Mapping Arctic mountain microclimates using satellite land surface temperature: insights from the Kevo valley, northern Finland

Arctic mountainous environments show pronounced spatial and temporal variability in near-surface air temperature (T_air), driven by complex terrain, frequent temperature inversions, seasonal snow cover, and strong seasonal contrasts in solar radiation. Local atmospheric and surface processes, such as cold-air pooling, can cause rapid temperature changes over short distances and timescales. These dynamics are important for understanding Arctic ecosystem change and climate sensitivity, but remain difficult to quantify using sparse in situ temperature observations alone. Satellite-derived land surface temperature (LST) provides spatially continuous information on surface thermal conditions and has increasingly been explored as a proxy for T_air. However, LST-T_air relationships in Arctic mountain environments are highly variable, complicating the application of satellite LST for characterising fine-scale T_air patterns.

 

This study uses a unique in situ T_air dataset from the Kevo valley in northern Finland (26.88–27.05°E, 69.72–69.78°N), which is characterised by strong topographic shading, seasonal snow cover and frequent temperature inversions, and is subjected to the polar night and continuous summer daylight. The dataset comprises 65 stations spanning elevations from 74 to 330 m and recording hourly T_air since 2007. These observations are used to evaluate satellite‑derived LST and to develop models for mapping local T_air using Landsat LST combined with terrain and surface variables, including elevation, slope orientation, snow cover and vegetation indices. We analyse higher spatial resolution LST from Landsat sensors together with coarser resolution LST from MODIS Terra/Aqua and Sentinel-3 SLSTR, examining how terrain, snow cover and surface properties influence LST-T_air relationships and the ability of different LST products to represent microclimate variability across the valley. A focused case study examines high-resolution thermal patterns during nighttime and polar-night conditions using Landsat 8/9 LST acquired from October 2024 to August 2025. Preliminary results indicate that strong apparent LST-T_air agreement is largely driven by the seasonal cycle, with correlations in MODIS LST decreasing from ~0.95 to ~0.74 after deseasonalisation. For Landsat, performance is highly sensitive to data quality, with good‑quality data aligning closely with T_air and poorer‑quality data producing large scatter and a cold bias.