Data for Spatially distributed overstory and understory leaf area index estimated from forest inventory data

This resource contains the data and scripts used for:
Goeking, S. A. and D. G. Tarboton, (2022). Spatially distributed overstory and understory leaf area index estimated from forest inventory data. Water. https://doi.org/10.3390/w1415241.

Abstract from the paper:
Abstract: Forest change affects the relative magnitudes of hydrologic fluxes such as evapotranspiration (ET) and streamflow. However, much is unknown about the sensitivity of streamflow response to forest disturbance and recovery. Several physically based models recognize the different influences that overstory versus understory canopies exert on hydrologic processes, yet most input datasets consist of total leaf area index (LAI) rather than individual canopy strata. Here, we developed stratum-specific LAI datasets with the intent of improving the representation of vegetation for ecohydrologic modeling. We applied three pre-existing methods for estimating overstory LAI, and one new method for estimating both overstory and understory LAI, to measurements collected from a probability-based plot network established by the US Forest Service’s Forest Inventory and Analysis (FIA) program, for a modeling domain in Montana, MT, USA. We then combined plot-level LAI estimates with spatial datasets (i.e., biophysical and re-mote sensing predictors) in a machine learning algorithm (random forests) to produce annual gridded LAI datasets. Methods that estimate only overstory LAI tended to underestimate LAI relative to Landsat-based LAI (mean bias error ≥ 0.83), while the method that estimated both overstory and understory layers was most strongly correlated with Landsat-based LAI (r2 = 0.80 for total LAI, with mean bias error of -0.99). During 1984-2019, interannual variability of under-story LAI exceeded that for overstory LAI; this variability may affect partitioning of precipitation to ET vs. runoff at annual timescales. We anticipate that distinguishing overstory and understory components of LAI will improve the ability of LAI-based models to simulate how for-est change influences hydrologic processes.

This resource contains one CSV file, two shapefiles (each within a zip file), two R scripts, and multiple raster datasets. The two shapefiles represent the boundaries of the Middle Fork Flathead river and South Fork Flathead River watersheds. The raster datasets represent annual leaf area index (LAI) at 30 m resolution for the entire modeling domain used in this study. LAI was estimated using method LAI4, which produced separate overstory and understory LAI datasets. Filenames contain years, e.g., "LAI4_2019" is overstory LAI for 2019; "LAI4under_2019" is understory LAI for 2019.

The CSV files in this Resource contain annual time series of LAI and ET ratio (annual evapotranspiration divided by annual precipitation) for the South Fork Flathead River and Middle Fork Flathead River watersheds, 1984-2019. LAI methods represented in this time series are LAI1 and LAI4 from the paper. LAI1 consists of only overstory LAI, and LAI4 consists of overstory (LAI4), understory (LAI4_under), and total (LAI4_total) LAI. For each LAI estimation method, summary statistics of the entire watershed are included (min, first quartile, median, third quartile, and max).

The two R scripts (R language and environment for statistical computing) summarize Forest Inventory & Analysis (FIA) data from the FIA database (FIADB) to estimate LAI at FIA plots.
1) FIADB_queries_public.r: Script for compiling FIA plot measurements prior to estimating LAI
2) LAI_estimation_public: Script for estimating LAI at FIA plots using the four methods described in this paper

Before running the R scripts, users must obtain several FIADB tables (PLOT, COND, TREE, and P2VEG_SUBP_STRUCTURE; all four tables must be renamed with lower-case names, e.g., "plot"). These tables can be obtained using one of two methods:
1) By downloading CSV files for the appropriate U.S. state(s) from the FIA DataMart (https://apps.fs.usda.gov/fia/datamart/datamart.html). If this method is used, the CSV files must be imported (read) into R before proceeding.
2) By using r package 'rFIA' to download the tables from FIADB for the U.S. state(s) of interest.

Note that publicly available plot coordinates are accurate within 1 km and are not true plot locations, which are legally confidential to protect the integrity of the sample locations and the privacy of landowners. Access to true plot location data requires review by FIA's Spatial Data Services unit, who can be contacted at [e-mail].

本资源包含用于以下研究的数据和脚本：Goeking, S. A. 和 D. G. Tarboton (2022) 从森林资源数据中估计的分布式的林冠层和亚林冠层叶面积指数。水。https://doi.org/10.3390/w1415241。

论文摘要如下：
摘要：森林变化影响水分通量如蒸散量（ET）和地表径流的相对大小。然而，关于地表径流对森林干扰和恢复的响应敏感性的知识仍有很多未知。一些基于物理学的模型认识到林冠层和亚林冠层对水文过程的不同影响，但大多数输入数据集包含的是总叶面积指数（LAI），而非单个林冠层。在此，我们开发了针对特定层级的LAI数据集，旨在提高生态水文模型中对植被的表征。我们应用了三种预先存在的估计林冠层LAI的方法，以及一种新方法来估计林冠层和亚林冠层的LAI，这些方法均应用于美国森林服务森林资源与分析（FIA）项目建立的基于概率的样地网络收集的测量数据，用于美国蒙大拿州MT的建模区域。随后，我们将样地级别的LAI估计值与空间数据集（即生物物理和遥感预测因子）结合，通过机器学习算法（随机森林）生成年度网格化的LAI数据集。仅估计林冠层LAI的方法往往低估了相对于基于Landsat的LAI的LAI（平均偏差误差≥0.83），而同时估计林冠层和亚林冠层的方法与基于Landsat的LAI的相关性最强（总LAI的r2 = 0.80，平均偏差误差为-0.99）。在1984-2019年期间，亚林冠层LAI的年际变化超过了林冠层LAI的变化；这种变化可能会影响降水在蒸散量与径流之间的分配。我们预期区分LAI的林冠层和亚林冠层成分将提高基于LAI的模型模拟森林变化如何影响水文过程的能力。

此资源包含一个CSV文件、两个shapefile（每个shapefile包含一个zip文件）、两个R脚本和多个栅格数据集。这两个shapefile代表Middle Fork Flathead河流和South Fork Flathead河流流域的边界。栅格数据集代表整个建模区域中30米分辨率的年度叶面积指数（LAI）。LAI使用方法LAI4进行估计，该方法产生了独立的林冠层和亚林冠层LAI数据集。文件名包含年份，例如，“LAI4_2019”是2019年的林冠层LAI；“LAI4under_2019”是2019年的亚林冠层LAI。

本资源中的CSV文件包含1984-2019年South Fork Flathead河流和Middle Fork Flathead河流流域的年度LAI和ET比率（年蒸散量除以年降水量）的时间序列。此时间序列中包含的LAI方法有论文中的LAI1和LAI4。LAI1仅包含林冠层LAI，而LAI4包含林冠层（LAI4）、亚林冠层（LAI4_under）和总LAI（LAI4_total）。对于每种LAI估计方法，整个流域的汇总统计量都包括（最小值、第一四分位数、中位数、第三四分位数和最大值）。

两个R脚本（R语言和统计计算环境）用于从FIA数据库（FIADB）中总结森林资源与分析（FIA）数据以估计FIA样地的LAI。
1) FIADB_queries_public.r：在估计LAI之前编译FIA样地测量的脚本
2) LAI_estimation_public：使用本文中描述的四种方法估计FIA样地LAI的脚本

在运行R脚本之前，用户必须获取几个FIADB表（PLOT、COND、TREE和P2VEG_SUBP_STRUCTURE；所有四个表都必须使用小写名称重命名，例如，“plot”）。这些表可以通过以下两种方法之一获取：
1) 从FIA DataMart（https://apps.fs.usda.gov/fia/datamart/datamart.html）下载相应美国州（s）的CSV文件。如果使用此方法，必须在继续之前将CSV文件导入（读取）到R中。
2) 使用r包'rFIA'从FIADB下载感兴趣的美国州（s）的表。

请注意，公开可用的样地坐标在1公里范围内是准确的，并非真实的样地位置，这些位置因保护样本位置的完整性和土地所有者的隐私而具有法律上的保密性。访问真实样地位置数据需要由FIA的空间数据服务部门进行审查，可以通过[e-mail]联系。