diff --git a/metomi/rose/etc/tutorial/cylc-forecasting-workflow/.validate b/metomi/rose/etc/tutorial/cylc-forecasting-workflow/.validate
index 38d306164d..d7331a932a 100755
--- a/metomi/rose/etc/tutorial/cylc-forecasting-workflow/.validate
+++ b/metomi/rose/etc/tutorial/cylc-forecasting-workflow/.validate
@@ -16,11 +16,9 @@
 # along with this program.  If not, see <http://www.gnu.org/licenses/>.
 
 set -eux
-APIKEY="$(head --lines 1 ../api-keys)"
 FLOW_NAME="$(< /dev/urandom tr -dc A-Za-z | head -c6)"
 cylc lint .
 cylc install --workflow-name "$FLOW_NAME" --no-run-name
-sed -i "s/DATAPOINT_API_KEY/$APIKEY/" "$HOME/cylc-run/$FLOW_NAME/flow.cylc"
 cylc validate --check-circular --icp=2000 "$FLOW_NAME"
 cylc play --no-detach --abort-if-any-task-fails "$FLOW_NAME"
 cylc clean "$FLOW_NAME"
diff --git a/metomi/rose/etc/tutorial/cylc-forecasting-workflow/bin/consolidate-observations b/metomi/rose/etc/tutorial/cylc-forecasting-workflow/bin/consolidate-observations
index 0a25a2e53e..e6babab133 100755
--- a/metomi/rose/etc/tutorial/cylc-forecasting-workflow/bin/consolidate-observations
+++ b/metomi/rose/etc/tutorial/cylc-forecasting-workflow/bin/consolidate-observations
@@ -61,7 +61,7 @@ def plot_wind_data(wind_x, wind_y, x_range, y_range, x_coords, y_coords,
                [x[0] for x in z_coords],
                [y[1] for y in z_coords],
                color='red')
-    plt.savefig('wind.png')
+    plt.savefig(f'{os.environ["CYLC_TASK_LOG_DIR"]}/wind.png')
 
 
 def get_wind_fields():
@@ -115,4 +115,5 @@ def main():
 
 
 if __name__ == '__main__':
+    util.sleep()
     main()
diff --git a/metomi/rose/etc/tutorial/cylc-forecasting-workflow/bin/forecast b/metomi/rose/etc/tutorial/cylc-forecasting-workflow/bin/forecast
index a3f73c1edd..c5e46cbf8b 100755
--- a/metomi/rose/etc/tutorial/cylc-forecasting-workflow/bin/forecast
+++ b/metomi/rose/etc/tutorial/cylc-forecasting-workflow/bin/forecast
@@ -150,11 +150,12 @@ def push_rainfall(rainfall, wind_data, step, resolution, spline_level):
                                 dim_x, dim_y, resolution, resolution,
                                 spline_level)
 
+    domain = util.parse_domain(os.environ['DOMAIN'])
+
     while True:
         out_of_bounds = []
         for itt in range(len(x_values)):
             try:
-                domain = util.parse_domain(os.environ['DOMAIN'])
                 lng = domain['lng1'] + x_values[itt]
                 lat = domain['lat1'] + y_values[itt]
 
@@ -242,6 +243,7 @@ def main(forecast_interval, forecast_iterations):
 
 
 if __name__ == '__main__':
+    util.sleep()
     try:
         args = [int(sys.argv[1]), int(sys.argv[2])]
     except IndexError:
diff --git a/metomi/rose/etc/tutorial/cylc-forecasting-workflow/bin/get-observations b/metomi/rose/etc/tutorial/cylc-forecasting-workflow/bin/get-observations
index 394add6012..329c25dc39 100755
--- a/metomi/rose/etc/tutorial/cylc-forecasting-workflow/bin/get-observations
+++ b/metomi/rose/etc/tutorial/cylc-forecasting-workflow/bin/get-observations
@@ -24,11 +24,8 @@ Usage:
     get-observations
 
 Environment Variables:
-    SITE_ID: The four digit DataPoint site code identifying the weather station
-        we are to fetch results for.
-    API_KEY: The DataPoint API key, required for getting live weather data.
-        If un-specified then get-observations will fall back to archive data
-        from the workflow directory.
+    SITE_ID: The four digit WMO (World Meteorological Organization)
+    site code identifying the weather station we are to fetch results for.
 
 """
 
@@ -40,10 +37,8 @@ import re
 
 import requests
 
+import util
 
-BASE_URL = ('http://datapoint.metoffice.gov.uk/public/data/'
-            'val/wxobs/all/json/{site_id}'
-            '?res=hourly&time={time}&key={api_key}')
 
 # Compass bearings for ordinate directions.
 # NOTE: We measure direction by where the winds have come from!
@@ -66,22 +61,52 @@ ORDINATES = {
     'NNW': '157.5'
 }
 
+class Meteorology:
+    """
+    Surface winds tend to be 20-30 degrees backed (vector anticlockwise)
+    from the winds which steer weather systems:
+    Friction with the ground surface tends to mean that land surface winds
+    are as slow as half the speed of the wind at 2000ft. This fudge factor is
+    a more conservative 1.5:
 
-class NoDataException(Exception):
-    ...
+    .. seealso::
 
+       [Source Book to the Forecaster's Reference Book](https://digital.nmla
+       .metoffice.gov.uk/IO_011f7cd4-50fc-4903-b556-d24480ea883d/), section
+       1.2
+    """
+    SURFACE_BACKING = 2
+    SURFACE_FRICTION = .66
+    KT_TO_MPH = 1.15078
+
+    @staticmethod
+    def process_direction(direction: str) -> str:
+        """Process raw wind direction:
+
+        * Convert direction from 10s of degrees to degrees.
+        * Convert from Oceanographic (wind to) to Meteorological (wind from)
+          convention.
+        * Surface Friction Correction
+        """
+        return str(
+            ((int(direction) + 18 - Meteorology.SURFACE_BACKING) % 36) * 10)
+
+    @staticmethod
+    def process_speed(speed: str) -> str:
+        """Process Raw wind speed
+
+        * Convert to KT to MPH
+        * Surface Friction Correction
+        """
+        return str(
+            (
+                int(speed) * Meteorology.KT_TO_MPH
+            ) / Meteorology.SURFACE_FRICTION
+        )
 
-def get_datapoint_data(site_id, time, api_key):
-    """Get weather data from the DataPoint service."""
-    # The URL required to get the data.
-    time = datetime.strptime(time, '%Y%m%dT%H%MZ').strftime('%Y-%m-%dT%H:%MZ')
-    url = BASE_URL.format(time=time, site_id=site_id, api_key=api_key)
-    req = requests.get(url)
-    if req.status_code != 200:
-        raise Exception(f'{url} returned exit code {req.status_code}')
-    # Get the data and parse it as JSON.
-    print('Opening URL: %s' % url)
-    return req.json()['SiteRep']['DV']['Location']
+
+class NoDataException(Exception):
+    ...
 
 
 def get_archived_data(site_id, time):
@@ -150,11 +175,12 @@ def synop_grab(site_id, time):
         raise NoDataException(
             f'Request for data failed, raw request was {req.text}')
 
-    # Parse the direction from 10's of degrees to degrees:
-    data['direction'] = str(int(data['direction']) * 10)
+    # * Parse the direction from 10's of degrees to degrees
+    # * Convert direction from to direction it's blowing to
+    data['direction'] = Meteorology.process_direction(data['direction'])
 
     # Convert data in KT to MPH:
-    data['speed'] = str(int(data['speed']) * 1.15078)
+    data['speed'] = Meteorology.process_speed(data['speed'])
 
     # Get lat and long from MO Data:
     lat, lon = reference_lat_long(site_id)
@@ -183,7 +209,7 @@ def get_nearby_site(site_id, badsites):
     return int(result[0]), dist
 
 
-def main(site_id, api_key=None):
+def main(site_id):
     cycle_point = os.environ['CYLC_TASK_CYCLE_POINT']
 
     # Try to get the information from SYNOPS.
@@ -200,13 +226,8 @@ def main(site_id, api_key=None):
             site_id, dist = get_nearby_site(site_id, badsites)
 
     if obs is None:
-        if api_key:
-            print('Attempting to get weather data from DataPoint...')
-            data = get_datapoint_data(site_id, cycle_point, api_key)
-        else:
-            print('No API key provided, falling back to archived data')
-            data = get_archived_data(site_id, cycle_point)
-
+        print('Obs unavailable, falling back to archived data')
+        data = get_archived_data(site_id, cycle_point)
         obs = extract_observations(data)
 
     # Write observations.
@@ -215,5 +236,5 @@ def main(site_id, api_key=None):
 
 
 if __name__ == '__main__':
-    main(os.environ['SITE_ID'],
-         os.environ.get('API_KEY'))
+    util.sleep()
+    main(os.environ['SITE_ID'])
diff --git a/metomi/rose/etc/tutorial/cylc-forecasting-workflow/bin/get-rainfall b/metomi/rose/etc/tutorial/cylc-forecasting-workflow/bin/get-rainfall
index 481b203765..811fa46231 100755
--- a/metomi/rose/etc/tutorial/cylc-forecasting-workflow/bin/get-rainfall
+++ b/metomi/rose/etc/tutorial/cylc-forecasting-workflow/bin/get-rainfall
@@ -24,9 +24,6 @@ Usage:
     get-rainfall
 
 Environment Variables:
-    API_KEY: The DataPoint API key, required for getting live weather data.
-        If un-specified then get-observations will fall back to archive data
-        from the workflow directory.
     DOMAIN: The area in which to generate forecasts for in the format
         (lng1, lat1, lng2, lat2).
     RESOLUTION: The length/width of each grid cell in degrees.
@@ -38,20 +35,20 @@ import math
 import os
 import shutil
 
-import requests
+import h5py
+from pathlib import Path
+import urllib
+import requests   # noqa: F401 - required implicitly by urllib.
 
-try:
-    from PIL import Image
-except ModuleNotFoundError:
-    # not all PIL installations are created equal
-    # sometimes we must import like this
-    import Image
-from mercator import get_offset, get_scale, pos_to_coord
+from mercator import get_scale, pos_to_coord
 import util
 
-
-URL = ('http://datapoint.metoffice.gov.uk/public/data/layer/wxobs/'
-       'RADAR_UK_Composite_Highres/png?TIME={time}&key={api_key}')
+S3URL = (
+    'https://met-office-radar-obs-data.s3-eu-west-2.amazonaws.com/radar/'
+    '{Y}/{m}/{d}/{YYYYmmddHHMM}_ODIM_ng_radar_rainrate_composite_1km_UK.h5'
+)
+DEBUG=os.environ['CYLC_DEBUG']=='true'
+CYLC_TASK_LOG_DIR=os.environ['CYLC_TASK_LOG_DIR']
 
 
 class Rainfall(object):
@@ -62,16 +59,6 @@ class Rainfall(object):
         resolution (float): The length of each grid cell in degrees.
 
     """
-    VALUE_MAP = {
-        (0, 0, 254, 255): 1,
-        (50, 101, 254, 255): 2,
-        (127, 127, 0, 255): 3,
-        (254, 203, 0, 255): 4,
-        (254, 152, 0, 255): 5,
-        (254, 0, 0, 255): 6,
-        (254, 0, 254, 255): 7
-    }
-
     def __init__(self, domain, resolution):
         self.resolution = resolution
         self.domain = domain
@@ -98,10 +85,23 @@ class Rainfall(object):
         """
         itt_x, itt_y = util.get_grid_coordinates(lng, lat, self.domain,
                                                  self.resolution)
-        try:
-            self.data[itt_y][itt_x].append(self.VALUE_MAP[value])
-        except KeyError:
-            pass
+                    
+        self.data[itt_y][itt_x].append(self.value_map(value))
+
+    @staticmethod
+    def value_map(v):
+        """Convert rainfall rate values into colour space values.
+
+        Checks if rainfall value above each threshold in turn.
+
+        TODO:
+            - Unit test this
+        """
+        thresholds = {32, 16, 8, 4, 2, 1, .5, .2}
+        for i, threshold in enumerate(sorted(thresholds, reverse=True)):
+            if v > threshold:
+                return 8 - i
+        return 0
 
     def compute_bins(self):
         """Return this dataset as a 2D matrix."""
@@ -114,25 +114,6 @@ class Rainfall(object):
         return self.data
 
 
-def get_datapoint_radar_image(filename, time, api_key):
-    """Retrieve a png image of rainfall from the DataPoint service.
-
-    Args:
-        filename (str): The path to write the image file to.
-        time (str): The datetime of the image to retrieve in ISO8601 format.
-        api_key (str): Datapoint API key.
-
-    """
-    time = datetime.strptime(time, '%Y%m%dT%H%MZ').strftime(
-        '%Y-%m-%dT%H:%M:%SZ')
-    url = URL.format(time=time, api_key=api_key)
-    req = requests.get(url)
-    if req.status_code != 200:
-        raise Exception(f'{url} returned exit code {req.status_code}')
-    with open(filename, 'bw') as png_file:
-        png_file.write(req.content)
-
-
 def get_archived_radar_image(filename, time):
     """Retrieve a png image from the archived data in the workflow directory.
 
@@ -147,8 +128,23 @@ def get_archived_radar_image(filename, time):
         filename)
 
 
+def get_amazon_radar_data(filename, time):
+    time = datetime.strptime(time, '%Y%m%dT%H%MZ')
+    url = S3URL.format(
+        Y=time.strftime('%Y'),
+        m=time.strftime('%m'),
+        d=time.strftime('%d'),
+        YYYYmmddHHMM=time.strftime('%Y%m%d%H%M')
+    )
+    print(f'[INFO] Getting data from {url=}')
+    data = urllib.request.urlopen(url).read()
+    with open(filename, 'wb') as fh:
+        fh.write(data)
+
+
+# def process_rainfall_data(filename, resolution, domain):
 def process_rainfall_data(filename, resolution, domain):
-    """Generate a 2D matrix of data from the rainfall data in the image.
+    """get_amazon_radar_dataGenerate a 2D matrix of data from the rainfall data in the image.
 
     Args:
         filename (str): Path to the png image to process.
@@ -160,39 +156,65 @@ def process_rainfall_data(filename, resolution, domain):
         list - A 2D matrix of rainfall data.
 
     """
+    print(f'[INFO] Analysing data from {filename}')
+    data = h5py.File(filename)['dataset1']['data1']['data']
     rainfall = Rainfall(domain, resolution)
 
-    image = Image.open(filename)
-    scale = get_scale(domain, image.width)
-    offset = get_offset(domain, scale)
+    _height, width = data.shape
+
+    scale = get_scale(domain, width)
+    offset = (-1100.8461538461539, 1400.6953225710452)
+
+    if DEBUG:
+        print(f'[INFO] {scale=}, {offset=}')
+        from matplotlib import pyplot as plt
+        Path(CYLC_TASK_LOG_DIR).mkdir(parents=True, exist_ok=True)
+        plt.imshow(data)
+        plt.savefig(f'{CYLC_TASK_LOG_DIR}/raw.data.png')
 
-    for itt_x in range(image.width):
-        for itt_y in range(image.height):
+    for itt_y, row in enumerate(data):
+        for itt_x, col in enumerate(row):
             lng, lat = pos_to_coord(
                 itt_x,
                 itt_y * (2. / 3.),  # Counter aspect ratio.
-                offset, scale)
-            rainfall.add(lng, lat, image.getpixel((itt_x, itt_y)))
+                offset,
+                scale
+            )
+            val = float(col)
+            # Original data uses -1 to indicate radar mask
+            val = 0 if val == -1 else val
+            rainfall.add(lng, lat, val)
+    data = rainfall.compute_bins()
 
-    return rainfall.compute_bins()
+    if DEBUG:
+        plt.imshow(data)
+        plt.savefig(f'{CYLC_TASK_LOG_DIR}/processed.data.png')
 
+    return data
 
 def main():
     time = os.environ['CYLC_TASK_CYCLE_POINT']
     resolution = float(os.environ['RESOLUTION'])
     domain = util.parse_domain(os.environ['DOMAIN'])
-    api_key = os.environ.get('API_KEY')
 
-    if api_key:
-        print('Attempting to get weather data from the DataPoint service.')
-        get_datapoint_radar_image('rainfall-radar.png', time, api_key)
+    # Acts as a switch - if a file-name is given, use that file-name
+    canned_data = os.environ.get('CANNED_DATA', 'fetch')
+
+    if canned_data == 'fetch':
+        print('[INFO] Attempting to get rainfall data from S3 bucket')
+        get_amazon_radar_data('radardata.h5', time)
+        canned_data = 'radardata.h5'
     else:
-        print('No API key provided, falling back to archived data')
-        get_archived_radar_image('rainfall-radar.png', time)
+        raise Exception('Not currently provided functionality')
+        # Not currently used.
+        #print(f'[INFO] Canned data provided: {canned_data}')
+        #get_archived_radar_data(canned_data, time)
+
+    data = process_rainfall_data(canned_data, resolution, domain)
 
-    data = process_rainfall_data('rainfall-radar.png', resolution, domain)
     util.write_csv('rainfall.csv', data)
 
 
 if __name__ == '__main__':
+    util.sleep(2)  # make the tutorial run a little slower
     main()
diff --git a/metomi/rose/etc/tutorial/cylc-forecasting-workflow/bin/post-process b/metomi/rose/etc/tutorial/cylc-forecasting-workflow/bin/post-process
index 3316152e8e..c834d644cf 100755
--- a/metomi/rose/etc/tutorial/cylc-forecasting-workflow/bin/post-process
+++ b/metomi/rose/etc/tutorial/cylc-forecasting-workflow/bin/post-process
@@ -101,6 +101,7 @@ def main(site_name, time):
 
 
 if __name__ == '__main__':
+    util.sleep()
     try:
         args = (sys.argv[1].lower(), int(sys.argv[2]))
     except IndexError:
diff --git a/metomi/rose/etc/tutorial/cylc-forecasting-workflow/flow.cylc b/metomi/rose/etc/tutorial/cylc-forecasting-workflow/flow.cylc
index 20c95cb34b..e9743ec3c3 100644
--- a/metomi/rose/etc/tutorial/cylc-forecasting-workflow/flow.cylc
+++ b/metomi/rose/etc/tutorial/cylc-forecasting-workflow/flow.cylc
@@ -1,6 +1,4 @@
 #!jinja2
-[scheduler]
-    UTC mode = True
 [task parameters]
     # A list of the weather stations we will be fetching observations from.
     station = camborne, heathrow, shetland, aldergrove
@@ -41,15 +39,13 @@
         [[[environment]]]
             # The dimensions of each grid cell in degrees.
             RESOLUTION = 0.2
+
             # The area to generate forecasts for (lng1, lat1, lng2, lat2)
-            DOMAIN = -12,48,5,61  # Do not change!
+            DOMAIN = -12,46,12,61
+
 
     [[get_observations<station>]]
         script = get-observations
-        [[[environment]]]
-            # The key required to get weather data from the DataPoint service.
-            # To use archived data comment this line out.
-            API_KEY = DATAPOINT_API_KEY
 
     [[get_observations<station=aldergrove>]]
         [[[environment]]]
@@ -69,10 +65,6 @@
 
     [[get_rainfall]]
         script = get-rainfall
-        [[[environment]]]
-            # The key required to get weather data from the DataPoint service.
-            # To use archived data comment this line out.
-            API_KEY = DATAPOINT_API_KEY
 
     [[forecast]]
         script = forecast 60 5  # Generate 5 forecasts at 60 minute intervals.
diff --git a/metomi/rose/etc/tutorial/cylc-forecasting-workflow/lib/python/util.py b/metomi/rose/etc/tutorial/cylc-forecasting-workflow/lib/python/util.py
index 2addf5c5f2..6450bbc161 100644
--- a/metomi/rose/etc/tutorial/cylc-forecasting-workflow/lib/python/util.py
+++ b/metomi/rose/etc/tutorial/cylc-forecasting-workflow/lib/python/util.py
@@ -21,9 +21,11 @@
 from copy import copy
 from contextlib import suppress
 import math
-import jinja2
+import os
 import sys
+import time
 
+import jinja2
 
 R_0 = 6371.  # Radius of the Earth (km).
 
@@ -280,13 +282,18 @@ def __call__(self, grid_x, grid_y):
         return z_val
 
 
-def parse_domain(domain):
-    bbox = list(map(float, domain.split(',')))
+def parse_domain(domain: str):
+    lng1, lat1, lng2, lat2 = list(map(float, domain.split(',')))
+    msg = "Invalid domain '{}' ({} {} >= {})"
+    if lng1 >= lng2:
+        raise ValueError(msg.format(domain, 'longitude', lng1, lng2))
+    if lat1 >= lat2:
+        raise ValueError(msg.format(domain, 'latitude', lat1, lat2))
     return {
-        'lng1': bbox[0],
-        'lat1': bbox[1],
-        'lng2': bbox[2],
-        'lat2': bbox[3]
+        'lng1': lng1,
+        'lat1': lat1,
+        'lng2': lng2,
+        'lat2': lat2,
     }
 
 
@@ -301,3 +308,12 @@ def generate_html_map(filename, template_file, data, domain, resolution):
                 lat2=domain['lat2'],
                 data=data
             ))
+
+
+def sleep(secs=4):
+    """Make the tutorials run a little slower so users can follow along.
+
+    (Only if not running in CI).
+    """
+    if 'CI' not in os.environ:
+        time.sleep(secs)