Review of „Long-term sea surface temperature time series from Malin Head, Ireland”, by Sarah Daves, Guy Westerbrook, Glenn Nolan, Rob Thomas, Eoghan Daly.

The manuscript carefully describes a very long (for ocean measurements) time-series of sea surface temperature measurements from a location close to Malin Head on Ireland. This is an important time-series since it documents in great detail warming trends and variability in the North Atlantic, and in particular the coastal Northern Ireland.

The main issue, or complication, with the time-series is the three different ways (segments) of doing the measurements have been used over time, i.e. evolved. Evolution to better measurements is obviously a good thing, but the authors have to do some adjustments to the measurements to make the data comparable. I think it is mostly transparent what they have done, and it appears to be a sensible approach.

Although there are adjustments to the diurnal cycle, I wonder if there is a bias from well (2 m depth), bucket (surface) and continuous (3-4 meters depth) in the time-series simply due to water depth? Was there, in addition to the diurnal signal, also a signal in temperature that was tide-dependent, i.e. the sensor being closer or further away from the sea surface?

The manuscript is well-written and document an important oceanographic time-series well, and I recommend that the manuscript should be published with minor adjustments.

Minor comments:

Figure 1: According to google maps, Malin Head is the northernmost point of contiguous Ireland, whereas the northernmost point of the republic of Ireland is the island of Inishtrahull. You might want to correct the legend of the figure.

Line 70: Well measurements: it is not clear how the measurements took place to me. Did the observers pump water through the pipe to the site 30 meters offshore, and measured the temperature there (I assume once enough water had been pumped to reflect ocean temperature)? Please explain.

Line 139: I might have missed it, but this is important: Did you adjust only the standardized data set for segment 2, or did you adjust also the full data set? Reading the ms and the conclusions makes me believe you only adjusted the standardized data set (which would be the correct way of doing this), but the wording on line 139 is not entirely clear about this. Please explain if adjustments were done to only the standardized data, or to both data sets.

Line 215: I suggest using the unit of °C/a for the annual warming (0.015)

Line 237: Why is there such a difference in warming trend for measured vs. standardized data? Please explain.

Line 300: Just out of curiosity, why is this data set called “Malin Head”, when it is actually from Portmore Pier? I am sure there are good reasons, but when reading about the tide-gauge station is called Portmore Pier, and I believe the temperature time series and the tide gauge is pretty much the same station, none of which are actually on the nearby Malin Head.

This paper describes a long and important series of measurements of near surface sea temperature recorded near Malin Head starting in 1958. Three periods of measurements are identified with different methods and sampling protocols and two datasets are presented: a subset of the measurements as recorded, and a “standardised daily dataset” which has attempted to account for the changes over time to provide a consistent 67 year daily record. The original and standardised data were easy to find and download as csv files.

My recommendation is that the paper may become suitable for acceptance following major revisions to the paper, and an enhancement of the data made publicly available.

Major comments

1) Not all of the observations described in the paper are available for download. The original data file should contain all of the measurements made from the pier in segment 3, including when the sensors were duplicated, and the 5 months of bucket sampled measurements made during 2010. A flag should be included indicating which sensor was used in the derivation of the standardised data.

2) To justify the adjustments made to the data to form the daily timeseries this paper needs to include a detailed comparison of the observations from each of the segments and each of the co-incident measurements (up to 3 for some periods). The size of the adjustments isn’t clear from this paper, but from the downloaded data seem to range up to a degree for segment 2, with a mean for that segment of about 0.5 degree. From the description in the paper segments 2 and 3 are located fairly close to each other, segments 1 and 2 were taken with similar methods but in different locations, segment 3 was taken at a fixed position relative to a varying sea level whereas segments 1 and 2 were likely to have been taken near the surface, and the sampling protocols were different for each of the segments (twice daily, daily and every 30 minutes). The relative importance of each of the different contributions to the overall adjustment needs to be shown (location, methods, water depth, diurnal sampling etc.). Looking at the data, alongside the description in the paper (mainly section 2.4), it’s not clear what the rationale is for the adjustments applied.

3) Evidence needs to be provided about the interaction between the diurnal variations and the tidal cycle to give confidence that this does not affect the adjustment approach. For example, are the derived diurnal cycles different when stratified according to the depth of the sensor/tidal level? Doing this might give more confidence in applying the diurnal cycle adjustment to the data from segment 1, and particularly segment 2, which will have likely been taken at more consistent depths relative to the sea surface.

4) The “error evaluation” in section 3.2 is actually an uncertainty analysis. The uncertainties should be presented separately for the raw data, and for the standardised timeseries. More use should be made of the measurements nearby in space and time, and comparison of the variability between segments, to inform the quantification of uncertainty. It isn’t clear from what is presented in this paper that the combination of the 3 segments into a daily time series is justified as an approach. Perhaps a monthly timeseries, or accepting that the 3 segments cannot robustly be combined might be a more appropriate outcome. Including a more thorough uncertainty analysis may help to show whether the adjustments and their uncertainty are sufficient to justify the generation of the long timeseries.

5) It was unclear to me what the purpose of the convolution procedure described at the end of section 2.4 was intended to achieve. Is half-hourly data not sufficient to produce a daily average? The convolution incurs data loss and makes the provenance of the record less clear. What is the justification for this step?

6) In section 4 there are 3 presentations of timeseries of the raw and standardised data, two for annual (Figures 3 and 5) and one for a seasonal decomposition (Figure 4). None of the trends presented have confidence intervals, which should include both the effects of autocorrelation in the time series and the uncertainty in the series.

Minor comments

The original data file contained a “depth” variable, which was 2.1m up to 2007 and 1m thereafter. This seemed odd compared with the description of the measurements which were described as 1) well measurements (1958-1991), 2) bucket measurements (1991-2007) and 3) continuous (2008 to present). The measurement depth for segment 3 would therefore have varied with tidal level. From an internet search the maximum tidal range seems to be about 4 metres. An explanation of the source of the depth information should be added, and ideally the depth of measurement for segment 3 should reflect the depth below the surface of the sensor.

Section 3, titled “data quality” is really about the integrity and processes of the data centre.

Section 3.1 “dashboard and visual qc” needs to provide more detailed information on the quality testing applied to the data. The SeaDataNet manual provides examples of qc methodologies, but is not prescriptive. The test limits, any climatology values and any other relevant information should be added.

The original data file for the period 2008-2012 includes a mix of data from mid-pier and end-pier. Only a single value has been included, and the only indication of which has been selected is a slight difference in the latitude and longitude values. It is stated that the end-pier is generally preferred as it is a more exposed location with a smaller diurnal cycle. However the paper does not include any information about of the differences in diurnal cycle between the two pier locations in the standardisation procedure described in section 2.4. As noted, all the measurements should be included in the original (raw) data record, with a flag to say which sensor was used in the derivation of the standardised data.