diff --git a/backend/Dockerfile.processing b/backend/Dockerfile.processing
index b6f3713..9467256 100644
--- a/backend/Dockerfile.processing
+++ b/backend/Dockerfile.processing
@@ -1,4 +1,4 @@
-# backend/Dockerfile.validation
+# backend/Dockerfile.processing
 FROM apache/beam_python3.11_sdk:2.60.0
 
 # Install system dependencies including GDAL
@@ -17,9 +17,21 @@ ENV C_INCLUDE_PATH=/usr/include/gdal
 COPY requirements.txt /app/requirements.txt
 RUN pip install -r /app/requirements.txt
 
+# Install additional dependencies needed for validation
+RUN pip install \
+    geopandas \
+    pandas \
+    shapely \
+    fsspec>=2024.10.0 \
+    gcsfs>=2024.10.0 \
+    pyarrow
+
 # Copy the dataflow directory (which includes setup.py)
 COPY dataflow /app/dataflow
 
 # Install the package in development mode
 WORKDIR /app/dataflow
 RUN pip install -e .
+
+# Set the working directory
+WORKDIR /app
diff --git a/backend/dataflow/validate_cadastral.py b/backend/dataflow/validate_cadastral.py
new file mode 100644
index 0000000..2c1ad97
--- /dev/null
+++ b/backend/dataflow/validate_cadastral.py
@@ -0,0 +1,95 @@
+import apache_beam as beam
+from apache_beam.options.pipeline_options import PipelineOptions
+import logging
+
+class ValidateGeometriesOptions(PipelineOptions):
+    @classmethod
+    def _add_argparse_args(cls, parser):
+        parser.add_argument('--dataset')
+        parser.add_argument('--input_bucket')
+        parser.add_argument('--output_bucket')
+
+class ReadParquetDoFn(beam.DoFn):
+    def __init__(self, input_bucket):
+        self.input_bucket = input_bucket
+    
+    def process(self, dataset):
+        import geopandas as gpd
+        path = f'gs://{self.input_bucket}/raw/{dataset}/current.parquet'
+        logging.info(f'Reading from {path}')
+        gdf = gpd.read_parquet(path)
+        yield {'dataset': dataset, 'data': gdf}
+
+class ValidateCadastralDoFn(beam.DoFn):
+    def process(self, element):
+        from shapely.geometry import Polygon, MultiPolygon
+        gdf = element['data']
+        dataset = element['dataset']
+        
+        # Fix invalid geometries
+        invalid_mask = ~gdf.geometry.is_valid
+        if invalid_mask.any():
+            gdf.loc[invalid_mask, 'geometry'] = gdf.loc[invalid_mask, 'geometry'].apply(
+                lambda geom: geom.buffer(0) if geom else None
+            )
+        
+        # Cadastral-specific validation
+        gdf = gdf[
+            gdf.geometry.is_valid &
+            gdf.geometry.apply(lambda x: isinstance(x, (Polygon, MultiPolygon))) &
+            ~gdf.geometry.is_empty &
+            gdf.geometry.apply(lambda x: 0.1 <= x.area <= 100_000_000)  # Area between 0.1m² and 100km²
+        ]
+        
+        stats = {
+            'total_rows': len(element['data']),
+            'valid_geometries': len(gdf)
+        }
+        
+        element['data'] = gdf
+        element['stats'] = stats
+        yield element
+
+class WriteResultsDoFn(beam.DoFn):
+    def __init__(self, output_bucket):
+        self.output_bucket = output_bucket
+    
+    def process(self, element):
+        import pandas as pd
+        dataset = element['dataset']
+        gdf = element['data']
+        stats = element['stats']
+        
+        # Write parquet
+        output_path = f'gs://{self.output_bucket}/validated/{dataset}/current.parquet'
+        logging.info(f'Writing to {output_path}')
+        gdf.to_parquet(
+            output_path,
+            compression='zstd',
+            compression_level=3,
+            index=False,
+            row_group_size=100000
+        )
+        
+        # Write stats
+        stats_path = f'gs://{self.output_bucket}/validated/{dataset}/validation_stats.csv'
+        pd.DataFrame([stats]).to_csv(stats_path, index=False)
+        
+        yield element
+
+def run(argv=None):
+    """Build and run the pipeline."""
+    pipeline_options = PipelineOptions(argv)
+    options = pipeline_options.view_as(ValidateGeometriesOptions)
+    
+    with beam.Pipeline(options=pipeline_options) as p:
+        (p 
+         | 'Create Dataset' >> beam.Create([options.dataset])
+         | 'Read Parquet' >> beam.ParDo(ReadParquetDoFn(options.input_bucket))
+         | 'Validate Geometries' >> beam.ParDo(ValidateCadastralDoFn())
+         | 'Write Results' >> beam.ParDo(WriteResultsDoFn(options.output_bucket))
+        )
+
+if __name__ == '__main__':
+    logging.getLogger().setLevel(logging.INFO)
+    run() 
\ No newline at end of file
diff --git a/backend/dataflow/validate_water_projects.py b/backend/dataflow/validate_water_projects.py
new file mode 100644
index 0000000..c6993c6
--- /dev/null
+++ b/backend/dataflow/validate_water_projects.py
@@ -0,0 +1,95 @@
+import apache_beam as beam
+from apache_beam.options.pipeline_options import PipelineOptions
+import logging
+
+class ValidateGeometriesOptions(PipelineOptions):
+    @classmethod
+    def _add_argparse_args(cls, parser):
+        parser.add_argument('--dataset')
+        parser.add_argument('--input_bucket')
+        parser.add_argument('--output_bucket')
+
+class ReadParquetDoFn(beam.DoFn):
+    def __init__(self, input_bucket):
+        self.input_bucket = input_bucket
+    
+    def process(self, dataset):
+        import geopandas as gpd
+        path = f'gs://{self.input_bucket}/raw/{dataset}/current.parquet'
+        logging.info(f'Reading from {path}')
+        gdf = gpd.read_parquet(path)
+        yield {'dataset': dataset, 'data': gdf}
+
+class ValidateWaterProjectsDoFn(beam.DoFn):
+    def process(self, element):
+        from shapely.geometry import Polygon, MultiPolygon
+        gdf = element['data']
+        dataset = element['dataset']
+        
+        # Fix invalid geometries
+        invalid_mask = ~gdf.geometry.is_valid
+        if invalid_mask.any():
+            gdf.loc[invalid_mask, 'geometry'] = gdf.loc[invalid_mask, 'geometry'].apply(
+                lambda geom: geom.buffer(0) if geom else None
+            )
+        
+        # Water projects-specific validation
+        gdf = gdf[
+            gdf.geometry.is_valid &
+            gdf.geometry.apply(lambda x: isinstance(x, (Polygon, MultiPolygon))) &
+            ~gdf.geometry.is_empty &
+            gdf.geometry.apply(lambda x: 1 <= x.area <= 10_000_000)  # Area between 1m² and 10km²
+        ]
+        
+        stats = {
+            'total_rows': len(element['data']),
+            'valid_geometries': len(gdf)
+        }
+        
+        element['data'] = gdf
+        element['stats'] = stats
+        yield element
+
+class WriteResultsDoFn(beam.DoFn):
+    def __init__(self, output_bucket):
+        self.output_bucket = output_bucket
+    
+    def process(self, element):
+        import pandas as pd
+        dataset = element['dataset']
+        gdf = element['data']
+        stats = element['stats']
+        
+        # Write parquet
+        output_path = f'gs://{self.output_bucket}/validated/{dataset}/current.parquet'
+        logging.info(f'Writing to {output_path}')
+        gdf.to_parquet(
+            output_path,
+            compression='zstd',
+            compression_level=3,
+            index=False,
+            row_group_size=100000
+        )
+        
+        # Write stats
+        stats_path = f'gs://{self.output_bucket}/validated/{dataset}/validation_stats.csv'
+        pd.DataFrame([stats]).to_csv(stats_path, index=False)
+        
+        yield element
+
+def run(argv=None):
+    """Build and run the pipeline."""
+    pipeline_options = PipelineOptions(argv)
+    options = pipeline_options.view_as(ValidateGeometriesOptions)
+    
+    with beam.Pipeline(options=pipeline_options) as p:
+        (p 
+         | 'Create Dataset' >> beam.Create([options.dataset])
+         | 'Read Parquet' >> beam.ParDo(ReadParquetDoFn(options.input_bucket))
+         | 'Validate Geometries' >> beam.ParDo(ValidateWaterProjectsDoFn())
+         | 'Write Results' >> beam.ParDo(WriteResultsDoFn(options.output_bucket))
+        )
+
+if __name__ == '__main__':
+    logging.getLogger().setLevel(logging.INFO)
+    run() 
\ No newline at end of file
diff --git a/backend/dataflow/validate_wetlands.py b/backend/dataflow/validate_wetlands.py
new file mode 100644
index 0000000..ce123ad
--- /dev/null
+++ b/backend/dataflow/validate_wetlands.py
@@ -0,0 +1,94 @@
+import apache_beam as beam
+from apache_beam.options.pipeline_options import PipelineOptions
+import logging
+
+class ValidateGeometriesOptions(PipelineOptions):
+    @classmethod
+    def _add_argparse_args(cls, parser):
+        parser.add_argument('--dataset')
+        parser.add_argument('--input_bucket')
+        parser.add_argument('--output_bucket')
+
+class ReadParquetDoFn(beam.DoFn):
+    def __init__(self, input_bucket):
+        self.input_bucket = input_bucket
+    
+    def process(self, dataset):
+        import geopandas as gpd
+        path = f'gs://{self.input_bucket}/raw/{dataset}/current.parquet'
+        logging.info(f'Reading from {path}')
+        gdf = gpd.read_parquet(path)
+        yield {'dataset': dataset, 'data': gdf}
+
+class ValidateWetlandsDoFn(beam.DoFn):
+    def process(self, element):
+        from shapely.geometry import Polygon, MultiPolygon
+        gdf = element['data']
+        dataset = element['dataset']
+        
+        # Fix invalid geometries
+        invalid_mask = ~gdf.geometry.is_valid
+        if invalid_mask.any():
+            gdf.loc[invalid_mask, 'geometry'] = gdf.loc[invalid_mask, 'geometry'].apply(
+                lambda geom: geom.buffer(0) if geom else None
+            )
+        
+        # Wetlands-specific validation
+        gdf = gdf[
+            gdf.geometry.is_valid &
+            gdf.geometry.apply(lambda x: isinstance(x, (Polygon, MultiPolygon))) &
+            ~gdf.geometry.is_empty
+        ]
+        
+        stats = {
+            'total_rows': len(element['data']),
+            'valid_geometries': len(gdf)
+        }
+        
+        element['data'] = gdf
+        element['stats'] = stats
+        yield element
+
+class WriteResultsDoFn(beam.DoFn):
+    def __init__(self, output_bucket):
+        self.output_bucket = output_bucket
+    
+    def process(self, element):
+        import pandas as pd
+        dataset = element['dataset']
+        gdf = element['data']
+        stats = element['stats']
+        
+        # Write parquet
+        output_path = f'gs://{self.output_bucket}/validated/{dataset}/current.parquet'
+        logging.info(f'Writing to {output_path}')
+        gdf.to_parquet(
+            output_path,
+            compression='zstd',
+            compression_level=3,
+            index=False,
+            row_group_size=100000
+        )
+        
+        # Write stats
+        stats_path = f'gs://{self.output_bucket}/validated/{dataset}/validation_stats.csv'
+        pd.DataFrame([stats]).to_csv(stats_path, index=False)
+        
+        yield element
+
+def run(argv=None):
+    """Build and run the pipeline."""
+    pipeline_options = PipelineOptions(argv)
+    options = pipeline_options.view_as(ValidateGeometriesOptions)
+    
+    with beam.Pipeline(options=pipeline_options) as p:
+        (p 
+         | 'Create Dataset' >> beam.Create([options.dataset])
+         | 'Read Parquet' >> beam.ParDo(ReadParquetDoFn(options.input_bucket))
+         | 'Validate Geometries' >> beam.ParDo(ValidateWetlandsDoFn())
+         | 'Write Results' >> beam.ParDo(WriteResultsDoFn(options.output_bucket))
+        )
+
+if __name__ == '__main__':
+    logging.getLogger().setLevel(logging.INFO)
+    run() 
\ No newline at end of file