How do I select a certain part of NumPy array and then the next part?

I would suggest doing it in pandas (let's assume your array is named x):

df = pd.DataFrame(x, columns=['year', 'month', 'max_temp'])

max_temps_per_year = df.groupby('year')['max_temp'].max()

print(max_temps_per_year)

If you want a pure numpy implementation and assuming your array is named x, you could:

1. first use a logical condition to select only the valid rows of your array:

cond = x[:, 0] == 1990  # cond is True/False based on if the year is 1990
sub = x[cond]  # subset of x that has only the rows satisfying the condition

2. then calculate the max temperature for the subset

max_temp = sub[:, 2].max()  # maximum temperature of year

If you do these steps iteratively, then you should get the max temperature for each year:

years = np.unique(x[:, 0])  # find unique years
max_temps = [(x[:, 0] == year)[:, 2].max() for year in years]  # same as above written as list comp

# print the results:
print('Year | Max temperature')
for year, temp in zip(years, max_temps):
    print('{:^4} | {:^15}'.format(year, temp))

This should work..

#col index of the year field
y=0
#col index of the temperature field
t=2
#Taking unique Years
n=np.unique(arr[:,y])
#Taking max temp for each Year
x=[np.max(arr[arr[:,y]==i,t].astype(np.float)) for i in n]
maxarr=np.append(n,x).reshape(len(n),len(x))
print(maxarr)

Output

[['1990' '1991']
 ['25.0' '21.0']]