Adapting methods from , MHWs have been detected when the TOPAZ daily sea surface temperature (SST) exceeds the 90th percentile for at least 5 consecutive days, allowing no more than two days below the threshold within the 5 d. Successive events separated by gaps of 2 d or fewer were considered part of the same MHW. MHW events and their characteristics were determined using the Python marineHeatWaves module (https://github.com/ecjoliver/marineHeatWaves/tree/master, last access: 9 February 2026). According to , the baseline SST climatology for the percentile should be based on at least 30 years of data. For this study, the 90th percentile threshold has been calculated for each grid point of each calendar day of the year using daily temperature data over 32 years (1991–2022) as a fixed climatological baseline. For each calendar day, temperatures from a 5 d window centred on that day, were used to determine the percentile. suggests that daily, threshold time series may need to be smoothed to extract a useful climatology from inherently variable data. Consequently, we applied a 31 d moving window to smooth the 90th percentile. MHWs were detected for the Svalbard West region by averaging SSTs over the region and identifying periods when this regional mean exceeded the 90th percentile.

As in , each MHW event is described by a set of metrics . Mean intensity (°C) is the average SST anomaly (SSTA) over the duration of the event. Maximum intensity (°C) is the highest SSTA during an event. SSTAs were calculated relative to the reference period 1991–2022. The event peak indicates the date of peak intensity. Cumulative intensity (°C d⁻¹) is the accumulation of SSTAs associated with MHWs for each year . The duration (in days) of a MHW is calculated as the time interval between the start and end times and frequency (in events) is the number of events that occurred in each year. To generate maps of MHW metrics in Svalbard West and surrounding seas, MHWs were detected individually for each grid point in the seas surrounding Svalbard. For this analysis, MHWs were not analysed north of the sea ice edge (sea ice concentration ≥ 15 %).

The mean start date for MHWs in the Arctic Ocean (1982–2020) has been reported to be in August . MHWs during summer can have a higher ecosystem impact compared to winter events. When a summer MHW occurs on top of already high summer ocean temperatures, the additional warming can push species toward or beyond their thermal limits . Hence, we focused on MHWs in Svalbard West initiated during the Arctic summer period from June to September.

Multiple studies use a 5 d criteria to define MHW events in the Arctic ; however, longer-duration events are shown to potentially have more significant ecological consequences. For example, a study by on Arctic zooplankton, showed that continuous exposure to high temperatures was more harmful than intermittent short periods. In the study, continuous exposure to high temperatures for 9 d led to a more significant reduction in survival compared to two separate 3 d heat-stress events. As a result, for the analysis of individual MHW events in Svalbard West, we selected prolonged MHW events lasting at least 10 d.

In this study, we have also detected subsurface MHWs in Svalbard West by determining to which depths the mean temperature profile for Svalbard West exceeded the 90th percentile during each detected surface MHW event. Events shallower or equal to 50 m were defined as shallow events. Those that extended from the surface to deeper than 50 m were defined as deep events. It is important to note that since we only detect events at the surface, this approach may overlook events without a surface expression.

Each MHW event in Svalbard West is assigned a category defined by based on its intensity. Each category is determined by the degree to which temperatures exceed the local climatology, by looking at multiples of the 90th percentile difference (2× twice, 3× three times, etc.) from the mean climatology. The categories are as follows: moderate (Category I), strong (Category II), severe (Category III) and extreme (Category IV). Assigning categories to MHW events can be extremely useful as it enables comparison of events across different regions and facilitates communication among experts and the general public .

Using methods adapted from , the ocean heat budget was determined for Svalbard West (77–80° N, 5–15° E). Considering a control ocean volume with surface A and vertical section S, where mass and salinity are conserved, the heat budget is given by the balance between advective and vertical heat flux terms. As in , we omit the lateral heat diffusion term as this is negligible compared to the surface and advective flux terms : 1∂Q∂t︸Qt=ρ0cp∫SVTdS︸OHT+∫AQsdA︸SHF, Qt is the ocean heat content tendency; OHT represents the advective ocean heat transport through S; cp is the specific heat capacity of seawater (3987 J (kg °C)⁻¹) and ρ0 is the density of seawater (1000 kg m⁻³); V and T represent the cross-sectional velocity and potential temperature, respectively; SHF indicates the net sea surface heat flux, Qs, over the surface A.

The net ocean heat transport into the Svalbard West domain is computed along each boundary – facing north, south, west, (the eastern boundary is excluded from the ocean heat transport calculation as it is bounded by land) at daily frequency, 2OHT=ρ0cp∫-z(λ)η∫λ1λ2V(T-Tref)dzdλ, where λ1 and λ2 are the coordinates of the section line, η is the sea-surface elevation (the value of the vertical coordinate z at the ocean surface) and z(λ) is the depth at each section (down to the ocean floor). A reference temperature (Tref) of 0 °C was used. We consider the net SHF term over the Svalbard West box area. SHF is represented by the surface downward heat flux in seawater product in TOPAZ and consists of combined solar irradiance, sensible heat flux, latent heat flux and longwave radiation. The OHT was calculated using this code – https://github.com/nansencenter/NERSC-HYCOM-CICE/tree/master/hycom/MSCPROGS/src/Section (last access: 9 February 2026).

We computed the daily ocean heat content (OHC) of the upper 300 m using methods adapted from , 3cpρ∫z1=-300mz2=0mTzdz, In the TOPAZ reanalysis, there are 22 vertical levels in the upper 300 m. Integration of the OHC from a depth of 300 m was chosen to ensure the incorporation of the Atlantic Water layer (AW). The intermediate depth layer, 0–300 m, is characterised by the greatest AW warming in the Arctic Ocean .

The TOPAZ reanalysis provides gridded data for the North Atlantic and Arctic region with a spatial resolution of 12.5 × 12.5 km . We analysed the period 1991–2022. The dataset uses 40 vertical levels in the ocean and variables are interpolated to these depth levels. The minimum and maximum depth of the layers are 0 and 4000 m respectively. The reanalysis data is a product of the Arctic Monitoring and Forecasting Centre and contains daily, monthly and yearly mean fields of the following variables: temperature, salinity, sea surface height, horizontal velocity, sea ice concentration, surface heat flux and sea ice thickness. For the fields above, we use re-gridded output downloaded from Copernicus Marine Services (https://data.marine.copernicus.eu/product/ARCTIC_MULTIYEAR_PHY_002_003/download, last access: 9 February 2026). However, for calculating ocean heat transport as part of the ocean heat budget, we use data from the original model grid (12.5 × 12.5 km).

The dataset is based on the latest reanalysis produced by the coupled ensemble data assimilation system – TOPAZ4 10.48670/moi-00007. The TOPAZ system is based on the Hybrid Coordinate Ocean Model (HYCOM, ) coupled to an EVP sea-ice model . TOPAZ is forced at the ocean surface with fluxes derived from 6-hourly atmospheric fluxes from ERA5 atmospheric reanalysis from ECMWF,. TOPAZ uses the deterministic version of the Ensemble Kalman filter DEnKF, for data assimilation. This data assimilation includes a 100-member ensemble production. Observations assimilated by TOPAZ include: SST from Operational Sea Surface Temperature and Sea Ice Analysis (OSTIA), along-track sea level anomalies from satellite altimeters, CS2SMOS ice thickness data, sea surface salinity based on the SMOS satellite (assimilated from 2013–2019), ice concentrations from OSI-SAF and in-situ temperature and salinity from hydrographic cruises and moorings collected from main global networks Argo, GOSUD, OceanSITES, World Ocean Database,.

To assess TOPAZ's capability in representing the hydrography close to Svalbard, we compared temperature data from observations to TOPAZ output. The observational datasets used for TOPAZ evaluation are described below; satellite and mooring data were used. Mooring data was interpolated to match TOPAZ vertical levels and daily averages of the resultant time series were obtained for plotting. For comparison with the mooring data, we chose a grid point from TOPAZ close to or at the location of each mooring. Temperature data from TOPAZ was compared to each mooring at the shallowest and deepest levels with sufficient valid mooring data to evaluate how TOPAZ represents the water column. The Pearson Correlation Coefficient (r) with TOPAZ data was calculated for several depths for each mooring. Where possible, at each chosen depth, we also show the correlation between monthly anomalies to ensure the correlation is not solely based on similarities in the seasonality between TOPAZ and the moorings.

Evaluation of TOPAZ showed a strong positive correlation between daily temperatures from the ISM mooring and TOPAZ TP1 at 50 m (r=0.75, p<0.05, Fig. S1a in the Supplement) and a slightly weaker correlation for monthly anomalies (r=0.72, p<0.05, Fig. S1b). The high correlation between TP1 and the mooring is demonstrated in Fig. S2; a moderate/strong correlation with the ISM mooring is also shown across Svalbard West. At 150 m, the correlation was lower (r=0.63 for daily averages and 0.62 for monthly anomalies, p<0.05, Fig. S3a, b). The climatology in TOPAZ and the ISM mooring were also compared for the period 2006–2022. At 50 m, there was an offset of 0.02 °C between the mooring and TP1, with warmer temperatures in TOPAZ. At 150 m, the offset was larger at -0.52 °C with warmer temperatures shown in the mooring.

At the location of the YPM mooring, a moderate correlation was found at 70 m (r=0.48, p<0.05) and 500 m (r=0.62, p<0.05, Fig. S4). In the east, TOPAZ showed a moderate, negative correlation with Storfjorden M1 and M2 moorings at 50 m (r=-0.35, p<0.05, Fig. S5) and a moderate, positive correlation with the M4 mooring at a bottom depth of 60 m (r=0.52, p<0.05, Fig. S6). Furthermore, compared to the Storfjorden M1 and M2 moorings, TOPAZ temperatures were too low in summer/autumn. In contrast, compared to the M4 mooring, TOPAZ temperatures were higher than observed during summer.

Storfjorden (M1–M2) and M4 experience intense water mass transformation due to sea ice freezing and the Storfjorden mooring is situated in a productive polynya. At both M1–M2 and M4 during winter, TOPAZ could not resolve the cooling processes related to ice formation and temperatures did not reach freezing as observed in the mooring data (Figs. S5, S6). Due to the short time series, monthly anomalies were not determined for these moorings. In summary, TOPAZ is shown to perform best west of Svalbard with limitations east of Svalbard.

Based on Fig. , Svalbard West (black dashed box) has experienced a clear increase in both the frequency and duration of MHW events. The MHW intensity in this region is overall high compared to other regions, especially over the West Spitsbergen Shelf (Fig. ) and is highest during summer (Fig. ). Analysing the SST averaged over Svalbard West (77–80° N, 5–15° E), we again see that the number and duration of MHWs has increased in recent decades, with a particular increase observed after 2011 (Fig. ). Between the periods 1991–2010 and 2011–2020, for the Svalbard West region, the mean MHW frequency has increased from 2 to 3 events per year, whilst the mean MHW duration has increased from 10 to 24 d. Little change is observed in the intensity of MHWs between the two periods with a decrease of 0.02 °C in 2011–2022 compared to 1991–2010. Before 2011, MHWs are only observed in 1991, 2006 and 2007 during summer (Fig. ). After 2011, events are shown to occur throughout the year, particularly in 2016. There is an overall increase in SST anomalies over the period of the reanalysis and long-lasting MHW events are paired with high SST anomalies of around 2 °C.

To quantify whether the detected MHW events are forced at the surface by air-sea heat fluxes or forced through increased ocean heat transport (OHT), the SHF and OHT for the Svalbard West region have been analysed for each MHW event.

In this study, we applied methods from , whereby MHWs are detected when the daily sea surface temperature (SST) exceeds the 90th percentile for at least 5 consecutive days, with no more than two below-threshold days. Additionally, we used a fixed baseline of 32 years (1991–2022) to calculate the percentile. The same approach is used by , when studying MHWs in the Barents Sea. A fixed baseline provides a stable reference for detecting long-term SST trends in MHW studies. While a fixed baseline is commonly used, some advocate for a shifting baseline to exclude the effects of climate change . and , however, have advised against using a shifting baseline as updating the baseline climatology over time can change how past events are classified. , compared a shifting baseline approach to a fixed baseline approach to study MHWs in the Barents Sea, and found that using a shifting baseline approach decreased the intensity of MHW events compared to a fixed baseline case.

Baseline length can also affect MHW metrics. In our study, when the baseline was adjusted to the last 10 years of the reanalysis (2011–2022), the timing of most summer MHW events remained largely unchanged (Table S1). However, a decrease in duration and intensity was found when the baseline was shortened. also found a general trend of decreasing average intensity with decreasing length of the baseline. In addition, we found that the decrease in duration was larger for deep events compared to shallow events. The above differences in MHW characteristics caused by changing the baseline highlights the need, as emphasised by , for the definition of MHWs to be standardised.

The MHW research community also debates whether temperature data should be detrended prior to MHW detection. Detrending, often by removing a linear trend , aims to separate MHWs from the long-term warming signal and is used in several global and Arctic studies . In the Arctic, warming is not necessarily linear and is strongly influenced by natural variability, including decadal thermohaline anomalies in the North Atlantic-Arctic region , that modulate Atlantic inflows to the Nordic Seas on multi-year to decadal scales . Such variability might explain part of the decadal variability in MHW frequency for the northern Nordic Seas (Fig. S7). Given these complexities, we argue that using raw (non-detrended) temperature data is more appropriate for Arctic MHW detection.

We find regional differences in MHW activity across the Svalbard Archipelago and its surrounding seas (Fig. ). The highest MHW intensity is located at water mass fronts. For example, high MHW intensity is found in the location of the Polar Front, southeast of Svalbard description of the Polar Front is given in. The above could be attributed to the high variability of SST in this region . High MHW intensity, including high frequency and duration in the last decade, is also found southwest of Svalbard along the Mohn Ridge, in the pathway of the NwAFC (Fig. ) in the Arctic Front. A high number of MHWs at the Mohn Ridge could be attributed to high seasonal and interrannual temperature variability in the region or due to strong temperature gradients, associated with the interaction of warm and saline Atlantic Water with colder and fresher Arctic Water . Lastly, lateral exchanges, such as eddies across the front, could trigger MHW events by bringing warmer water into a normally colder water region.

Large changes in MHW frequency and duration is found on the West Spitsbergen Shelf (Fig. ). One reason for this could be the shoaling of Atlantic Water (AW) associated with the Atlantification seen in western Svalbard fjords and north and west of Svalbard. Atlantification refers to the increased influence of AW in the Arctic driven by recent warming of the AW inflow . AW has been observed higher in the water column and along shallower isobaths in Isfjorden and Kongsfjorden for location see Fig. . North of Svalbard, Atlantification can be observed in the eastern Eurasian Basin, as evidenced by a weakening of the halocline and a shoaling of the intermediate-depth AW layer . Furthermore, on the shelf southwest of Svalbard, reported a 8 % yr⁻¹ increase in the volume fraction of AW. Since AW inflow along the West Spitsbergen shelf is warming , a shoaling and increased presence of warming AW can lead to higher MHW frequency and duration.

The variability of AW temperature west of Svalbard has been shown to be associated with the strength of the Greenland Sea Gyre (GSG) circulation influenced by the anomalous wind stress curl over the Nordic Seas . A stronger GSG circulation increases the AW flow speed west of Svalbard, leading to increased oceanic heat content and higher AW temperature therein . Thus, increased AW temperature driven by strong GSG circulation may also be responsible for the increase in MHW events in Svalbard West.

MHWs in Svalbard West exhibited substantial spatial variability, ranging from local events confined to the archipelago to widespread events extending beyond Svalbard, and from shallow surface anomalies to depths reaching up to 600 m (Figs. , ). Previous studies, for example, , define the vertical structure of MHWs by averaging temperature anomalies from in-situ profiles over the event duration. Instead of averaging, we examined the day-to-day vertical progression of each MHW by identifying the depth levels exceeding the 90th percentile on each day of the event. Understanding the horizontal and vertical extent of individual MHW events, represents a novel set of criteria that provide a valuable framework for future MHW characterisation.

As noted in Sect. 2.1, since MHWs in this study are detected using SST, our methods overlook events that lack a surface expression. Global studies have shown that a significant proportion of MHWs occur in the subsurface without leaving a detectable SST signature ; this therefore represents an important limitation of our study. Studies on subsurface MHWs in the Arctic are limited. detected bottom MHWs in the Barents Sea, which were shown to have a longer duration than surface events. Bottom MHWs could thus imply a greater potential for ecosystem damage, underscoring the need for future studies to consider events that do not exhibit a surface signal. have proposed a classification scheme for subsurface MHW events that distinguishes between mixed‐layer, deep, thermocline, full‐depth, submerged and benthic MHWs. Applying such detailed classification to describe the vertical extent of events can improve our understanding of ecosystem impacts, by helping to identify which species and habitats are most likely to be affected.

Our results show that advective heating is the primary driver of most summer MHWs in Svalbard West, with positive OHT anomalies exceeding SHF anomalies in all but two events (Fig. b). Poleward heat transport by boundary currents can drive subsurface warming, and through vertical mixing can elevate SSTs and produce a MHW . During all the events identified in this study, mean OHT into Svalbard West across its southern boundary exceeded the combined outflow through the northern and western boundaries (Table S2). This net heat gain can elevate temperature anomalies and ultimately drive MHW development. Furthermore, all events show a positive anomaly at the southern boundary (Fig. a), implying anomalous heat input into the region that contributes to the development of MHWs. In contrast to our results, Arctic-wide assessments identify surface heat fluxes as the dominant driver of summer MHWs , with lateral advection generally considered as a secondary driver. However, also highlighted that OHT can become the leading contributor to MHWs at the main Arctic gateways, where heat advection is particularly pronounced. The importance of advection in driving MHWs in our region is further supported by , who found that enhanced AW transport contributed to the onset of the 2016 Barents Sea MHW.

Despite the leading role of OHT in driving MHWs in Svalbard West, atmospheric heating also contributes to their development. Positive mean SHF anomalies are found for all but two events (Fig. b), indicating anomalous heat input from the atmosphere during the majority of the MHWs. Negative anomalies are shown for the 2011 and 2019 MHW, which also display the smallest horizontal extent compared to the other events. Positive SHF anomalies are also found at the start date of each MHW, indicating enhanced heat input for events initiated in summer and reduced heat loss for those initiated in late summer/early autumn (Fig. ). For the events with a positive mean SHF (five out of eight events, Table S2), atmospheric heating contributes to a net heat input at the ocean surface, which directly warms the upper ocean. Anomalous atmospheric heating can trigger or enhance SST anomalies, contributing to MHW development. SST anomalies driven by an anomalous SHF can, through vertical mixing and entrainment, lead to positive subsurface temperature anomalies and contribute to driving deep MHW events . The atmospheric heating effect is further intensified by oceanic heat advection, as described above. In previous studies, an interplay between the ocean and atmosphere has proven to be important for the onset of MHW events, for example in the Barents Sea . In addition, the 2016 event identified in this present study has been linked to atmospheric forcing: suggested that warm air temperature anomalies of approximately ≥2 °C played a key role in its development over the Barents Sea, while reported that reduced heat loss to the atmosphere during the winter of 2015–2016 contributed to its onset.