diff --git a/forestatrisk/project/allocate_deforestation.py b/forestatrisk/project/allocate_deforestation.py
new file mode 100644
index 0000000..7516c23
--- /dev/null
+++ b/forestatrisk/project/allocate_deforestation.py
@@ -0,0 +1,121 @@
+"""Allocating deforestation."""
+
+import os
+
+from osgeo import gdal
+import pandas as pd
+
+opj = os.path.join
+opd = os.path.dirname
+
+
+def allocate_deforestation(riskmap_juris_file, defor_rate_tab,
+                           defor_juris_ha, years_forecast,
+                           project_boundaries,
+                           output_file="defor_project.csv",
+                           verbose=False):
+    """Allocating deforestation.
+
+    :param riskmap_juris: Raster file with classes of deforestation
+      risk at the jurisdictional level.
+
+    :param defor_rate_tab: CSV file including the table with
+      deforestation rates for each deforestation class.
+
+    :param defor_juris_ha: Expected deforestation at the
+      jurisdictional level (in hectares).
+
+    :param years_forecast: Length of the forecasting period (in years).
+
+    :param project_boundaries: Vector file for project boundaries.
+
+    :param output_file: Output file with deforestation
+      allocated to the project.
+
+    :param verbose: If True, print messages.
+
+    """
+
+    # Callback
+    cback = gdal.TermProgress_nocb if verbose else 0
+
+    # Creation options
+    copts = ["COMPRESS=DEFLATE", "BIGTIFF=YES"]
+
+    # ---------------------------------------
+    # Crop riskmap to project boundaries
+    # ---------------------------------------
+
+    out_dir = opd(output_file)
+    ofile = opj(out_dir, "project_riskmap.tif")
+    gdal.Warp(ofile, riskmap_juris_file,
+              cropToCutline=True,
+              warpOptions=["CUTLINE_ALL_TOUCHED=TRUE"],
+              cutlineDSName=project_boundaries,
+              creationOptions=copts,
+              callback=cback)
+
+    # ---------------------------------------
+    # Compute number of pixels for each class
+    # ---------------------------------------
+
+    nvalues = 65535
+    with gdal.Open(ofile) as ds:
+        band = ds.GetRasterBand(1)
+        counts = band.GetHistogram(0.5, 65535.5, nvalues, 0, 0)
+    data = {
+        "cat": [i + 1 for i in range(65535)],
+        "counts": counts}
+    df_count = pd.DataFrame(data)
+
+    # Upload deforestation rates
+    df_rate = pd.read_csv(defor_rate_tab)
+
+    # -----------------------------
+    # Compute deforestation density
+    # -----------------------------
+
+    # Pixel area
+    pixel_area = df_rate.loc[0, "pixel_area"]
+
+    # Correction factor, either ndefor / sum_i p_i
+    # or theta * nfor / sum_i p_i
+    sum_pi = (df_rate["nfor"] * df_rate["rate_mod"]).sum()
+    correction_factor = defor_juris_ha / (pixel_area * sum_pi)
+
+    # Absolute deforestation rate
+    df_rate["rate_abs"] = df_rate["rate_mod"] * correction_factor
+
+    # Deforestation density (ha/pixel/yr)
+    df_rate["defor_dens"] = df_rate["rate_abs"] * pixel_area / years_forecast
+
+    # -----------------------------
+    # Join tables
+    # -----------------------------
+
+    df_project = df_count.merge(right=df_rate, on="cat", how="left")
+
+    # Annual deforestation (ha) for project
+    defor_project = (df_project["counts"] * df_project["defor_dens"]).sum()
+
+    # Save results
+    data = {
+        "period": ["annual", "entire"],
+        "deforestation (ha)": [round(defor_project, 1), round(defor_project * years_forecast, 1)]
+    }
+    res = pd.DataFrame(data)
+    res.to_csv(output_file, header=True, index=False)
+
+
+# Test
+import os
+os.chdir(os.path.expanduser("~/deforisk/MTQ-tuto-bak/outputs/far_models/forecast/"))
+allocate_deforestation(
+    riskmap_juris_file="prob_icar_t3.tif",
+    defor_rate_tab="defrate_cat_icar_forecast.csv",
+    defor_juris_ha=4000,  # About 400 ha/yr in MTQ on 2010-2020.
+    years_forecast=10,
+    project_boundaries="project_boundaries.gpkg",
+    output_file="defor_project.csv",
+    verbose=False,
+)