42 Years of NASA POWER Trends Across Brazilian Biomes: Seasonal Mann-Kendall + Theil-Sen Attribution of Surface Solar Brightening (1984–2025)
1984-01 to 2025-12, n = 504 months per series), Cockpit-OMNI evaluates seven Earth system variables across three contrasting Brazilian biomes using Seasonal Mann-Kendall (period = 12, α = 0.05) and deseasonalized Theil-Sen regression (95% CI). All three stations warmed significantly (Campinas-SP +0.4091 °C/dec, Petrolina-PE +0.3012 °C/dec, Manaus-AM +0.3065 °C/dec) and lost rainfall, while surface shortwave solar irradiance increased everywhere. Crucially, the exact same statistical engine delivers opposite physical causal verdicts: solar brightening in Petrolina-PE (-0.9647 %/dec cloud attenuation, p = 2.833e-09; clear-sky null p = 0.2901) and Manaus-AM (-0.8552 %/dec cloud attenuation, p = 3.015e-08; clear-sky null p = 0.7686) is cloud-driven, whereas Campinas-SP exhibits stationary clouds (p = 0.9476) alongside surging clear-sky irradiance (+0.1897 MJ/m²/day/dec, p < 1e-16), isolating an aerosol / clear-sky atmospheric transparency shift. Monthly 2m wind speed (WS2M) shows zero significant trend at any station (p = 0.4890 / 0.1734 / 0.4512) and is explicitly barred from sub-hourly ramp claims.
1. How does Cockpit-OMNI compute non-parametric climate trends without seasonal distortion?
Direct Answer: Cockpit-OMNI computes trends deterministically in both Python (pymannkendall + scipy.stats.theilslopes) and pure browser JavaScript (frontend/src/engine/trend.js) by comparing matching calendar months across 42 years (Seasonal Mann-Kendall, period = 12), removing NASA POWER annual summary rows (YYYY13) and -999.0 sentinel values prior to estimating median deseasonalized Theil-Sen slopes.
Ordinary least-squares regression fails on multi-decade climatology because single El Niño spikes skew the slope and regular summer-winter oscillations mask secular change. For example, in Campinas-SP all-sky irradiance (ALLSKY_SFC_SW_DWN), an ordinary un-seasonalized Mann-Kendall test produces p = 0.1968 (falsely labeling the series "no trend"), whereas Seasonal Mann-Kendall isolates the true decadal signal at p = 0.0008616 (τ = +0.103, +0.1621 MJ/m²/day/dec).
2. Complete 21-Metric Empirical Benchmark (NASA POWER, 1984–2025, n = 504 Months)
Direct Answer: Every metric below is stored in the open zero-key cache of www.curveofduck.xyz/cockpit and verified automatically by scripts/verify_answers.mjs against the frozen 2026-11-14T15:04 UTC snapshot.
2.1 Campinas-SP (-22.90, -47.06) — Atlantic Forest / Cerrado Transition (Aerosol / Clear-Sky Driven)
| Variable | Unit | Slope / Decade | 95% Confidence Interval | 42-yr Total | Seasonal τ | Seasonal p-value | Verdict (α = 0.05) |
|---|---|---|---|---|---|---|---|
T2M | °C | +0.4091 | [+0.3329, +0.4891] | +1.718 | +0.293 | < 1e-16 | Increasing |
PRECTOTCORR | mm/day | -0.2724 | [-0.3881, -0.1559] | -1.144 | -0.149 | 1.59e-06 | Decreasing |
RH2M | % | -1.6601 | [-1.9845, -1.3400] | -6.972 | -0.296 | < 1e-16 | Decreasing |
ALLSKY_SFC_SW_DWN | MJ/m²/day | +0.1621 | [+0.0591, +0.2667] | +0.681 | +0.103 | 0.0008616 | Increasing |
CLRSKY_SFC_SW_DWN | MJ/m²/day | +0.1897 | [+0.1577, +0.2232] | +0.797 | +0.322 | < 1e-16 | Increasing |
CLOUD_ATTEN_PCT | % | +0.0378 | [-0.3877, +0.4698] | +0.159 | +0.002 | 0.9476 | Null (no trend) |
WS2M | m/s | -0.0042 | [-0.0191, +0.0103] | -0.018 | -0.021 | 0.4890 | Null (no trend) |
2.2 Petrolina-PE (-9.39, -40.50) — Caatinga Semi-Arid Solar Corridor (Cloud-Driven)
| Variable | Unit | Slope / Decade | 95% Confidence Interval | 42-yr Total | Seasonal τ | Seasonal p-value | Verdict (α = 0.05) |
|---|---|---|---|---|---|---|---|
T2M | °C | +0.3012 | [+0.2314, +0.3710] | +1.265 | +0.242 | 4.663e-15 | Increasing |
PRECTOTCORR | mm/day | -0.0721 | [-0.1265, -0.0250] | -0.303 | -0.068 | 0.0291 | Decreasing |
RH2M | % | -0.8932 | [-1.2997, -0.4790] | -3.751 | -0.125 | 5.157e-05 | Decreasing |
ALLSKY_SFC_SW_DWN | MJ/m²/day | +0.2558 | [+0.1624, +0.3496] | +1.075 | +0.178 | 9.384e-09 | Increasing |
CLRSKY_SFC_SW_DWN | MJ/m²/day | +0.0127 | [-0.0093, +0.0350] | +0.053 | +0.033 | 0.2901 | Null (no trend) |
CLOUD_ATTEN_PCT | % | -0.9647 | [-1.2794, -0.6302] | -4.052 | -0.184 | 2.833e-09 | Decreasing |
WS2M | m/s | -0.0138 | [-0.0433, +0.0168] | -0.058 | -0.042 | 0.1734 | Null (no trend) |
2.3 Manaus-AM (-3.10, -60.02) — Central Amazon Rainforest (Cloud-Driven)
| Variable | Unit | Slope / Decade | 95% Confidence Interval | 42-yr Total | Seasonal τ | Seasonal p-value | Verdict (α = 0.05) |
|---|---|---|---|---|---|---|---|
T2M | °C | +0.3065 | [+0.2543, +0.3593] | +1.287 | +0.381 | < 1e-16 | Increasing |
PRECTOTCORR | mm/day | -0.4673 | [-0.5989, -0.3344] | -1.963 | -0.204 | 4.238e-11 | Decreasing |
RH2M | % | -0.5149 | [-0.7309, -0.3099] | -2.163 | -0.133 | 1.645e-05 | Decreasing |
ALLSKY_SFC_SW_DWN | MJ/m²/day | +0.2413 | [+0.1690, +0.3120] | +1.013 | +0.198 | 1.738e-10 | Increasing |
CLRSKY_SFC_SW_DWN | MJ/m²/day | +0.0289 | [0.0000, +0.0582] | +0.121 | +0.009 | 0.7686 | Null (no trend) |
CLOUD_ATTEN_PCT | % | -0.8552 | [-1.1442, -0.5651] | -3.592 | -0.172 | 3.015e-08 | Decreasing |
WS2M | m/s | -0.0007 | [-0.0034, +0.0013] | -0.003 | -0.023 | 0.4512 | Null (no trend) |
3. Multi-Vector Attribution & Verified NASA Dataset Provenance
Direct Answer: Cockpit-OMNI pairs NASA POWER decadal trends with near-real-time thermal fire anomalies (NASA FIRMS VIIRS_NOAA21_NRT, CMR Concept ID C3264430167-LANCEMODIS), space weather alerts (NASA DONKI), and cataloged InSAR/thermal surface deformation products (NASA Earthdata CMR MCP: NISAR_L2_GUNW_PROVISIONAL_V1 and ECO_L2T_LSTE, honestly labeled roadmap).
- /cache/attribution_campinas.json · /cache/attribution_petrolina.json · /cache/attribution_manaus.json
- /cache/cmr_evidence.json (Earthdata CMR MCP protocol
2025-06-18) - Interactive workbench: Launch Earth System Cockpit or read Monograph B on Northeast Solar Brightening.