diff --git a/aimodel/src/parse_args.py b/aimodel/src/parse_args.py
index c7c6053..c8aa9f7 100644
--- a/aimodel/src/parse_args.py
+++ b/aimodel/src/parse_args.py
@@ -24,6 +24,7 @@ Available subcommands:
 	train-predict		Make predictions using a model trained through the train subcommand.
 	train-mono			Train a mono rainfall → water depth model.
 	train-mono-predict	Make predictions using a model trained through the train-mono subcommand.
+	rainfall-stats		Calculate statistics about the rainfall radar data.
 
 For more information, do src/index.py <subcommand> --help.
 """)
diff --git a/aimodel/src/subcommands/rainfall_stats.py b/aimodel/src/subcommands/rainfall_stats.py
new file mode 100755
index 0000000..8bd8d1d
--- /dev/null
+++ b/aimodel/src/subcommands/rainfall_stats.py
@@ -0,0 +1,85 @@
+import io
+import json
+import os
+import sys
+import argparse
+import re
+
+from loguru import logger
+import tensorflow as tf
+from lib.dataset.batched_iterator import batched_iterator
+
+from lib.io.handle_open import handle_open
+from lib.ai.RainfallWaterMono import RainfallWaterMono
+from lib.dataset.dataset_mono import dataset_mono_predict
+from lib.io.find_paramsjson import find_paramsjson
+from lib.io.readfile import readfile
+from lib.vis.segmentation_plot import segmentation_plot
+
+
+MODE_JSONL = 1
+MODE_PNG = 2
+
+def parse_args():
+	parser = argparse.ArgumentParser(description="Output water depth image segmentation maps using a given pretrained mono model.")
+	# parser.add_argument("--config", "-c", help="Filepath to the TOML config file to load.", required=True)
+	parser.add_argument("--input", "-i", help="Path to input directory containing the .tfrecord(.gz) files to predict for. If a single file is passed instead, then only that file will be converted.", required=True)
+	parser.add_argument("--reads-multiplier", help="Optional. The multiplier for the number of files we should read from at once. Defaults to 0. When using this start with 1.5, which means read ceil(NUMBER_OF_CORES * 1.5). Set to a higher number of systems with high read latency to avoid starving the GPU of data. SETTING THIS WILL SCRAMBLE THE ORDER OF THE DATASET.")
+	parser.add_argument("--batch_size", help="Optional. The batch size to calculate statistics with. Can be larger than normal since we don't have a model loaded. Default: 1024")
+	return parser
+
+def run(args):
+	
+	if (not hasattr(args, "read_multiplier")) or args.read_multiplier == None:
+		args.read_multiplier = 4
+	if (not hasattr(args, "batch_size")) or args.batch_size == None:
+		args.batch_size = 1024
+	
+	
+	sys.stderr.write(f"\n\n>>> This is TensorFlow {tf.__version__}\n\n\n")
+	
+	# Note that if using a directory of input files, the output order is NOT GUARANTEED TO BE THE SAME. In fact, it probably won't be (see dataset_mono for more details).
+	dataset = dataset_mono_predict(
+		dirpath_input=args.input,
+		parallel_reads_multiplier=args.read_multiplier
+	)
+	
+	# for items in dataset_train.repeat(10):
+	# 	print("ITEMS", len(items))
+	# 	print("LEFT", [ item.shape for item in items[0] ])
+	# print("ITEMS DONE")
+	# exit(0)
+	
+	
+	logger.info("RAINFALL STATS")
+	
+	calc_mean = []
+	calc_stddev = []
+	calc_max = []
+	
+	i = 0
+	for batch in batched_iterator(dataset, tensors_in_item=2, batch_size=args.batch_size):
+		rainfall_actual_batch, water_actual_batch = batch
+		
+		rainfall_flat = tf.reshape(rainfall_actual_batch, [-1])
+		
+		batch_mean = tf.math.reduce_mean(rainfall_flat)
+		batch_stddev = tf.math.reduce_std(rainfall_flat)
+		batch_max = tf.math.reduce_max(rainfall_flat)
+		
+		print("BATCH", "mean", batch_mean, "stddev", batch_stddev, "max", batch_max)
+		
+		calc_mean.append(batch_mean)
+		calc_stddev.append(batch_stddev)
+		calc_max.append(batch_max)
+		
+		i += 1
+	
+	
+	calc_mean	= tf.math.reduce_mean(tf.stack(calc_mean))
+	calc_max	= tf.math.reduce_max(tf.stack(calc_max))
+	
+	print("STDDEV VALUES", tf.stack(calc_stddev).numpy().tolist())
+	print("OVERALL", "mean", calc_mean.numpy().tolist(), "max", calc_max.numpy().tolist())
+	
+	logger.write(">>> Complete\n")
\ No newline at end of file
diff --git a/aimodel/src/subcommands/train_mono_predict.py b/aimodel/src/subcommands/train_mono_predict.py
index c0e169f..4e3c5c1 100644
--- a/aimodel/src/subcommands/train_mono_predict.py
+++ b/aimodel/src/subcommands/train_mono_predict.py
@@ -149,15 +149,12 @@ def do_jsonl(args, ai, dataset, model_params, do_argmax=False):
 	for batch in batched_iterator(dataset, tensors_in_item=2, batch_size=model_params["batch_size"]):
 		rainfall_actual_batch, water_actual_batch = batch
 		
-		print("DEBUG:do_jsonl rainfall_actual_batch", rainfall_actual_batch.shape)
-		print("DEBUG:do_jsonl water_actual_batch", water_actual_batch.shape)
 		water_predict_batch = ai.embed(rainfall_actual_batch)
 		water_actual_batch = tf.unstack(water_actual_batch, axis=0)
 		rainfall_actual_batch = tf.unstack(rainfall_actual_batch, axis=0)
 		
 		i_batch = 0
 		for water_predict in water_predict_batch:
-			print("DEBUG:do_jsonl water_predict", water_predict.shape)
 			# [ width, height, softmax_probabilities ] → [ batch, width, height ]
 			if do_argmax:
 				water_predict = tf.math.argmax(water_predict, axis=-1)