Texas Statewide Landcover, Ecological Systems, and Percentage Canopy Classifications (10-meter Resolution) (2021)

Statewide landcover, ecosystem, and percentage canopy cover classifications for Texas were mapped at a spatial resolution of 10-meters. Classifications were run for 16 zones across the state corresponding to available cloud-free multitemporal Sentinel-2 satellite imagery for each zone. For each zone, RandomForest classifications were run using data stacks comprised of spectral bands from three dates (winter, early growing season, late growing season/leaf-off) of imagery, as well as multiple vegetation indices (NDVI, EVI2, MSAVI2). Over 50,000 training points were selected from ground trips and high-resolution aerial image surveys to run the entire pixel-based classification. The overlapping zones were merged using a feathering algorithm to produce a single statewide land-cover classification map. The landcover mapping results were further refined using multiple spatial masks (e.g., urban, water, crop) along with logical rulesets and ancillary data. To map ecological systems, the land-cover classification was then intersected with an enduring features dataset derived primarily from soil map-unit polygons (gSSURGO) and other geophysical variables. Additionally, we produced a statewide percentage canopy cover map at a 10-meter spatial resolution using multiple techniques. For the western 2/3 of the Texas, a nested machine learning approach was used (Sunde et al., 2020), and for the eastern 1/3 of the state, a combination of LiDAR derived training data and machine learning was used.

This dataset includes four items:

"TX_10m_landcover_2021.zip" - Statewide landcover classification for Texas (10-meter spatial resolution)

"TX_10m_ecoclass_map_2021.zip" - Statewide ecological systems classification for Texas (10-meter spatial resolution)

"TX_10m_canopy_cover_2021.zip" - Statewide percentage canopy cover map for Texas (10-meter spatial resolution)

"TX_lu_ecoclass_key.xlsx" - Table containing keys for the mapped landcover, canopy, and ecological systems classes

(To facilitate display of the datasets within ESRI software, .lyr files are included in the respective archive folders)

This work was funded by the Texas A&M Forest Service.

本数据集针对德克萨斯州完成了全域土地覆被（landcover）、生态系统及冠层覆盖百分比的分类制图，空间分辨率达10米。本次分类覆盖全州16个分区，每个分区均采用对应时段的无云多时相哨兵二号（Sentinel-2）卫星影像开展分类。针对每个分区，研究采用由三个成像时段（冬季、生长季早期、生长季晚期/落叶期）的影像光谱波段，以及多种植被指数（归一化差分植被指数NDVI、增强型植被指数2 EVI2、修正土壤调整植被指数2 MSAVI2）构成的数据集栈，开展随机森林（RandomForest）分类。本次全像素分类模型的训练样本共计超5万个，均采集自野外实地调查与高分辨率航空影像解译。重叠分区通过羽化（feathering）算法进行拼接，最终生成单张全域土地覆被分类图。后续通过多类空间掩码（如城镇、水体、耕地）、逻辑规则集及辅助数据，对土地覆被制图成果进行了精细化修正。为完成生态系统制图，研究将土地覆被分类结果与主要基于土壤图斑多边形（gSSURGO）及其他地球物理变量生成的持久特征数据集进行空间叠加。此外，本数据集还采用多种技术生成了空间分辨率为10米的全域冠层覆盖百分比分布图。其中，德克萨斯州西部2/3区域采用嵌套式机器学习方法（Sunde等，2020）完成制图，东部1/3区域则结合激光雷达（LiDAR）衍生训练数据与机器学习技术开展制图。

本数据集包含以下四项内容：
"TX_10m_landcover_2021.zip"——德克萨斯州全域10米分辨率土地覆被分类成果
"TX_10m_ecoclass_map_2021.zip"——德克萨斯州全域10米分辨率生态系统分类成果
"TX_10m_canopy_cover_2021.zip"——德克萨斯州全域10米分辨率冠层覆盖百分比分布图
"TX_lu_ecoclass_key.xlsx"——包含本次制图中土地覆被、冠层覆盖及生态系统分类代码对照表的表格

（为便于在ESRI软件中加载展示数据集，各压缩包文件夹内均附带.lyr格式图层文件）

本研究由德克萨斯农工大学林业服务处（Texas A&M Forest Service）资助。