You'll have to introduce a day variable...by locating the positions where the hour returns to 0 if is always a 0-hour reading or when the diff() is <0 if not; then fill between those with either an arbitrary day (but sequential) number or an actual date if have it.
Once that's done,
[g,da,hr]=findgroups(M(:),M(:)); % grouping variable by day, hour
tot=splitapply(@sum,M(:,3),g); % sum by group
res=table(da,hr,tot); % assemble results in table
the above assumes the day number is prepended to M as the first column.
