Took me a minute... if you use Windowing functions (requires SQL 2012 or later), then this is easy. The PARTITION basically groups the records together... then you can use ROW_NUMBER() to number the members in each group, and then take the first one...

/* create the table -- since I'm grouping by the Date, I split EventDate and EventTime into separate columns */

CREATE TABLE #WeatherEvent (

EventDate DATE

,EventTime TIME

,Reading DECIMAL(3,1)

,ReadingID INT IDENTITY(1,1)

);

GO

/* insert some data... note the dates with multiple readings */

INSERT INTO #WeatherEvent (EventDate, EventTime, Reading)

VALUES ('11/19/2016', '10:10:26 PM',55),

('11/19/2016', '9:10:26 AM',65),

('11/20/2016', '1:21:26 AM', 54),

('11/20/2016', '4:45:22 PM', 72),

('11/21/2016', '1:02:27 AM', 49.9),

('11/22/2016', '08:00:26 AM', 50.3),

('11/22/2016', '4:45:32 PM', 90.3),

('11/23/2016', '10:25:26 AM', 43.8);

Read up on ROW_NUMBER() and windowing functions... Itzik Ben Gan's book is great if youre ready for it.

SELECT EventDate

, EventTime AS DailyHighTime

, Reading As DailyHigh

FROM (

SELECT EventDate

, EventTime

, Reading

, ROW_NUMBER() OVER (PARTITION BY EventDate ORDER BY Reading DESC) As ReadingID

FROM #WeatherEvent) we

WHERE we.ReadingID=1;