High resolution and high cadence time series of land surface categories, land use land cover, and land use land cover changes

Name: High resolution and high cadence time series of land surface categories, land use land cover, and land use land cover changes
Creator: Zenodo
Published: 2023-06-20T10:42:31+08:00

Zenodo2023-06-20 更新2026-05-26 收录

数据链接：

https://zenodo.org/record/7924340 数据链接链接失效反馈

官方服务：

资源简介：

A prototype of monthly, 10 m resolution land surface categories, land use land cover (LULC) cover, and LULC change maps derived from Sentinel-2 data over three areas within Belgium, Portugal, and Sicily for the period 2018-2020. The LULC and LULC change maps were independently validated by IIASA. All products were generated within the framework of the RapidAI4EO project, funded from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 101004356. The data description can be found below. The validation report of the LULC and LULC change maps can be found in validation_LULC.pdf and validation_change.pdf, respectively, and the validation dataset can be found in Lesiv et al. (2023). Data description Increasing the cadence of the land cover updates from the typical (multi-)annual to monthly cadence poses several challenges. First, several land cover types are difficult to discriminate without any knowledge of temporal dynamics. For instance, croplands are characterized by a dynamic of vegetation growth and a harvest period (i.e. cycles of bare soil, sparsely vegetated and vegetated periods). This contrasts with grasslands that often lack the harvest period resulting in a bare soil cover. Without this temporal information, it is difficult to distinguish a vegetated cropland field from grassland. Second, phenological changes may introduce a large intra-class variability and thus also confusion between classes. For example, the shedding of leaves during autumn or wilting of herbaceous vegetation in dry summer periods introduces spectral variability within land cover classes. To overcome these challenges, we developed a workflow with two main phases. The first phase aims to map land surface categories (LSC) at a monthly resolution. The next phase uses the resulting monthly LSC probability time series to classify land cover. Land surface category (LSC) These LSC represent basic, observable bio-geophysical properties (categories) of the Earth surface that can be predicted directly from individual monthly composites. LSC classes contain a set of vegetated and non-vegetated surface categories. Discrete LSC classification legend: Map code Land cover class 11 Tree (leaf-on) 12 Shrubland (leaf-on) 13 Grassland 14 Woody vegetation (leaf-off) 15 Wilted herbaceous vegetation 21 Bare/sparse vegetation 22 Water 24 Built-up In order to predict the LSC, we trained a CatBoost model (Dorogush et al., 2018) using a DEM, spectral bands and vegetation indices, country, the timing (month) of the spectral data, and the pseudo-probability of a U-Net model trained to segment built-up surfaces as input. Labels were derived by post-processing the land cover labels of the ESA WorldCover product (Zanaga et al., 2021). Please note that the collection of these labels was suboptimal, likely having an impact on the LULC and change maps generated in the prototype. Land use land cover After predicting LSC over the three AOI’s, we trained a CatBoost model using the LSC probabilities over a window of one year, country, and the timing (month) as independent variable. The use of LSC probabilities over multiple months allows to incorporate information about dynamics, which is necessary to discriminate some classes (e.g. cropland and grassland or cropland and bare). Similar to the LSC labels, the LULC labels were derived from the ESA WorldCover product v100 (year 2020), resulting in a similar legend system. The use of a moving window approach to predict LULC allows to (i) incorporate temporal information that is necessary to discriminate land cover classes and (ii) is expected to lead to more consistent land cover maps. It however has the disadvantage that (i) no land cover predictions are available at the beginning and the end of the time series and (ii) the timing of the predicted land cover change is not always accurate. To resolve these issues, we applied a post-processing step that compares and integrates the LULC predictions and cleaned LSC predictions. Discrete LC classification legend: Map code Land cover class 10 Tree cover 20 Shrubland 30 Grassland 40 Cropland 50 Built-up 60 Bare/sparse vegetation 80 Permanent water bodies 90 Herbaceous wetland Land use land cover change Monthly change maps were finally derived from the land cover maps. The pixel values within the change maps represent the percentage of pixels that changed with respect to the previous month over an area of 90x90m. The maps contain values between 0-100, with larger values assigned to larger change patches. A value of 100 indicates that all pixels within an area of 90x90m around the pixel were flagged as change. Files The zip files contain the following data: lsc.zip: land surface category maps over the three AOI’s lc.zip: LULC maps over the three AOI’s change.zip: change maps over the three AOI’s These maps are generated for each month over the period 2018-2020 for each of the tiles (see tiles.gpkg for an overview of all tiles). The files names use the following naming convention: “tile-year-month.tif”. References Myroslava Lesiv, Halyna Bun, & Martina Duerauer. (2023). Validation data set on land cover changes for RapidAI4EO project [Data set]. Zenodo. https://doi.org/10.5281/zenodo.7825963 Dorogush, A. V., Ershov, V., & Gulin, A. (2018). CatBoost: gradient boosting with categorical features support. arXiv preprint arXiv:1810.11363. Zanaga, D., et al.., 2021. ESA WorldCover 10 m 2020 v100. https://doi.org/10.5281/zenodo.5571936

本数据集为2018-2020年期间，基于哨兵二号（Sentinel-2）数据，针对比利时、葡萄牙及西西里岛三个研究区（Area of Interest, AOI）生成的月度、10米分辨率地表类别、土地利用/土地覆被（Land Use Land Cover, LULC）及LULC变化图原型。相关LULC及LULC变化图已由国际应用系统分析研究所（IIASA）独立完成验证。所有产品均在RapidAI4EO项目框架下生成，该项目由欧盟地平线2020（Horizon 2020）研究与创新计划资助，资助协议编号为101004356。数据描述详见下文。LULC及LULC变化图的验证报告分别见`validation_LULC.pdf`与`validation_change.pdf`，验证数据集可参考Lesiv等人（2023）的研究。 数据描述 将土地覆被更新频率从典型的（多年/年度）提升至月度尺度，会带来诸多挑战。其一，若不借助时间动态信息，部分土地覆被类型难以区分。例如，农田具有植被生长与收获的动态周期（即裸土、稀疏植被与植被覆盖的循环阶段），而草地通常无收获周期，仅存在裸土覆盖阶段。若缺乏此类时序信息，便难以区分处于植被覆盖期的农田与草地。其二，物候变化会引发显著的类内变异性，进而导致类别间混淆。例如，秋季落叶或夏季干旱期草本植被枯萎，会在同一土地覆被类别内引入光谱差异。为应对上述挑战，我们开发了包含两个主要阶段的工作流：第一阶段旨在生成月度分辨率的地表类别（Land Surface Category, LSC）图；第二阶段则利用生成的月度LSC概率时间序列，完成土地覆被分类。 地表类别（LSC） 此类LSC代表可直接从单张月度合成影像预测的地球表面基础、可观测的生物物理属性（类别）。LSC类别涵盖多组植被与非植被覆盖类型。离散LSC分类图例如下： | 地图编码 | 土地覆被类别 | | --- | --- | | 11 | 乔冠林（叶茂期） | | 12 | 灌丛（叶茂期） | | 13 | 草地 | | 14 | 木本植被（落叶期） | | 15 | 萎蔫草本植被 | | 21 | 裸地/稀疏植被 | | 22 | 水体 | | 24 | 建成区 | 为预测LSC，我们采用数字高程模型（DEM）、光谱波段与植被指数、国家信息、光谱数据的采集月份，以及经过训练的U-Net模型对建成区表面进行分割得到的伪概率作为输入，训练了CatBoost模型（Dorogush等人，2018）。标签源自对欧洲空间局世界覆被（ESA WorldCover）产品（Zanaga等人，2021）的土地覆被标签进行后处理得到。需注意，此类标签的采集并非最优，可能对本原型生成的LULC及变化图产生影响。 土地利用/土地覆被（LULC） 在对三个研究区完成LSC预测后，我们以一年时间窗口内的LSC概率、国家信息及采集月份作为自变量，训练了CatBoost模型。利用跨多个月份的LSC概率，可纳入动态信息，这对区分部分类别（如农田与草地、农田与裸地）至关重要。与LSC标签类似，LULC标签源自2020年版欧洲空间局世界覆被产品v100，采用与之一致的图例体系。采用移动窗口法预测LULC，可实现：(i) 纳入区分土地覆被类别所需的时序信息；(ii) 生成更为一致的土地覆被图。但该方法也存在两点局限：(i) 时间序列首尾时段无土地覆被预测结果；(ii) 预测的土地覆被变化时间未必完全准确。为解决这些问题，我们加入了后处理步骤，对LULC预测结果与清洗后的LSC预测结果进行比对与整合。离散LULC分类图例如下： | 地图编码 | 土地覆被类别 | | --- | --- | | 10 | 乔木覆被 | | 20 | 灌丛 | | 30 | 草地 | | 40 | 农田 | | 50 | 建成区 | | 60 | 裸地/稀疏植被 | | 80 | 永久性水体 | | 90 | 草本湿地 | 土地利用/土地覆被变化 最终从土地覆被图中生成月度变化图。变化图内的像素值代表90×90米区域内，相较于上月发生变化的像素占比。地图取值范围为0-100，数值越高代表变化斑块越大。取值100表示该像素周边90×90米区域内的所有像素均被标记为发生变化。 文件说明 压缩包包含以下数据： - `lsc.zip`：三个研究区的地表类别图 - `lc.zip`：三个研究区的LULC图 - `change.zip`：三个研究区的变化图上述地图为2018-2020年期间每个月针对所有瓦片生成，所有瓦片的概览可参考`tiles.gpkg`。文件命名遵循以下约定：`tile-年-月.tif`。 参考文献 1. Myroslava Lesiv, Halyna Bun, & Martina Duerauer. (2023). 用于RapidAI4EO项目的土地覆被变化验证数据集[数据集]. Zenodo. https://doi.org/10.5281/zenodo.7825963 2. Dorogush, A. V., Ershov, V., & Gulin, A. (2018). CatBoost: 支持类别特征的梯度提升树模型. arXiv预印本arXiv:1810.11363. 3. Zanaga, D., et al., 2021. ESA WorldCover 10 m 2020 v100. https://doi.org/10.5281/zenodo.5571936

提供机构：

Zenodo

创建时间：

2023-06-20