Mapping Forest Characteristics at Fine Resolution across Large Landscapes of the Southeastern United States Using NAIP Imagery and FIA Field Plot Data

Abstract

Accurate information is important for effective management of natural resources. In the field of forestry, field measurements of forest characteristics such as species composition, basal area, and stand density are used to inform and evaluate management activities. Quantifying these metrics accurately across large landscapes in a meaningful way is extremely important to facilitate informed decision-making. In this study, we present a remote sensing based methodology to estimate species composition, basal area and stand tree density for pine and hardwood tree species at the spatial resolution of a Forest Inventory Analysis (FIA) program plot (78 m by 70 m). Our methodology uses textural metrics derived at this spatial scale to relate plot summaries of forest characteristics to remotely sensed National Agricultural Imagery Program (NAIP) aerial imagery across broad extents. Our findings quantify strong relationships between NAIP imagery and FIA field data. On average, models of basal area and trees per acre accounted for 43% of the variation in the FIA data, while models identifying species composition had less than 15.2% error in predicted class probabilities. Moreover, these relationships can be used to spatially characterize the condition of forests at fine spatial resolutions across broad extents.

A New Pseudoinvariant Near-Infrared Threshold Method for Relative Radiometric Correction of Aerial Imagery

“…Compared to satellite imagery programs such as Landsat and MODIS, the NAIP data are acquired from aerial platforms using different sensors that have lower radiometric resolutions and the information on sensor characteristics is less available and consistent over time [1]. Because of the small swath of NAIP images (approximately 7 km × 8 km), the acquisition of imagery for one region usually involves multiple flights that may last weeks or potentially even months, and the mosaics of NAIP images have artifacts associated with atmospheric interference, viewing geometry, illumination, shadows, time of the day, and plant phenology [1,14]. More importantly, NAIP images are distributed in the format of digital numbers (DNs), which are integer values to facilitate computation and transmission and to scale brightness for convenient display [15].…”

“…This is important to the calculations of many multispectral indices, such as the normalized difference vegetation index (NDVI) and the normalized difference water index (NDWI), that are very sensitive to atmospheric effects. However, many studies directly applied NAIP without correction for spectral analysis such as NDVI calculation, land cover classification, and vegetation cover estimation [3,4,6,8,9,[11][12][13][14]17]. This is mainly due to a lack of easy access to radiometric response data for the sensors used and the lack of methods for retrieving surface reflectance from NAIP DN values [1].…”

A Comparison of Standard Modeling Techniques Using Digital Aerial Imagery with National Elevation Datasets and Airborne LiDAR to Predict Size and Density Forest Metrics in the Sapphire Mountains MT, USA

“…Similarly, fine resolution passive airborne imagery produced by organizations such as the National Agricultural Imagery Program (NAIP) [5] have also been used to estimate forest metrics [6][7][8][9]. In particular, strong relationships have been developed between BAWD, QMD, BAH, and TPH using spectral values, measures of image texture [10], and topographic variables derived from imagery such…”

Chimera: A Multi-Task Recurrent Convolutional Neural Network for Forest Classification and Structural Estimation

“…However, since these plots are non-forested by definition, these locations are not-visited and have no tree attributes recorded but may still have tree cover (e.g., urban and residential areas) [61]. Following the methods of Hogland et al [35], who visually inspected plot locations against the corresponding reference NAIP image, we manually inspected images corresponding to these non-visited plots. For locations where there were clearly no trees visible in the imagery, we labeled them as 'none' and attributed them with AGB, QMD, basal area, and canopy cover values of zero.…”

“…Predictions of forest structure metrics have been shown to improve when including image textures as predictors, e.g., gray level co-occurrence matrix (GLCM), standard deviation of gray levels (SDGL). However, these, and other more complex textural features require laborious hand-crafting to identify and implement in model fitting [34][35][36]. CNNs have the ability to identify similar relevant image textures from the data alone without human assistance, which allows them to be applied to solve generalized image feature detection problems [32,37].…”