Comparative Detective Case Studies (Or Click Any Coordinate on the GIS Map):
1. WHAT IS CHANGING?
T2M (°C)
2 m Air Temperature
Thermodynamic air temperature at 2 metres above ground.
Reanalysis Model vs. Satellite: Not MODIS thermal skin LST.
2. WHERE IS IT CHANGING?
Δ 1998 km
Pala, KL vs. Jaisalmer, RJ
Pala, Kerala:
9.71°N, 76.68°E
Jaisalmer, RJ:
26.92°N, 70.91°E
0.5° × 0.625° MERRA-2 grid footprints (~55 km regional cells).
3. BY HOW MUCH? (SEN'S SLOPE)
95% CI
Pala, KL:
--
Jaisalmer, RJ:
--
Outlier-resistant median of 946 pairwise slopes per decade.
4. IS IT SIGNIFICANT? (α = 0.05)
Mann–Kendall
Pala, KL:
--
Jaisalmer, RJ:
--
Two-sided tie-corrected p-value, Kendall's τ & Z-score.
ZERO-DEPENDENCY HTML5 CANVAS GEOSPATIAL ENGINE
0.5° × 0.625° MERRA-2 Footprints & 3D Orthographic Globe
Click Map Assigns:
Click anywhere to set Region A/B • Drag to pan/rotate • Scroll to zoom
A (Lat/Lon):
B (Lat/Lon):
Region A (0.5°×0.625° MERRA-2 Cell)
Region B (0.5°×0.625° MERRA-2 Cell)
HUMID TROPICAL MONSOON
--
Pala, Kerala
9.71°N, 76.68°E • MERRA-2 GridSen's Slope
--
MK p-value
--
Kendall τ
--
95% Confidence Interval:
--
OLS Linear Slope (vs Sen's):
--
Mann–Kendall S & Z-Score:
--
44-Year Mean (1981–2024):
--
ARID THAR DESERT
--
Jaisalmer, Rajasthan
26.92°N, 70.91°E • MERRA-2 GridSen's Slope
--
MK p-value
--
Kendall τ
--
95% Confidence Interval:
--
OLS Linear Slope (vs Sen's):
--
Mann–Kendall S & Z-Score:
--
44-Year Mean (1981–2024):
--
ARGON AI DETECTIVE
EARTH SYSTEM TREND DETECTIVE • GEOPHYSICAL SYNTHESIS
CHART 1 • 1981–2024 MULTI-DECADAL TRAJECTORY & THEIL–SEN ROBUST TREND
2 m Air Temperature (°C) — Annual Series, Sen's Slope & OLS Comparison
Pala, KL (Sen's Slope Dashed)
Jaisalmer, RJ (Sen's Slope Dashed)
Peak Outlier Year
n = 44 Years (1981–2024)
CHART 2 • CLIMATOLOGICAL ANOMALY & SKIN-AIR DIAGNOSTIC
1981–2010 WMO Baseline Anomaly Departures (Δ)
528/528 Months Valid
COMPLETE MULTI-VARIABLE STATISTICAL INVESTIGATION MATRIX (1981–2024)
Click any row to load directly into the primary workstation
All 5 NASA MERRA-2 Variables Across Annual, SW Monsoon (JJAS) & Pre-Monsoon (MAM) Windows
| Variable | Temporal Window | Region A Sen's Slope [95% CI] | Region A Mann–Kendall | Region B Sen's Slope [95% CI] | Region B Mann–Kendall |
|---|
NASA SPACE APPS SCIENTIFIC VALIDITY & METHODOLOGY REFERENCE
Deep Space Blue (#07173F) on White (#FFFFFF) • WCAG AAA
Non-Interchangeable Variables: Measurements vs. Reanalysis Models vs. Derived Spectral Indices
| Variable / Dataset | Physical Quantity Measured or Modelled | Units | Data Category | Resolution, Missing Data & Scientific Limitations |
|---|---|---|---|---|
| T2M / T2M_MAX (POWER) | Thermodynamic air temperature at 2 metres height | °C | Reanalysis Model (MERRA-2) | 0.5° × 0.625° (~55 km) assimilation grid. Captures regional climate state; not a point thermometer in Pala or a micro-topographic district heatmap. |
| TS (POWER) vs. MODIS LST | Radiative skin temperature of land/canopy surface | °C | Model Skin vs. Satellite | TS is gap-free MERRA-2 model skin temperature. Orbital MODIS thermal LST suffers from heavy cloud-cover dropouts during Kerala's JJAS monsoon. Never conflate skin temperature with 2 m air temperature. |
| PRECTOTCORR (POWER) | Bias-corrected total rain/snow precipitation | mm | Merged Reanalysis-Gauge | Monthly POWER values are served in mm/day; EARTH PULSE multiplies each month by exact calendar days in that month (including leap years) before summing Annual, JJAS, and MAM totals. |
| RH2M (POWER) | Relative humidity at 2 metres height | % | Reanalysis Model (MERRA-2) | Ratio of actual water vapor pressure to saturation vapor pressure. Crucial for contrasting humid tropical heat stress in Kerala against dry sensible heating in Rajasthan. |
| MODIS NDVI & ISRO MOSDAC | Normalized Difference Vegetation Index & Monsoon Validation | Unitless | Derived Index & Partner Context | Red/NIR reflectance ratio proxy for canopy greenness; contextualized alongside Space Agency Partner ISRO Bhuvan / MOSDAC Indian monsoon regional climatology references. |
1. Avoiding Seasonal Autocorrelation
Running a naive t-test on 528 raw monthly records is invalid because seasonal monsoon cycles violate serial independence. Slicing into Annual (12-mo), JJAS (4-mo), and MAM (3-mo) series ($n=44$ years) ensures rigorous Mann–Kendall inference.
2. Theil–Sen vs. Outlier Distortion
Extreme events (such as the 2018 Kerala floods, flagged with a Rocket Red callout marker on Chart 1) pull ordinary least squares (OLS) regression slopes. Theil–Sen computes the median slope across all $\binom{44}{2} = 946$ year pairs.
3. Significance ≠ Attribution
A Mann–Kendall $p\text{-value} < 0.05$ confirms a non-random monotonic trend over 1981–2024. Distinguishing global greenhouse forcing from local irrigation or urbanization requires coupled physical attribution modeling (correlation $\neq$ causation).