Systematic instrumental weather recording (e.g. using thermometers, rain-gauges etc.) began in the days when the only way such data could be obtained & logged was by someone reading & noting same. For this reason, the meteorological 'day' tended to be built around the 'human-activity' day. With most observers being available 7 to 9am local time, the climatological start/end for the day has settled in many European countries (though not all), to 0900 UTC (or to 0900 local/clock-time elsewhere). For those stations only noting maxima, minima, rainfall etc. once per day, the maximum temperature read at 0900UTC is credited to the previous day, on the assumption that on the majority of occasions, the highest value would have happened during that day's afternoon. The minimum is credited to the current day, on the assumption that, again for the most part, the lowest temperature would have occurred in the hour or so around local dawn - i.e. the day of reading the thermometer. Rainfall for this 09-09 period is apportioned to the day occupying the greatest part of the 24hr, that is it is 'thrown-back' to the previous day's date. This means of course that a severe thunderstorm producing a lot of rain in the period after midnight will have it's rainfall credited to the previous day!
A better way of noting (and recording) these data is to further divide the 'climatological' day into two periods: 0900 to 2100 & 2100 to 0900 UTC. This of course means that thermometers and rain-gauges have to be read, reset/emptied at 2100UTC. However, as with the 09-09 method, anomalies will occur. For example, if a major air-mass change occurs after 2100UTC (e.g. cold to warm in winter), then the highest temperature in the 24hr period may occur overnight, yet the logged maximum will be that recorded 0900 to 2100. And rainfall will still be credited to the date at the start of each period, i.e. the 21 - 09 UTC rainfall (read at 0900UTC) is again 'thrown-back' and added to the 09-21 total.
A further complication arises from the requirement of those providing information for use in SYNOP bulletins. As there is no 'main' SYNOP at 0900UTC, in Europe the night minimum (18 to 06) is reported in the 2-group of the 333 section of the 0600 SYNOP. (Note that during the long winter nights, the actual minimum could well occur after 0600UTC.) The day maximum (06-18) is reported in the 1-group of the same section at 1800 UTC. The rainfall (or melted snowfall) at these (and other times), is recorded using the 6-group in the main SYNOP (see here).
With the advent of high-quality electronic systems, it is very easy to log data according to the 'normal' day, 00-24 UTC, and users of such find it cumbersome to comply with the above scheme(s). This FAQ makes no comment upon this debate, except to note that all methods of recording climatological data throw up anomalies, and there is a large body of historical data tied to the 09-09, or 09-21-09 standard. The important point, especially when noting extreme temperatures and notable rainfall, is to annotate your own records (and report same where possible), so that researchers in years to come can pick out the deviations from the 'assumed' diurnal cycle of temperature, or the 'noteworthy' precipitation event.