本数据集旨在探讨气候变化和人类活动对中国干旱区土壤多功能性的影响。通过从 Web of Science 和 中国知网 检索 1900 年至 2023 年 5 月的相关文献，筛选出 841 项野外研究，涵盖 18,189 组观测数据。数据包括土壤功能相关指标、气候数据（如年均气温、降水量、氮沉降等）、生态系统类型、土壤深度和处理强度等信息。利用 H2O AutoML 包构建预测模型，分析未来（本世纪末）气候变化（温度、降水、氮沉降）和人类活动（如耕作、放牧、退耕还林等）对土壤多功能性的影响。数据质量通过 10 折交叉验证和多个评价指标（如 R² 和 RMSE）确认，结果表明所构建的模型能有效预测未来土壤多功能性变化，具有较高的可靠性和精度。", "ds_source": "
1.文献数据收集: 我们通过 Web of Science 和 CNKI 检索 1900–2023 年干旱区全球变化控制实验，筛选包含对照—处理对比、明确处理强度与持续时间、可获得至少一项土壤指标的研究，共纳入 841 项野外实验、18189 组观测数据。\n
2. 变量提取: 从文献中提取土壤功能指标均值、重复数及环境因子（纬度、经度、海拔、生态系统类型、MAT、MAP、土壤深度、处理强度等）。数据以表格直接读取，图形数据使用 WebPlotDigitizer 数字化；缺失气候变量利用 WorldClim v2.1 补充。\n
3. 空间数据: 基于陆地生态系统分布数据获取生态系统类型，利用“getData” 函数提取海拔。\n
4.未来气候数据: 采用 CMIP6 ACCESS 模型获取中国干旱区历史（1970–2000 年）与 SSP1-2.6 情景下 21 世纪末（2081–2100 年）温度、降水和氮沉降的变化量。", "ds_process_way": "
1.机器学习预测模型构建: 使用 H2O AutoML，以 lnRRSMF 为响应变量，MAT、MAP、海拔、生态系统类型、土壤深度与处理强度/持续时间为预测因子，选择最优模型用于预测。\n
2. 气候变化预测数据集构建: 整合干旱区未来温度、降水、氮沉降变化值及环境因子，形成用于预测的气候变化输入数据。\n
3. 人类活动处理强度设定: 放牧强度标准化为 1 头/公顷；耕作、围栏、退耕还林等生态工程处理持续时间统一设定到 21 世纪末（75 年）。\n
4. 土壤剖面设置: 构建 0–30 cm（3 cm 间隔）共 10 层的土壤数据，以覆盖文献中 80% 的采样深度，提高预测稳健性。\n
5. 模型预测与结果输出: 使用 h2o.predict 对每层土壤数据进行预测，结果取平均后转换为变化百分比，并生成栅格格式用于制图。", "ds_quality": "
模型验证与选择:采用 10 折交叉验证与固定随机种子，基于 RMSE 和 R² 选择最优模型。总体而言，不同全球变化因子的模型均具有可接受至较高的预测能力，可用于评估未来情景下土壤多功能性的变化。", "ds_acq_start_time": "2000-01-01 00:00:00", "ds_acq_end_time": "2021-12-31 00:00:00", "ds_acq_place": "中国干旱区", "ds_acq_lon_east": 135.05, "ds_acq_lat_south": 18.17, "ds_acq_lon_west": 73.46000000000001, "ds_acq_lat_north": 53.54, "ds_acq_alt_low": null, "ds_acq_alt_high": null, "ds_share_type": "apply-access", "ds_total_size": 7174994, "ds_files_count": 2, "ds_format": "tif", "ds_space_res": "5000m", "ds_time_res": "年", "ds_coordinate": "WGS84", "ds_projection": "", "ds_thumbnail": "1a00eaa4-eb6b-4cb9-95c2-e326837b4964.png", "ds_thumb_from": 2, "ds_ref_way": "", "paper_ref_way": "", "ds_ref_instruction": "", "ds_from_station": null, "organization_id": "6d0aa454-9b64-4be5-b0cd-4cc796e6aea0", "ds_serv_man": "李红星", "ds_serv_phone": "0931-4967592", "ds_serv_mail": "ncdc@lzb.ac.cn", "doi_value": "", "subject_codes": [ "170.1535" ], "quality_level": 3, "publish_time": "2026-03-13 18:44:29", "last_updated": "2026-03-13 18:44:29", "protected": false, "protected_to": "2028-03-01 00:00:00", "lang": "zh", "cstr": "11738.11.NCDC.DESERTIFICATION.DB7160.2026", "i18n": { "en": { "title": "A dataset on the response of soil multifunctionality in arid regions of China to climate change and human activities at the end of the 21st century", "ds_format": "tif", "ds_source": "
1. Literature data collection: We searched for global change control experiments in arid regions from 1900 to 2023 through Web of Science and CNKI, and selected studies that included control treatment comparisons, clarified treatment intensity and duration, and obtained at least one soil indicator. A total of 841 field experiments and 18189 sets of observation data were included.\n
2. Variable extraction: Extract the mean, replicates, and environmental factors (latitude, longitude, altitude, ecosystem type, MAT, MAP, soil depth, treatment intensity, etc.) of soil functional indicators from literature. Data is directly read in tables, and graphical data is digitized using WebPlotDigitizer; Missing climate variables are supplemented using WorldClim v2.1.\n
3. Spatial data: Based on the distribution data of terrestrial ecosystems, obtain ecosystem types and use the \"getData\" function to extract altitude.\n
4. Future climate data: The CMIP6 ACCESS model is used to obtain the changes in temperature, precipitation, and nitrogen deposition in China's arid regions from 1970 to 2000 and under SSP1-2.6 scenarios at the end of the 21st century (2081 to 2100).", "ds_quality": "
Model validation and selection: Using 10 fold cross validation and fixed random seeds, select the optimal model based on RMSE and R ². Overall, models with different global change factors have acceptable to high predictive power and can be used to assess changes in soil multifunctionality under future scenarios.", "ds_ref_way": "", "ds_abstract": "
This dataset aims to explore the impact of climate change and human activities on soil multifunctionality in arid regions of China. By searching relevant literature from Web of Science and China National Knowledge Infrastructure from 1900 to May 2023, 841 field studies covering 18189 sets of observational data were selected. The data includes indicators related to soil function, climate data (such as annual average temperature, precipitation, nitrogen deposition, etc.), ecosystem type, soil depth, and treatment intensity. Construct a predictive model using the H2O AutoML package to analyze the impact of future (late 21st century) climate change (temperature, precipitation, nitrogen deposition) and human activities (such as cultivation, grazing, and returning farmland to forests) on soil multifunctionality. The data quality was confirmed through 10 fold cross validation and multiple evaluation indicators (such as R ² and RMSE), and the results showed that the constructed model can effectively predict future changes in soil multifunctionality, with high reliability and accuracy.", "ds_time_res": "年", "ds_acq_place": "China's arid regions", "ds_space_res": "5000m", "ds_projection": "", "ds_process_way": "
1. Construction of machine learning prediction model: Using H2O AutoML with lnRRSMF as the response variable, MAT、MAP、 Select the optimal model for prediction based on factors such as altitude, ecosystem type, soil depth, and treatment intensity/duration.\n
2. Construction of climate change prediction dataset: Integrate future temperature, precipitation, nitrogen deposition changes, and environmental factors in arid areas to form climate change input data for prediction.\n
3. Setting of human activity processing intensity: Normalize grazing intensity to 1 head per hectare; The duration of ecological engineering treatments such as cultivation, fencing, and returning farmland to forests will be uniformly set until the end of the 21st century (75 years).\n
4. Soil profile setting: Construct 10 layers of soil data ranging from 0-30 cm (3 cm interval) to cover 80% of the sampling depth in the literature and improve prediction robustness.\n
5. Model prediction and result output: Use H2O.redist to predict each layer of soil data, take the average of the results and convert it into a percentage change, and generate a grid format for mapping.", "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, "ds_topic_tags": [ "中国干旱区", "土壤功能", "气候变化", "人类活动" ], "ds_subject_tags": [ "气候学" ], "ds_class_tags": [], "ds_locus_tags": [ "中国干旱区" ], "ds_time_tags": [ 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010, 2011, 2012, 2013, 2014, 2015, 2016, 2017, 2018, 2019, 2020, 2021, 2100 ], "ds_contributors": [ { "true_name": "高斯远", "email": "52263903011@stu.ecnu.edu.cn", "work_for": "华东师范大学", "country": "中国" }, { "true_name": "刘慧颖", "email": "hyliu@des.ecnu.edu.cn", "work_for": "华东师范大学", "country": "中国" } ], "ds_meta_authors": [ { "true_name": "高斯远", "email": "52263903011@stu.ecnu.edu.cn", "work_for": "华东师范大学", "country": "中国" } ], "ds_managers": [ { "true_name": "刘慧颖", "email": "hyliu@des.ecnu.edu.cn", "work_for": "华东师范大学", "country": "中国" } ], "category": "其他" }