Mitigating the Impact of Field and Image Registration Errors through Spatial Aggregation

Abstract

Remotely sensed data are commonly used as predictor variables in spatially explicit models depicting landscape characteristics of interest (response) across broad extents, at relatively fine resolution. To create these models, variables are spatially registered to a known coordinate system and used to link responses with predictor variable values. Inherently, this linking process introduces measurement error into the response and predictors, which in the latter case causes attenuation bias. Through simulations, our findings indicate that the spatial correlation of response and predictor variables and their corresponding spatial registration (co-registration) errors can have a substantial impact on the bias and accuracy of linear models. Additionally, in this study we evaluate spatial aggregation as a mechanism to minimize the impact of co-registration errors, assess the impact of subsampling within the extent of sample units, and provide a technique that can be used to both determine the extent of an observational unit needed to minimize the impact of co-registration and quantify the amount of error potentially introduced into predictive models.

Using Forest Inventory Data with Landsat 8 Imagery to Map Longleaf Pine Forest Characteristics in Georgia, USA

“…This suggests that for the extent of a FIA plot there can be substantial variation in summarized estimates and that estimated mean values can vary significantly from true mean values, especially for trees less than 12.7 cm in DBH occurring on heterogeneous plots. To draw stronger relationships between plot measurements and Landsat 8 metrics, plots should be sampled using a design that covers more of the extent of the plot [46] than what is currently specified within the FIA plot protocol [24]. Furthermore, field plot layouts, such as described in Reference [46], should also help to improve the accuracy and precision of models that link field plot data to imagery.…”

Multi-Scale Remote Sensing-Assisted Forest Inventory: A Glimpse of the State-of-the-Art and Future Prospects

“…Motivated by the recent trends and global interest, the Remote Sensing special issue "Remote Sensing-Based Forest Inventories from Landscape to Global Scale" hosted nine peer-reviewed papers adopting various modern applications of passive and active remote sensing data for multi-scale forest inventory applications. This special issue is enriched with a series of independent, though contextually related, recent studies from diverse geographical domains of the globe, including the near-Arctic Canada [10], Northern United States [11,12], Northern Japan [13], Southern Spain [14,15], Central Italy [16], Southern Poland [17] and Western Germany [18].…”

“…This was shown to reduce the variance of the one-phase simple random sampling estimator by an average of 43% and 25% for district and state levels, respectively. Furthermore, scale was also addressed by a number of other papers in a range of realms, for example by [12] (spatial aggregation of field and image objects for co-registration) and [10] (tree-and plot-level estimation of height and basal area by means of stereo WorldView-3 imagery). All in all, drawing attention to the general topic of scale (as one of the challenging, yet still insufficiently addressed issues in forestry remote sensing) was definitely within the essential focuses when abstracting this special issue by the guest editors, which was well represented via a number of original studies published here.…”

“…Spectral information from multi-temporal, optical Landsat imagery was also shown by [16] to enable good approximations for forest recovery via the use of multiple vegetation indices, highlighting the tremendous information content within the time series of medium-resolution satellite imagery for monitoring forest stand dynamics on a regional scale and beyond. Finally, [12] followed the important though often underestimated preprocessing step of image co-registration by suggesting a simulated solution based on spatial aggregation over adjacent pixels, which helps with precisely quantifying the co-registration error between the ground control points and satellite data.…”

“…Moreover, very high resolution, often mono-temporal satellite imagery (UAV and aerial photogrammetry, WorldView-3 data), was employed by [10,11,13], which mainly aimed at quantifying local forest structural details at high geometric resolutions. High-resolution optical data, including RapidEye and SPOT with geometric resolutions <10 m, were applied by [18], whereas medium-resolution (Landsat and Sentinel-2 series) [12,16,17] and low-resolution [14] data were also represented, mainly due to their proven ability for monitoring forest dynamics over timespans. In this realm, further applications of such multi-temporal series of data for regional and statewide inventory and analysis of forest structure and health are expected in the near future, either standalone or in combination with other costly, i.e., high resolution, sources of data.…”