Danika Fay Wellington
Suming Jin
Heather J. Tollerud
Jesslyn F. Brown
Jon Dewitz
Neal J. Pastick
Christopher P. Barber
Austin O'Brien
Mark Spanier
Rylie Fleckenstein
2026
<div id="sp0075" class="u-margin-s-bottom">Land cover information is essential for understanding Earth’s surface dynamics and how vegetation, water, soil, climate, and terrain interact. The National Land Cover Database (NLCD) has been the authoritative source for consistent U.S. land cover mapping. To extend NLCD’s temporal resolution and reduce production latency, we developed the Land Cover Artificial Mapping System (LCAMS)—a prototype spatiotemporal deep learning framework piloted as the foundation for the new Annual NLCD.</div><div class="u-margin-s-bottom"><br data-mce-bogus="1"></div><div id="sp0080" class="u-margin-s-bottom">LCAMS builds on concepts from legacy NLCD and the U.S. Geological Survey Land Change Monitoring, Assessment, and Projection (LCMAP) initiatives. It employs a loosely coupled two-stage architecture consisting of independent but functionally interdependent spatial and temporal models. Spatial models extract per-year information from Landsat data, while the temporal models refine the spatial outputs to enforce inter-annual consistency—critical for reliable land change monitoring. LCAMS produces annual 30 m resolution land cover and impervious surface outputs, with region-specific fine-tuning to generalize across diverse landscapes and temporal dynamics.</div><div class="u-margin-s-bottom"><br data-mce-bogus="1"></div><div id="sp0085" class="u-margin-s-bottom">Validation was conducted using an independent dataset of 1925 randomly sampled plots from five U.S. Landsat Analysis Ready Data (ARD) tiles spanning 1985-2021, selected for spatial and temporal variability. This dataset was used consistently to evaluate LCAMS, Legacy NLCD, and LCMAP. Using the NLCD legend, LCAMS achieved<span> 72.1 ± 1.60%</span><span class="math"><span id="MathJax-Element-1-Frame" class="MathJax_SVG" data-mathml="<math xmlns="http://www.w3.org/1998/Math/MathML"><mn is="true">72.1</mn><mo linebreak="goodbreak" is="true">&#xB1;</mo><mn is="true">1.60</mn><mi mathvariant="normal" is="true">%</mi></math>"></span></span><span> </span>overall agreement, compared to<span> 71.1 ± 1.7%</span><span class="math"><span id="MathJax-Element-2-Frame" class="MathJax_SVG" data-mathml="<math xmlns="http://www.w3.org/1998/Math/MathML"><mn is="true">71.1</mn><mo linebreak="goodbreak" is="true">&#xB1;</mo><mn is="true">1.7</mn><mi mathvariant="normal" is="true">%</mi></math>"></span></span><span> </span>agreement for Legacy NLCD. Using the LCMAP legend, LCAMS achieved<span> 83.4 ±</span><span class="math"><span id="MathJax-Element-3-Frame" class="MathJax_SVG" data-mathml="<math xmlns="http://www.w3.org/1998/Math/MathML"><mn is="true">83.4</mn><mo linebreak="goodbreak" is="true">&#xB1;</mo><mn is="true">1.22</mn><mi mathvariant="normal" is="true">%</mi></math>"></span></span><span> 1.22% </span>agreement, compared to 84.6<span> ±</span><span class="math"><span id="MathJax-Element-4-Frame" class="MathJax_SVG" data-mathml="<math xmlns="http://www.w3.org/1998/Math/MathML"><mn is="true">84.6</mn><mo linebreak="goodbreak" is="true">&#xB1;</mo><mn is="true">1.11</mn><mi mathvariant="normal" is="true">%</mi></math>"></span></span><span> 1.11% </span>agreement for LCMAP. Overall, LCAMS delivers comparable accuracy while offering higher thematic resolution, longer temporal coverage, and automated production of annual 30 m CONUS land cover.</div>
application/pdf
10.1016/j.rse.2026.115347
en
Elsevier
A framework for integrating spatiotemporal deep learning methods with landsat for annual land cover and impervious surface mapping
article