AbstractA network of rain‐gauges has been optimized to reconstruct the spatially representative summer monsoon rainfall series (1871–1984) of India prepared by averaging the normalized rainfall (actual divided by mean) of 306 rain‐gauges. Three approaches have been attempted: (i) random selection, (ii) fitting a multiple linear regression and (iii) selecting objectively a subset of a few gauges, the mean of which shows the highest correlation coefficient (CC) with the all‐India series. By averaging a randomly selected 45 rain‐gauges, an average CC with the all‐India rainfall series ofca.0.94 was achieved, and by averaging 100 gauges the average CC wasca.0.97. By fitting a multiple linear regression between the all‐India series and the normalized rainfall series of 306 gauges, a maximum of 34 independent gauges (parameters) were identified and the estimated rainfall series showed a highest CC of 0.9969 with the full series. In the third approach, a subset of 34 rain‐gauges was identified following an objective criterion; the mean rainfall of this subset showed a CC of 0.9931 with the all‐India series. For routine updating of the all‐India summer monsoon rainfall series, however, 35 rain‐gauges were selected by applying the third approach to 64 randomly chosen India Meteorological Department (IMD) rain‐gauges better known for their timely availability of data. The reconstructed (CC = 0.9866) all‐India summer monsoon rainfall series for the peri