the highest credibility, while the credibility of forest is slightly lower. The highest producer accuracy is in grassland (0.98), and the lowest is in urban land (0.64). From the perspective of user accuracy and producer accuracy, the internal texture features of bare land, cultivated land, and grassland have high uniformity and low complexity, which are significantly different from other surrounding land features and have high classification accuracy. Overall, the distinction between different types of objects is good and the classification accuracy is high.\n
Based on the first classification result, the grassland was extracted, and the vegetation moisture index (NDMI) was added. Then, a random forest classification method was used to identify and extract irrigated grassland and rainfed grassland. The overall accuracy of the classification result was 96%, and the Kappa coefficient was 0.92. Among them, the user accuracy and producer accuracy of rain fed grassland are 0.95 and 0.98, respectively, while the user accuracy of irrigated grassland is 99%, and the mapping accuracy is 0.78, indicating that the recognition accuracy of rain fed grassland and irrigated grassland is relatively high. Overall, the classification effect is good, and the classification results have passed the consistency test with high reliability, indicating that this method has good performance in the classification and extraction of irrigated grasslands and rainfed grasslands.", "ds_ref_way": "", "ds_abstract": "
This study is based on the Google Earth Engine (GEE) and uses Sentinel-2 remote sensing images with a resolution of 10 meters from May to September 2020. The spectral features of the terrain are combined with random forest classification methods, and the Normalized Difference Moisture Index (NDMI) is added to classify the irrigated and rainfed grasslands in the study area. A dataset of irrigated grasslands in the middle and lower reaches of the Bayin River is produced. This dataset can provide a foundation for surface process simulation, water resource planning, and more
", "ds_time_res": "年", "ds_acq_place": "Middle and Lower Reaches of Bayin River in Delingha, Qinghai Province", "ds_space_res": "10m", "ds_projection": "UTM Zone_47N", "ds_process_way": "First, call all available images from May 2020 to September 2020 (vegetation growth period) from GEE, sort them according to the \"cloudy pixel percentage\" related to images, select the image with the lowest cloud percentage and cut it to get the best cloud free (10 m: red, green, blue, NIR) image in the middle and lower reaches of Bayin River during this period.\n
Secondly, based on Sentinel-2 images in the agricultural bands (B11, B8, B2) and shortwave infrared bands (B12, B8A, B4), vegetation indices such as NDVI, GCVI, LSWI, MNDWI, and NDBI were constructed, along with terrain feature indices (slope). Then, feature selection was carried out using field sampling points, and random forest classification methods were used to extract seven land use types: forest land, cultivated land, grassland, buildings, water bodies, bare land, and wetlands. Verify classification accuracy through confusion matrix calculation.\n
Then, based on Sentinel-2 images, the water index NDMI was calculated, and a random forest classification method was used to select features of the grassland again, extracting irrigated grassland.\n
Finally, the accuracy of the results was verified through the confusion matrix, and a dataset for the distribution of irrigated grasslands was created.", "ds_ref_instruction": "" } }, "submit_center_id": "ncdc", "data_level": 0, "license_type": "CC BY 4.0", "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": [ "灌溉草地空间分布数据集", "植被水分指数", "巴音河中下游", "谷歌地球引擎" ], "ds_subject_tags": [ "地理学" ], "ds_class_tags": [], "ds_locus_tags": [ "青海省", "德令哈", "巴音河中下游" ], "ds_time_tags": [ 2020 ], "ds_contributors": [ { "true_name": "金彦香", "email": "jinyx13@lzu.edu.cn", "work_for": "青海师范大学地理科学学院", "country": "中国" }, { "true_name": "毛旭锋", "email": "maoxufeng@yeah.net", "work_for": "青海师范大学地理科学学院", "country": "中国" }, { "true_name": "郑丽", "email": "lzheng2022@126.com", "work_for": "青海师范大学", "country": "中国" }, { "true_name": "秦艳红", "email": "qinyh1128@163.com", "work_for": "青海师范大学", "country": "中国" }, { "true_name": "金鑫", "email": "jinx13@lzu.edu.cn", "work_for": "青海师范大学地理科学学院", "country": "中国" }, { "true_name": "杜凯", "email": "20211028@qhnu.edu.cn", "work_for": "青海师范大学", "country": "中国" } ], "ds_meta_authors": [ { "true_name": "郑丽", "email": "lzheng2022@126.com", "work_for": "青海师范大学", "country": "中国" }, { "true_name": "金鑫", "email": "jinx13@lzu.edu.cn", "work_for": "青海师范大学地理科学学院", "country": "中国" } ], "ds_managers": [ { "true_name": "金彦香", "email": "jinyx13@lzu.edu.cn", "work_for": "青海师范大学地理科学学院", "country": "中国" }, { "true_name": "毛旭锋", "email": "maoxufeng@yeah.net", "work_for": "青海师范大学地理科学学院", "country": "中国" }, { "true_name": "郑丽", "email": "lzheng2022@126.com", "work_for": "青海师范大学", "country": "中国" }, { "true_name": "秦艳红", "email": "qinyh1128@163.com", "work_for": "青海师范大学", "country": "中国" }, { "true_name": "金鑫", "email": "jinx13@lzu.edu.cn", "work_for": "青海师范大学地理科学学院", "country": "中国" }, { "true_name": "杜凯", "email": "20211028@qhnu.edu.cn", "work_for": "青海师范大学", "country": "中国" } ], "category": "遥感及产品" }