A quasi-stationary multicell thunderstorm affected parts of mid-Suffolk on Sunday 25 July 2021. The epicentre was the village of Brettenham, where extreme rainfall resulted in over 180mm falling in under 2 h. Local flooding occurred, including significant crop damage. The amount of rain observed in such a short duration challenges UK rainfall records. A detailed mesoanalysis is constructed to analyse the meteorological conditions present, using surface observations from a combination of conventional and privately owned automatic weather stations. During the afternoon of Sunday 25 July 2021, a series of heavy showers and thunderstorms affected portions of East Anglia, London, central southern and southeast England. These largely developed along marked convergence zones in a rather slack pressure pattern, and their slow movement resulted in locally high rainfall totals. Unofficial rain gauge measurements suggested as much as 120mm may have fallen in the Woodford area of northeast London and in the eastern suburbs of Hertford, with around 90mm also close to Haverhill (Suffolk). These areas all experienced flash flooding (BBC News, 2021a,b,c), but the main focus of this paper is mid-Suffolk where extreme but highly localised rainfall occurred that could challenge some UK rainfall records. This article uses a variety of datasets to explore the synoptic background and evolution, observed rainfall totals, impacts on the local area and comparisons with other significant short-duration rainfall events in the United Kingdom. To gain an understanding into the meteorological conditions present before and during this event, a mesoanalysis was constructed using data from over 350 privately owned automatic weather stations (AWS) across Norfolk, Suffolk, Essex and adjacent portions of Cambridgeshire and Hertfordshire. This data, provided by Weather Underground (WU), 1 was merged with existing synoptic observations from AWS operated by the Met Office to create a gridded dataset. While the exposure of each private AWS will likely vary considerably, no bias correction was applied to the raw data since the general siting of each AWS is not known and the rather slack pressure pattern on this day made it difficult to determine appropriate correction values with the nearest Met Office AWS. Nonetheless, some elements of filtering were employed in an attempt to reduce the number of potential erroneous measurements being used in the gridded fields. Wind speed and direction time-series for all private AWS between 0000 utc and 2359 utc on 25 July 2021 were subjectively examined, and sites with consistently erratic speed or direction, and/or poorly calibrated wind direction, were removed. Furthermore, for air and dewpoint temperature a threshold of two standard deviations was applied, with any values falling outside of this range manually checked and subsequently removed if necessary. All filtered observations at each timestep of interest were interpolated onto a 0.025° (~2.5km) horizontal resolution grid using the inverse-distance weighting method with a power of 2. A slight smoothing factor was then applied to avoid any one specific observation dominating too much. For each timestep, the observation from each contributing AWS closest to and within the 10min leading up to the observation time was used. A cut-off 500hPa low, originally located over the Bay of Biscay on Friday 23 July, slowly migrated northeastwards across the Brest peninsula on Saturday 24 July, while deepening. By the Sunday morning (25 July), the centre of the upper low was located over the eastern English Channel and this slowly filled in-situ through the day (Figure 1a). Cold air aloft associated with this upper low, and attendant areas of divergence due to positive vorticity advection (PVA), created an environment increasingly favourable for deep moist convection – and hence thunderstorms. At the surface, a slack pressure pattern dominated with a small low centre over the Cherbourg peninsula at 0000 utc 25 July; this feature drifted eastwards across northern France during the day (Figure 1b), and then northeastwards to become centred near Lille by 1800 utc. A separate small surface low, initially over northwest Germany at 0000 utc 25 July, shifted northwestwards across the North Sea while rotating around the parent low over northern France. However, the main axis between these two features appeared to remain to the southeast and then east of the UK. Much of East Anglia experienced extensive low cloud and mist during the morning, as confirmed by surface observations and visible satellite imagery, with air temperatures close to the dewpoint temperature of 16–17°C. Some patchy light rain and drizzle was evident on radar, in part associated with a frontolysing occlusion shown in Figure 1(b). Breaks in the low cloud eventually developed over the southern North Sea and migrated westwards, into Suffolk and southeast Norfolk during the mid to late morning, and it is these areas where the air temperature increased quickest. Examination of mesoanalyses during the morning revealed that the surface winds were broadly northeasterly across the whole of East Anglia, but there was evidence of veering near eastern coasts during the late morning and early afternoon, presumably as a sea breeze developed in response to the increased thermal contrast between land and sea. The onshore breeze became increasingly perpendicular to the coast through the afternoon – the exact wind direction varied along the east coast due to the local orientation of the coastline.Where this east or southeasterly onshore wind met the synoptic northeasterly surface wind inland, a marked convergence zone (hereafter CZ) developed along a northeast–southwest line from southeast Norfolk to London. The shape and position of this boundary changed through the afternoon, in part due to the influence of outflow from nearby thunderstorms. ERA5 reanalysis (Hersbach et al., 2020) soundings suggest that an air temperature of 22–23°C was required for convective initiation (Figure 2). These values were reached in south Norfolk and across Suffolk by 1130 utc, with maximum temperatures of 24.9°C recorded at Cavendish and 24.2°C at both Charsfield and Wattisham during the afternoon. The first showers appeared as radar echoes around 1230 utc, and over the following hours (Figure 3) multiple clusters of thunderstorms developed along the CZ and drifted to the west–southwest before weakening, their westerly outflow winds at the surface helping to reinforce the existing CZ away to the east. Of particular interest was a thunderstorm that developed around 1440 utc near the A140, to the southwest of Eye (Suffolk). According to radar reflectivity, this storm reached peak intensity around 1500 utc whilst drifting southwestwards towards Woolpit (Suffolk), and spawned a daughter cell on its southern flank near Stowmarket shortly afterwards. This new thunderstorm grew rapidly in size and intensity over subsequent radar scans, and appeared to become quasi-stationary over Brettenham between 1520 utc and 1720 utc. Figure 2 reveals some slight directional and speed shear between the ground and 700hPa (approximately 0–3km) and a vorticity generation parameter (VGP; Rasmussen and Blanchard, 1998) of 0.11, suggesting the potential for updraughts to tilt away from downdraughts and thereby aiding cell organisation and longevity. There is evidence in radar data of back-building as daughter cells continued to develop near Stowmarket and drifted southwestwards over the same areas. The storm motion is estimated from Figure 2 at ~8kn, using 75% of the 0–6km shear. The mesoanalysis in Figure 4 highlights that the area of new cell development coincides with a particular distortion in the shape of the CZ, to the northeast of Stowmarket. This could perhaps be an artefact of the gridding of unevenly distributed observations, but evidence of diverging winds in the vicinity of Brettenham and a notable reduction in 2m temperature well away from the rain shield suggests outflow may have caused a portion of the CZ to bow outwards (i.e. towards the south and east), while remaining anchored on the northern flank. This therefore indicates particularly strong convergence given that winds flowing from the northeast, east, southeast and southwest all met in the vicinity of Stowmarket, allowing unimpeded advection of warm, unstable air over east Suffolk into the area, then subjected to forced ascent from the surface to generate new thunderstorm cells. This may explain, at least partly, why this particular area experienced prolonged and intense rainfall. It is clear by 1700 utc in Figure 3 that the multiple clusters of thunderstorms over west Suffolk and Essex had developed a substantial cold pool at the surface (this became even more significant by 1800 utc), and westerly outflow winds likely contributed to the eastward movement of the CZ, perhaps also aided by a weakening sea breeze by this stage as air temperatures began to reduce following peak surface heating, and the onshore flow became backed to the east once again. By 1800 utc it became increasingly difficult to identify the CZ as the surface winds had broadly reverted back to a northeasterly in most areas, as was the case during the morning. It seems the eastwards surge of the CZ and subsequent weakening, coupled with a reduction in air temperature in all areas, ultimately curtailed new thunderstorm development and the existing storms gradually decayed over the following hours. As highlighted in Figure 5, the storm was very localised with rainfall accumulations largely restricted to rural areas. Some residents of Stowmarket commented on social media that they had heard thunder for several hours but barely received a drop of rain. The highest rainfall measurements in the vicinity from the official Environment Agency network for the 24 h ending 0900 utc 26 July 2021 were 86.7mm at Buxhall, 83.0mm at Felsham and 55.5mm at Cockfield (Rookery Farm). Given no obvious rainfall detected by radar in the morning, and again during the overnight period, it is highly likely that all of these readings represent rainfall received from thunderstorms during the afternoon and early evening hours on 25 July. It is clear from Figure 5 that Brettenham was the epicentre. A farmer on the southwest side of the village, who had previously been a voluntary observer for the Environment Agency from 1997 to 2013 and has continued to keep rainfall records with the same equipment since (Figure 6), measured 181.3mm for the day. A neighbour with an AWS approximately 0.1km to the northeast recorded 161.0mm, although this figure may have been higher in reality considering tipping bucket rain gauges can underestimate accumulations during high rain rates (Burt, 2005). Another farmer at Pound Farm, ~0.8km to the southwest, recorded an estimated 200mm, using a 105mm manual rain gauge that had to be emptied twice – the first time it had overflowed, and the second occurrence had not quite reached full capacity. Given these broadly similar observed values in close proximity to each other, and the equipment used, suggests the 181.3mm is likely reliable. Another (former) resident of Brettenham kindly provided monthly rainfall measurements that both he and his father had collected in the village from 1955 to 2015, including a spell at Old Buckenham School for Anglian Water. The most recent 30-year average for July from this dataset (1986–2015) is 59.0mm, and during the whole 61-year record only one month produced rainfall higher than July 2021: June 1985 with 191.0mm. Unfortunately, daily figures were not available; however, analysis of HadUK-Grid (Hollis et al., 2021) for the nearest grid point suggests that this rainfall was likely spread over multiple days through the month. As such, rainfall of the intensity experienced on 25 July 2021 has never been recorded in Brettenham before (at least not since 1955). Estimates from Figure 5 suggest that the area of 100mm accumulations or higher was approximately 6.3km at its widest, creating some steep rainfall gradients in the vicinity. An unofficial gauge to the north of Hitcham, 2.1km to the southeast of Brettenham, measured 43mm, suggesting a gradient of ~66mmkm−1. This is comparable to the south Norfolk storm on 16 August 2020 (Holley et al., 2021). According to the local parish council, at least 28 dwellings experienced some internal flooding, mostly due to run-off from roads or fields, including the local school. The intensity of the rain, accompanied by small hail, was said to have flattened some crops, in particular a narrow strip of land running from Hitcham through Brettenham and Dux Street towards Felsham, according to another local farmer. Several large diameter trees were also reportedly knocked down – all occurring at the beginning of the crucial harvest period. According to Suffolk Fire and Rescue Service, 37 incidents were reported related to flooding in the Haverhill area between 1545 utc and 2100 utc (Suffolk County Council, personal communication, 15 September 2021) from a separate thunderstorm complex well to the west. However, despite 83 cloud-to-ground lightning strikes detected in the same domain as Figure 5 by the Météorage network 2 , no lightning-related incidents were reported. An attempt has been made to verify the duration in which the rain fell at Brettenham in Figure 7. Since a complete set of high-resolution temporal data is not available from a rain gauge, composite radar reflectivity is used as a substitute for both the grid cell with the highest estimated accumulation (r max = 148.4mm, located to the north of Brettenham and therefore slightly displaced to the northeast of the highest observed values), and the nearest grid cell to the Snowdon rain gauge (r near = 100.4mm, on the south side of the village). Figure 7 suggests that there was a short-duration shower broadly between 1430 and 1500 utc, with an accumulation of 0.8mm at the nearby AWS. This was not noted by the farmer, and so it is possible this small amount of rain may need to be subtracted from the final event total. The exact start time of the main thunderstorm varies depending on data source; the observer in the village reported the event began at 1540 utc and ended at 1720 utc, however data from the AWS suggests rain had already accumulated between 1529 and 1534 utc (since the data were largely uploaded in 5min intervals). A short power cut occurred in the village, apparently only lasting 2–3min, but there was a prolonged data outage at the AWS with no observations between 1606 utc (when it had recorded 62.2mm) and 1729 utc (when the full event total of 161.0mm was reported). Regular observations from the AWS were restored thereafter but with no additional rainfall, suggesting the rain had ceased at or before 1729 utc. Therefore, the total duration of the event was probably between 100 and 120min – but no more than 2 h – and is likely to have produced around 180.5mm when taking into account the earlier shower. According to radar data for r max (Figure 7), the storm reached a peak rain rate of 210mmh−1 at 1610 utc (perhaps contaminated somewhat by small hail), with rain rates exceeding 100mmh−1 noted on 11 separate scans in total, of which 9 were consecutive suggesting that rainfall of at least this intensity was maintained over a 45min period (assuming no weakening between each 5min scan). It is worth noting that the AWS had recorded more rainfall prior to the power cut than either r max or r near at the same time, suggesting that the composite reflectivity product had under-estimated the intensity of the rain. The highest 1-h accumulation estimated by r max is 108.8mm, occurring between 1545 utc and 1645 utc (and 93.4mm for a whole hour ending at 1700 utc). However, if a simple scaling factor is applied such that the event total radar estimate is scaled to the observed Snowdon gauge value, then the highest 1-h accumulation could in theory have been closer to ~134mm. Both of these potential values exceed the current UK 1-h rainfall record of 92mm at Maidenhead on 12 July 1901 3 (Meteorological Office, 1926; Ross et al., 2009). Other notable short-duration events in the UK, over similar durations and based on autographic records, include 90mm in 55min at Eskdalemuir, Dumfries-shire, on 26 June 1953 (Meteorological Office, 1953) and 86mm in 60min at Lesnewth, Cornwall on 16 August 2004 (Burt, 2005). ‘Eye’ observations, but with good supporting evidence, also indicated falls of 110mm in 58min at Wheatley, Oxfordshire on 9 June 1910 (Webb, 2011) and 97mm in 45min at Orra Begg, Northern Ireland on 1 August 1980 (Woodley, 1981). It is plausible that the rainfall at Brettenham also challenges the UK 2-h record; the 193mm at Walshaw Dean Lodge (West Yorkshire) on 19 May 1989 (Acreman, 1989; Collinge et al., 1990) is disputed by some (as detailed in, for example, Collier (1991)), and so the next highest accepted 2-h rainfall record is currently 155mm in 109min at Hewenden Reservoir (also West Yorkshire) on 11 June 1956 (Collinge et al., 1992). Even accounting for a small reduction to the event total for the preceding shower, the rainfall at Brettenham would also exceed this 2-h record for the amount of rain and possibly also the duration in which it fell. Estimates from satellite data indicate that cloud top temperatures were as low as −52°C, which would suggest cloud heights peaked at ~34 000ft (~10 400m). Assuming the ~1100Jkg−1 convective available potential energy (CAPE) estimated in Figure 2 was evenly distributed through the troposphere and air parcels started from the surface at rest, a mean vertical velocity of 11.7ms−1 is calculated when taking into account entrainment of surrounding air and effects of precipitation loading by halving the theoretical maximum updraft speed at the equilibrium level (w max = √(2 × CAPE)). Therefore, an air parcel from the boundary layer would in theory reach the tropopause in just under 14min, resulting in a rain rate of very approximately 140mmh−1 assuming all precipitable water vapour (PWV) available in the column (estimated at 33mm in Figure 2) was condensed out and fell to the ground as precipitation. In reality, this rain rate may be closer to 70mmh−1 since typically 50% or more of precipitation tends to evaporate into the surrounding air as it descends, especially when the air is rather dry. This event is remarkably similar to 16 August 2020 in south Norfolk, when a multicell thunderstorm cluster produced as much as 240mm of rain and set a new UK August daily rainfall record (Holley et al., 2021). Both events occurred in an environment with relatively high PWV, slow storm motion and enhanced low-level convergence aiding continuous backbuilding. Synoptic surface winds were northeasterly in both events within a slack surface pressure pattern, and while winds aloft were slightly different both days involved a mid/upper-level low that had originated from the Bay of Biscay. Aside from subtle differences between the various indices listed in this paper and Holley et al. (2021), the main difference between the two events is that the Brettenham storm was of much shorter duration than south Norfolk in 2020, but still produced impressive rainfall accumulations nonetheless. Clusters of heavy showers and thunderstorms affected portions of East Anglia, central southern and southeast England on 25 July 2021, producing local flash flooding in places. Of particular interest was a quasi-stationary multicell thunderstorm that affected parts of mid-Suffolk, especially the village of Brettenham where 181.3mm of rain was recorded by a storage rain gauge that was formerly part of the Environment Agency registered network, most of which fell in less than 2 h. The thunderstorms developed along a marked CZ where a northeasterly synoptic surface wind met an east or southeasterly onshore wind, likely enhanced by a sea breeze. Light steering winds aloft, combined with backbuilding and high tropospheric moisture, resulted in prolonged torrential downpours and flash flooding in the village. The rainfall observed, especially when considering the relatively short duration, is unprecedented in the local area and potentially challenges some UK records. The authors would like to thank colleagues at Weatherquest for their comments and access to various resources, Neil Klotz (Environment Agency) and local residents and farmers, in particular Roger Bere, for providing rainfall measurements and information on impacts in the area. We are also very grateful to Weather Underground for enabling use of home AWS data in our mesoanalysis, and more especially the owners of the private AWS. Finally, special thanks are offered to the anonymous reviewers for their useful feedback on the manuscript.