diff --git a/src/climatevision/data/README.md b/src/climatevision/data/README.md
new file mode 100644
index 0000000..ee3ea52
--- /dev/null
+++ b/src/climatevision/data/README.md
@@ -0,0 +1,46 @@
+# Data Pipeline
+
+Sentinel-2 ingestion, band mapping, and preprocessing for ClimateVision.
+
+## Modules
+
+| File | Purpose |
+|------|---------|
+| `gee_downloader.py` | Download real Sentinel-2 tiles from Google Earth Engine for a given bbox + date range. Falls back to a labelled synthetic tile (`is_synthetic: true`) when GEE credentials are missing. |
+| `band_mapping.py` | Single source of truth for which spectral bands each analysis type requires. Reads from `config.yaml`. |
+| `preprocessing.py` | Cloud masking (SCL band), normalisation, resampling 20m bands to 10m, tiling to 256×256. |
+| `transforms.py` | Augmentation pipeline (flips, rotations, spectral jitter) for training DataLoaders. |
+| `sampling.py` | Tile sampling strategies (random, balanced, stratified by region). |
+| `quality.py` | Per-tile QA (cloud %, NaN ratio, band coverage). |
+| `validation.py` | Schema validation for incoming requests and downloaded tiles. |
+
+## Analysis-Type Band Contract
+
+Every analysis type has its own band list in `config.yaml`. The pipeline must use `get_bands_for_analysis(analysis_type)` — never hardcode band lists.
+
+| Analysis | Bands | Channels |
+|----------|-------|----------|
+| `deforestation` | B04, B03, B02, B08 | 4 |
+| `ice_melting` | B02, B03, B04, B11 | 4 |
+| `flooding` | B03, B08, B11 | 3 |
+
+## Cloud Masking
+
+`apply_scl_cloud_mask(image, scl_band)` zeroes out pixels classified as cloud, shadow, snow/ice, or no-data using the Sentinel-2 Scene Classification Layer (SCL). This must run **before** the model forward pass.
+
+Valid SCL classes kept: 4 (vegetation), 5 (bare soil), 6 (water), 7 (low cloud), 10 (thin cirrus).
+Masked out: 0 (no-data), 1 (saturated), 2 (dark), 3 (shadow), 8/9 (medium/high cloud), 11 (snow/ice).
+
+## Synthetic Fallback
+
+If GEE auth fails, the downloader returns a deterministic synthetic tile seeded by the bbox so the same region always yields the same fallback. The metadata always includes `is_synthetic: true` so the API can warn the caller.
+
+## Environment
+
+```
+GEE_PROJECT_ID=your-project-id
+GEE_SERVICE_ACCOUNT=svc@project.iam.gserviceaccount.com
+GEE_SERVICE_ACCOUNT_KEY=secrets/gee-key.json
+```
+
+Run `python scripts/setup_gee.py` to verify credentials.
diff --git a/team_docs/Adeolu_Mary_Oshadare_Role.pdf b/team_docs/Adeolu_Mary_Oshadare_Role.pdf
new file mode 100644
index 0000000..51b940e
Binary files /dev/null and b/team_docs/Adeolu_Mary_Oshadare_Role.pdf differ
diff --git a/tests/test_band_mapping.py b/tests/test_band_mapping.py
new file mode 100644
index 0000000..4f5832a
--- /dev/null
+++ b/tests/test_band_mapping.py
@@ -0,0 +1,83 @@
+"""Smoke tests for analysis-aware Sentinel-2 band mapping."""
+from __future__ import annotations
+
+import pytest
+
+from climatevision.data.band_mapping import (
+    SCL_BAND,
+    SENTINEL2_BAND_ORDER,
+    get_band_indices,
+    get_bands_for_analysis,
+    get_bands_for_analysis_with_scl,
+    get_model_config,
+    is_analysis_enabled,
+    list_enabled_analysis_types,
+)
+
+
+def test_sentinel2_band_order_has_13_bands():
+    assert len(SENTINEL2_BAND_ORDER) == 13
+    assert SENTINEL2_BAND_ORDER[0] == "B01"
+    assert SENTINEL2_BAND_ORDER[-1] == "B12"
+
+
+def test_deforestation_uses_four_bands():
+    bands = get_bands_for_analysis("deforestation")
+    assert len(bands) == 4
+    assert set(bands) == {"B02", "B03", "B04", "B08"}
+
+
+def test_flooding_uses_three_bands():
+    bands = get_bands_for_analysis("flooding")
+    assert len(bands) == 3
+    assert "B11" in bands
+
+
+def test_ice_melting_uses_swir():
+    bands = get_bands_for_analysis("ice_melting")
+    assert "B11" in bands
+
+
+def test_scl_appended_for_cloud_masking():
+    bands = get_bands_for_analysis_with_scl("deforestation")
+    assert SCL_BAND in bands
+    assert bands[-1] == SCL_BAND
+
+
+def test_scl_not_duplicated():
+    bands_with_scl = get_bands_for_analysis_with_scl("deforestation")
+    bands_again = get_bands_for_analysis_with_scl("deforestation")
+    assert bands_with_scl.count(SCL_BAND) == 1
+    assert bands_again.count(SCL_BAND) == 1
+
+
+def test_band_indices_resolve_correctly():
+    indices = get_band_indices(["B04", "B03", "B02", "B08"])
+    assert indices == [3, 2, 1, 7]
+
+
+def test_band_indices_rejects_unknown():
+    with pytest.raises(ValueError, match="Unknown"):
+        get_band_indices(["B99"])
+
+
+def test_band_indices_rejects_scl_directly():
+    with pytest.raises(ValueError, match="SCL"):
+        get_band_indices([SCL_BAND])
+
+
+def test_enabled_analysis_types_include_active_three():
+    enabled = list_enabled_analysis_types()
+    for name in ("deforestation", "ice_melting", "flooding"):
+        assert name in enabled, f"{name} should be enabled"
+
+
+def test_disabled_analysis_types():
+    assert not is_analysis_enabled("drought")
+    assert not is_analysis_enabled("wildfire")
+
+
+def test_model_config_carries_channels_and_classes():
+    cfg = get_model_config("flooding")
+    assert cfg["in_channels"] == 3
+    assert cfg["num_classes"] == 3