Field Notes: "5,976 km to the Arctic in an EV, towing a caravan" Curated data tables, Martijn Hermans, 2026. What this is Aggregate tables behind the article's figures. They come from the car's own logs (Teslamate), 3 to 20 September 2026. Positions, timestamps finer than a day, battery logs and any spending are deliberately left out. Files segments.csv One row per one-kilometre stretch of road inside a drive (2,393 rows; 2,208 used in the models). segment id: drive number and stretch index drive drive number (as logged) km stretch length, km (odometer) speed_kmh mean speed over the stretch kwh_per_100km battery energy per 100 km: battery power (about 4 readings per second) added up over the stretch (measured) temp_c outside temperature reported by the car, degrees C (measured) grade_pct net slope over the stretch from the car's elevation readings, percent headwind_ms along-road wind at about 1.5 m height, m/s, from Open-Meteo's archived weather model (modeled; negative = tailwind) wet_road_modeled 1 if the weather model had rain in the preceding hours (modeled, not observed) caravan tow / solo, from the owner's trip diary label_confidence confirmed / probable / uncertain used_in_models 0 for stretches with a data gap over 10 seconds or outside the analysis rules drives.csv One row per analysed drive longer than 5 km (50 rows). kwh_per_100km_rated_range_proxy the car's rated-range drop times 0.152 kWh/km, for comparison (reads 4 to 5% above measured) climb_m_per_km metres climbed per km excluded_from_main_models 1 for the one drive with a stale GPS fix after a ferry days.csv Kilometres per local day, with coarse places from the trip diary. model_segments.csv Everything the models use, one row per one-kilometre stretch (2,393 rows). drive, date_local, hour_local drive number, local date and hour the stretch started northern_norway 1 if the drive's mean latitude is above 65 N (used for one sensitivity check) seg stretch index within the drive (0 = first kilometre) dur_s, dist_m duration (s) and distance (m, odometer) e_kwh, e_pos_kwh, e_regen_kwh battery energy: net, drawn, and regenerated (negative), kWh (measured) S2, S3 integral of v ds and v^2 ds over the stretch (v in m/s), for the drag term v_mean_kmh, v0, v1 mean speed (km/h); speed at start and end (m/s) max_gap_s, moving_frac longest gap between power readings (s); share of time moving gps_stale True where the car held a wrong position (after the Kiel ferry) car_temp, climate_on outside temperature from the car (C); climate control on (share) soc0, soc1 battery percentage at start and end label, confidence caravan tow/solo from the trip notes, and its confidence dh_car, dh_dem, grade_car net climb (m) from the car's altitude and from an elevation model; grade (fraction) elev_suspect True where the two elevation sources disagree badly wx_T, rho weather-model temperature (C); air density at road level (kg/m3) (modeled) headwind10 along-road wind at 10 m (m/s, modeled; negative = tailwind) headwind_base, _smooth, _rough the same scaled to 1.5 m for terrain roughness 0.1, 0.03 and 0.5 m wet, wet_strict, wet_loose modeled wet-road flags under three rain rules reproduce.py Recomputes the trip averages, the same-speed comparison and regressions M0 to M5 (drive-clustered errors) from these files. Python 3 and numpy only. Output matches the article. Not included, and why - Positions, direction of travel, absolute elevation and exact times. They would show where we slept and where family lives. model_segments.csv keeps every model input without them, so the road-load fit can be redone from the equation and fixed values on the method page. - 3,474 km driven before logging started have no data. Units: km, km/h, kWh per 100 km, degrees C, m/s, percent.