20 cm resolution mapping of tundra vegetation communities provides an ecological baseline for important research areas in a changing Arctic environment

Abstract

Arctic tundra vegetation communities are spatially heterogeneous and may vary dramatically from one meter to the next. Consequently, representing Arctic tundra vegetation communities accurately requires very high resolution raster maps (<5 m grid cell size). However, using remotely sensed data to produce maps with sufficient spatial detail at an extent appropriate for understanding landscape-scale ecological patterns is challenging. In this study, we used predictor layers derived from airborne lidar and high-resolution (∼5 cm) 4-band airborne imagery to classify vegetation communities at 20 cm spatial resolution for three landscapes (12.5 km 2 total) near the Toolik Lake research area in the Alaskan Low Arctic. The maps were built using a Random Forest model that was trained and tested on 800 ground reference plots, using classes derived from commonly used legends on existing polygon maps of the area. Withheld test plots (25% of dataset) had a balanced map accuracy of 0.57, kappa of 0.47, and weighted (fuzzy) kappa of 0.65. These maps provide high-resolution plant community information that can serve as important baseline reference data for vegetation monitoring and change detection in this rapidly changing tundra ecosystem, and as validation for coarser scale maps. They also permit fine-scale characterization of landscape phenomena such as community-level nutrient dynamics and wildlife habitat suitability in an important Arctic research site. Our approach demonstrates that very high resolution mapping results can be achieved and validated by integrating high-resolution remote-sensing datasets from multiple sensors in a machine learning model trained on simple field reference data.

Use of Very High Spatial Resolution Commercial Satellite Imagery and Deep Learning to Automatically Map Ice-Wedge Polygons across Tundra Vegetation Types

“…In this exploratory study, we primarily investigated the interoperability of deep learning model predictions across heterogeneous tundra landscapes. Arctic tundra vegetation exhibits a significantly higher degree of heterogeneity over small spatial scales [ 52 ]. Further research is needed to better understand how trained models behave across other tundra vegetation types and regions.…”

“…In future research, we aim to systematically probe further into model interoperability considering multi-faceted factors. Moreover, Arctic tundra landscapes cover spatially isolated ponds, lakes, marshes, river, and stream corridor wetlands, which representing highly heterogeneous features, varying in soil moisture, vegetation composition, elevation, surficial geology, ground ice content, soil thermal regimes and surface hydrology [ 51 , 52 ]. Fine-scale differences in microtopography, limit the ability to comprehend local scale controls on regional to global scale patterns which, is an important factor in characterizing IWPs in Arctic varying tundra areas [ 62 ].…”

“…Although semantically we pursue an abstracted geo object of polygon, its low-level motifs (spectral, spatial, textural characteristics) and high-level meanings are greatly influenced by the landscape where it stands on. Arctic tundra represents a complex and heterogeneous mosaic shaped by the earth processes, markedly influencing vegetation, hydrology, or soil characteristics [ 51 , 52 ]. When investigating how particular targets of interest or geo objects are presented in image modality across a region essentially, we require a baseline fabric that decomposes the heterogeneous system into meaningful patches with underlain ecological functions.…”

Remote Sensing of Tundra Ecosystems Using High Spectral Resolution Reflectance: Opportunities and Challenges

“…Land cover maps with classifications designed for Arctic vegetation types are typically limited in spatial or temporal range (Chasmer et al, 2014;Greaves et al, 2019), precluding comprehensive study of Arctic vegetation dynamics, or are coarse in spatial or temporal resolution (e.g., gridded 1 km CAVM) (Raynolds et al, 2019), precluding accurate characterization of the high level of spatial heterogeneity and temporal variability in Arctic vegetation. Bartsch et al (2016) suggested that a 30 m spatial grain, which is the proposed spatial resolution for SBG, is sufficient for capturing many of the dynamics of Arctic land cover.…”

Reducing model uncertainty of climate change impacts on high latitude carbon assimilation

“…poor data coverage outside of North America (Neigh et al, 2013;Nelson et al, 2009); (2) inconsistent coverage by key global imaging systems, including imaging spectroscopy (IS), thermal infrared (TIR), Lidar, and synthetic aperture radar (SAR)-as they are not available, have poor spatial coverage, coarse resolution, or have not regularly covered northern high latitudes; and (3) the fact that the low species diversity in the ABR is offset by high sub-pixel heterogeneity (Greaves et al, 2019;Riihimäki et al, 2019;Yang, Meng, et al, 2020).…”