diff --git a/lidar_pipeline/rendering.py b/lidar_pipeline/rendering.py
index 86c07b8..b27146a 100644
--- a/lidar_pipeline/rendering.py
+++ b/lidar_pipeline/rendering.py
@@ -189,7 +189,8 @@ for _key, _info in RGB_LEGENDS.items():
 def _apply_colormap(data, tif_file, resolution=None):
     """Apply the registered colormap normalization to data based on filename.
 
-    Returns (data, cmap, title, legend_label, description, is_rgb).
+    Returns (data, cmap, title, legend_label, description, is_rgb, vmin, vmax)
+    où vmin/vmax sont les bornes physiques de la plage rendue (None si sans objet).
     """
     name = str(tif_file).lower()
 
@@ -197,7 +198,7 @@ def _apply_colormap(data, tif_file, resolution=None):
     for key in RGB_LEGENDS:
         if key in name:
             info = RGB_LEGENDS[key]
-            return data, None, info['title'], info['legend'], info['description'], True
+            return data, None, info['title'], info['legend'], info['description'], True, None, None
 
     # Find matching colormap — sort by key length descending so 'mslrm' matches before 'lrm'
     for key in sorted(COLORMAPS.keys(), key=len, reverse=True):
@@ -208,7 +209,7 @@ def _apply_colormap(data, tif_file, resolution=None):
 
             if len(valid_data) == 0:
                 logger.warning(f"    Aucune donnée valide pour {Path(tif_file).name} — colormap ignorée")
-                return data, 'terrain', Path(tif_file).stem.replace('_', ' ').title(), '', '', False
+                return data, 'terrain', Path(tif_file).stem.replace('_', ' ').title(), '', '', False, None, None
 
             vmin = vmax = None
 
@@ -253,7 +254,7 @@ def _apply_colormap(data, tif_file, resolution=None):
                     data = np.clip((data - vmin) / max(vmax - vmin, 0.001), 0, 1)
 
             legend = info['legend'].format(vmin=vmin or 0, vmax=vmax or 0)
-            return data, info['cmap'], info['title'], legend, info['description'], False
+            return data, info['cmap'], info['title'], legend, info['description'], False, vmin, vmax
 
     # Default: terrain colormap with percentile stretch
     valid_data = np.asarray(data.compressed() if hasattr(data, 'compressed') else data.flatten())
@@ -265,7 +266,7 @@ def _apply_colormap(data, tif_file, resolution=None):
     span = max(p98 - p2, 1e-6)
     data = np.clip((data - p2) / span, 0, 1)
     title = Path(tif_file).stem.replace('_', ' ').title()
-    return data, 'terrain', title, 'Altitude normalisée', '', False
+    return data, 'terrain', title, 'Altitude normalisée', '', False, p2, p98
 
 
 def _download_location_map(min_x, max_x, min_y, max_y):
@@ -453,7 +454,7 @@ def tif_to_png(tif_file, vis_dir, resolution, keep_tif=False, source_info=None,
                     nan_mask = nan_mask  # keep for later
 
             # Apply colormap
-            data, cmap, title, legend_label, description, is_rgb_result = _apply_colormap(data, tif_file, resolution=resolution)
+            data, cmap, title, legend_label, description, is_rgb_result, cmap_vmin, cmap_vmax = _apply_colormap(data, tif_file, resolution=resolution)
 
             # Apply NaN mask: make zones without data transparent
             has_nan_mask = nan_mask is not None and not is_rgb_result
@@ -461,8 +462,12 @@ def tif_to_png(tif_file, vis_dir, resolution, keep_tif=False, source_info=None,
                 # data is normalized 0-1 from _apply_colormap; apply cmap to get RGBA
                 # Save the colormap for colorbar before converting to RGBA
                 saved_cmap = plt.get_cmap(cmap) if isinstance(cmap, str) else cmap
-                saved_vmin = float(np.nanmin(data)) if not nan_mask.all() else 0
-                saved_vmax = float(np.nanmax(data)) if not nan_mask.all() else 1
+                # Bornes physiques de la colorbar (unités réelles, pas 0-1)
+                if cmap_vmin is not None and cmap_vmax is not None:
+                    saved_vmin, saved_vmax = float(cmap_vmin), float(cmap_vmax)
+                else:
+                    saved_vmin = float(np.nanmin(data)) if not nan_mask.all() else 0
+                    saved_vmax = float(np.nanmax(data)) if not nan_mask.all() else 1
                 rgba = saved_cmap(data)  # (H, W, 4) float RGBA
                 rgba[nan_mask, 3] = 0.0  # transparent where no data
                 data = rgba
@@ -806,9 +811,14 @@ def tif_to_crop(tif_file, vis_dir, resolution, keep_tif=False, quality=98, outpu
                 data = np.moveaxis(data, 0, -1)
             else:
                 data = src.read(1)
+                # Nodata → NaN : sinon les pixels bord (ex. -9999, 3.4e38)
+                # polluent les percentiles d'étalonnage de la colormap
+                if src.nodata is not None:
+                    data = data.astype(np.float32, copy=False)
+                    data[data == src.nodata] = np.nan
 
         # Apply colormap normalization
-        data, cmap_name, title, legend_label, description, is_rgb_result = _apply_colormap(data, tif_file, resolution=resolution)
+        data, cmap_name, title, legend_label, description, is_rgb_result, _cvmin, _cvmax = _apply_colormap(data, tif_file, resolution=resolution)
 
         if not is_rgb_result:
             data = np.where(np.isnan(data), 0.0, data)