Data used in hurdle modeling of spruce colonization of tundra in northwest Arctic Alaska

<b>Data and Methods Description</b>This dataset supports the manuscript <i>“Conifer colonization of Arctic tundra depends on greening and distance beyond treeline”</i> (Dial et al. 2025, submitted to <i>Ecology Letters</i>). The data combine extensive field observations of white spruce (<i>Picea glauca</i>) colonists with multi-decadal satellite-derived vegetation trends across northwestern Arctic Alaska.<b>Field Data</b>Between June and August 2021, we applied a high-resolution field-sampling protocol known as the <b>pixel-walking method</b> (Wong et al. 2024 <i>Global Change Biology</i>).<b>Sampling design:</b> 29 strip transects (“pixel-plots”) ranging from 30 to 60 m wide and totaling 310.6 km in length (1,371 ha in area).<b>Location:</b> Western Brooks Range, Alaska (153.0–163.2° W; 67.4–68.3° N), across six spatial blocks representing different watersheds and elevation zones.<b>Variables recorded in dataset:</b>pixel_plot = sample strip-transect (either 30m or 60m wide) plot identifier as mm_dd in 2021block = six-level spatial block to address spatial autocorrelation in mixed-effects modelingarea_ha = area of pixel-plot in hectaresColonist_count = number of white spruce (<i>Picea glauca) </i>colonists observed on pixel-plot. Colonists are located &gt;0.5 km from nearest treeline.Colonists_01 = occurrence of white spruce colonists in pixel-plot (0 = none-observed, 1 = one or more observed)Block = two-level block as Arctic alpine or Low ArcticKmeans = classification of pixel plot based on k = 3 clusters using three indicator taxa whose total overall relative abundance was between 5% and 10%: dry = dryas (as in <i>Dryas</i> spp. of dwarf shrubs), eri = evergreen ericaceous shrubs, tus = tussock (sedge <i>Eriophorum vaginatum</i>)southmost_lat = southernmost laitude of pixel-plotdist_to_treeline_km = distance beyond treeline as straightline distance to nearest treeline published in Dial et al. 2022 and Maher et al. 2020NIRV_decade = mean Landsat (30m) 2000-2021 pixel-wise NIRV Thiel-Sen trend (per decade) within pixel-plot found as area-weighted valueEVI2_decade = mean Landsat (30m) 2000-2021 pixel-wise EVI2 Thiel-Sen trend (per decade) within pixel-plot found as area-weighted valueNDVI_decade = mean Landsat (30m) 2000-2021 pixel-wise NDVI Thiel-Sen trend (per decade) within pixel-plot found as area-weighted valuemin_m_asl = minimum elevation sampled along pixel-plot midpoint for 60m wide transects and along edge of 30m wide transectsavg_m_asl = mean elevation sampled along pixel-plot midpoint for 60m wide transects and along edge of 30m wide transectsmax_m_asl = maximum elevation sampled along pixel-plot midpoint for 60m wide transects and along edge of 30m wide transectsavg_Lon mean longitude along pixel-plot midpoint for 60m wide transects and along edge of 30m wide transectspixels = number of pixels with real-valued trends of NIRV, EVI2 and NDVI in the pixel-plotnirv_slope_dec = mean pixel-wise decadal Thiel-Sen trend of NIRV from pixels with real-valued trends of NIRV, EVI2 and NDVI in the pixel-plotevi2_slope_dec = mean pixel-wise decadal Thiel-Sen trend of EVI2 from pixels with real-valued trends of NIRV, EVI2 and NDVI in the pixel-plotndvi_slope_dec = mean pixel-wise decadal Thiel-Sen trend of NDVI from pixels with real-valued trends of NIRV, EVI2 and NDVI in the pixel-plotNDVI_2000_est = mean pixel-wise decadal Thiel-Sen intercept of NDVI from pixels with real-valued trends of NIRV, EVI2 and NDVI in the pixel-plotNIRV_2000_est = mean pixel-wise decadal Thiel-Sen intercept of NIRV from pixels with real-valued trends of NIRV, EVI2 and NDVI in the pixel-plotEVI2_2000_est = mean pixel-wise decadal Thiel-Sen intercept of EVI2 from pixels with real-valued trends of NIRV, EVI2 and NDVI in the pixel-plot<b>Remote-Sensing Data</b>We extracted <b>Landsat surface reflectance (2000–2021)</b> from Google Earth Engine and processed it using the <b>LandsatTS R package</b> (Berner et al. 2023 <i>Ecography</i>).Three spectral vegetation indices (SVIs) were calculated per pixel: NDVI, NIRV, and EVI2.Cross-sensor calibration, atmospheric correction, and phenological modeling were applied to estimate annual peak greenness for each pixel.We retained 17,876 pixels with valid SVI trends across all three indices.Each pixel was assigned to its corresponding field pixel-plot polygon, producing a dataset that links ground observations with remotely sensed vegetation trends.<b>Analytical Methods</b>We modeled white spruce colonization using <b>spatial fixed-effects hurdle models</b> implemented in <i>glmmTMB</i> (R 4.3). This dataset is the one used for the modeling.The <b>binomial component</b> modeled colonist occurrence (presence–absence) as a function of SVI trend (proxy for growing-condition improvement).The <b>truncated count component</b> modeled non-zero colonist abundance as a function of distance-beyond-treeline (proxy for dispersal limitation).Spatial autocorrelation was accounted for using random intercepts by block.Each row is a pixel-plot level set of observations:Field observations (colonist counts, vegetation composition, GNSS coordinates as midpoints of extremes of pixel-plots).Pixel-wise SVI trends for NDVI, NIRV, and EVI2.