全球30米不透水面產品（MSMT_IS30）2015年
下載鏈接：https://zenodo.org/record/3505079#.YV2oWZpBxnI
http://www.geodata.cn/data/datadetails.html?dataguid=214900664506554&docid=1
論文鏈接：https://essd.copernicus.org/articles/12/1625/2020/essd-12-1625-2020.html
全球30米不透水面數據產品（2015年），該產品由中國科學院空天信息創新研究院劉良雲研究員團隊生產。
人工不透水面地表因光譜和空間結構異常復雜，高精度的全球不透水面提取存在極大挑戰。傳統方法僅依賴於光學數據或雷達數據的制圖策略往往很難將不透水面地表和裸地完全區分開來，從而導致了不透水面地類存在較為嚴重的誤分現象。該團隊提出了基於多源多時相遙感數據的不透水面提取算法和基於GEE平台的全球不透水面產品生產框架。首先，利用GlobeLand30地表覆蓋產品、VIIRS夜間燈光數據和MODIS EVI植被指數產品，自動提取了全球高置信度的人工不透水面分類的訓練樣本。其次，利用多時相Landsat-8 OLI反射率特征、Sentinel-1 SAR結構特征和SRTM/ASTER DEM地形特征，采用隨機森林分類模型，以5°網格進行了逐區塊地自適應隨機森林建模。最后，利用GEE雲平台的數據、存儲和計算資源以及隨機森林分類模型，逐區塊地生產了不透水面產品，並經過地理拼接生產了2015年全球30米不透水面產品 MSMT_IS30-2015。
The GlobeLand30 data were also used to automatically derive the global impervious and nonimpervious training samples.
做全球的不透水面可以選擇不同大陸的一些典型城市來展示，以及不同大陸上的網格區域來進行一些操作，比如變量重要性度量
全球30米分辨率(1985-2015)城市用地動態擴張數據集 NUACI
下載鏈接
論文鏈接
The NUACI-based maps, developed using the spectral index-based method applied to Landsat and DMSP OLS NTL imagery, are multitemporal global 30 m impervious surface datasets (Liu et al., 2018). This map has been validated as having an overall accuracy of 0.81–0.84 and kappa coefficient of 0.43–0.50 at the global level (Liu et al., 2018).
這套數據感覺精度有點低
FROM_GLC數據（GAIA）
下載鏈接
論文鏈接
FROM-GLC, first produced in 2010, was the first 30 m resolution global land-cover dataset and was produced by the supervised classification of 8900 Landsat images (Gong et al., 2013). The second generation of FROM-GLC from 2015 (FROM-GLC-2015; http://data.ess.tsinghua.edu.cn/, last access: 8 July 2020) was produced by using multiseasonal Landsat imagery acquired between 2013 and 2015 and incorporates the day of year, geographical coordinates and elevation data (Li et al., 2017).
GHSL數據（全球人居圖層）
The GHSL, a global information baseline describing the spatial evolution of human settlements in the past 40 years, was developed by using a symbolic machine learning model trained by the collected high-resolution samples, multitemporal Landsat imagery in the epochs 1975, 1990, 2000 and 2015 (Florczyk et al., 2019). The GHSL impervious surface map at 30 m for 2015 (GHSL-2015; https://ghsl.jrc.ec.europa.eu/download.php, last access: 8 July 2020) achieved an overall accuracy of 96.28 % and kappa coefficient of 0.3233 validated using Land Use/Cover Area frame Survey reference data (Pesaresi et al., 2016).
GlobeLand30
GlobeLand30 is an operational 30 m global land-cover dataset produced using the pixel–object–knowledge-based method (POK-based) approach in 2000 and 2010 (Chen et al., 2015). In this study, the global impervious product derived from GlobeLand30 in 2010 (GlobeLand30-2010; http://www.globallandcover.com/GLC30Download/index.aspx, last access: 8 July 2020) was produced by combining pixel-based classification, multiscale segmentation and manual editing based on the high-resolution imagery and had been validated as having a user's accuracy of 86.7 %.
HBASE-2010
下載鏈接
The HBASE dataset was the first global 30 m dataset of artificial impervious cover derived from the Global Land Survey (GLS) Landsat data for 2010. It was produced by combining meter-resolution training data (exceeding 20 million), OpenStreetMap, VIIRS NTL, GLS Landsat SR and MODIS normalized difference vegetation index (NDVI) products and achieved a kappa coefficient of 0.91 using scene-level cross validation in Europe (Wang et al., 2017a, b).
論文：
Wang, P., C. Huang, E. C. Brown de Colstoun, J. C. Tilton, and B. Tan. 2017. Global Human Built-up And Settlement Extent (HBASE) Dataset From Landsat. Palisades, NY: NASA Socioeconomic Data and Applications Center (SEDAC).