diff --git a/call_efast.py b/call_efast.py
index a67e10d..a170a9a 100644
--- a/call_efast.py
+++ b/call_efast.py
@@ -100,6 +100,7 @@ def prepare_s2(season, site_position, site_name, cleaning_strategy="aggressive",
 
         temp_normalized = s2_output_dir / f"temp_{s2_file.name}"
         with rasterio.open(s2_file) as src:
+            # S2 L2A: use DN/10000 (Earth Search AWS differs from ESA spec; offset caused many transparent pixels)
             data = src.read().astype("float32") / 10000.0
             profile = src.profile.copy()
             profile.update({"dtype": "float32", "nodata": 0})
diff --git a/post_process.py b/post_process.py
index f2bcc43..1e1fb48 100644
--- a/post_process.py
+++ b/post_process.py
@@ -31,8 +31,13 @@ def process_cropped(season, site_position, site_name, cleaning_strategy="aggress
             valid = ~np.isnan(data) & (data > 0.001)
             rows = np.any(valid, axis=(0, 2))
             cols = np.any(valid, axis=(0, 1))
-            r0, r1 = np.where(rows)[0][[0, -1]]
-            c0, c1 = np.where(cols)[0][[0, -1]]
+            row_idx = np.where(rows)[0]
+            col_idx = np.where(cols)[0]
+            if len(row_idx) == 0 or len(col_idx) == 0:
+                print(f"[PROCESS] Skipping {fusion_file.name} (no valid pixels)")
+                continue
+            r0, r1 = row_idx[0], row_idx[-1]
+            c0, c1 = col_idx[0], col_idx[-1]
             w, h = c1 - c0 + 1, r1 - r0 + 1
             window = windows.Window(c0, r0, w, h)
             data_crop = src.read(window=window)
diff --git a/run.py b/run.py
index 6c95225..3bd7116 100644
--- a/run.py
+++ b/run.py
@@ -19,12 +19,12 @@ def run_pipeline(season, site_position, site_name):
     """Run pipeline from existing raw data (no downloads)."""
     try:
         # Download steps (skipped - use existing data in data/{site_name}/{season}/raw/)
-        download_s2(season, site_position, site_name)
+        #download_s2(season, site_position, site_name)
         # download_s3(season, site_position, site_name)
         # download_phenocam(season, site_position, site_name)
         # download_phenocam_greenness(season, site_position, site_name)
 
-        print(f"Generating NDVI for raw data: {site_name}, {season}")
+        #print(f"Generating NDVI for raw data: {site_name}, {season}")
         #create_ndvi_timeseries_raw(season, site_position, site_name)
 
         print(f"Running EFAST fusion for all scenarios: {site_name}, {season}")