Longest consecutive period above threshold using rle and for loop - r

Question

I have four years of streamflow data for one month and I'm trying to figure out how to extract the longest consecutive period at or above a certain threshold for each of the four years. In the example below, the threshold is 4. I want to try to accomplish this using a for loop or possibly one of the apply functions, but I'm not sure how to go about it.
Here's my example dataframe:
year <- c(rep(2009,31), rep(2010, 31), rep(2011, 31), rep(2012, 31))
day<-c(rep(seq(1:31),4))
discharge <- c(4,4,4,5,6,5,4,8,4,5,3,8,8,8,8,8,8,8,1,2,2,8,8,8,8,8,8,8,8,8,4,4,4,5,6,3,1,1,3,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,9,10,3,3,3,3,3,3,1,1,3,8,8,8,8,8,8,8,8,8,1,2,2,8,8,3,8,8,8,8,8,8,4,4,4,5,6,3,1,1,3,3,3,3,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,9,3)
df<-data.frame(cbind(year, day, discharge))
df$threshold<-ifelse(discharge>=4,1,0)
In this example, the threshold column is coded as 1 if the discharge is at or above the threshold and 0 if not. I'm able to partially get my desired output for one year (2009 in the example below), with the following code:
rl2009<-with(subset(df,year==2009),rle(threshold))
cs2009 <- cumsum(rl2009$lengths)
index2009<-cbind(cs2009[rl2009$values == 1] - rl2009$length[rl2009$values == 1] + 1,
cs2009[rl2009$values == 1])
df2009<-data.frame(index2009)
df2009 #ouput all periods when flow is above threshold
df2009$X3<-df2009$X2-df2009$X1+1
max2009<-df2009[which.max(df2009$X3),]
max2009 #output the first and longest period when flow is above threshold
For 2009, there are three time periods when the discharge equals or exceeds 4, but the period from day 1 to day 10 is chosen because it is the first of the longest period above the threshold. X1 represents the start of the time period, X2 the end of the time period, and X3 the number of days in the time period. If there is more than one period with the same number of days, I want to select the first of such periods.
My desired output for all four years is below:
year X1 X2 X3
2009 1 10 10
2010 9 31 23
2011 10 18 9
2012 12 30 19
The actual data includes many more years and many streams, so it's not feasible to do this for each year individually. If anyone has any thoughts on how to achieve this, it'd be greatly appreciated. Thanks.

Answer 1

Simply, generalize your process with a defined function such as threshold_find and pass dataframes subsetted for each year into it which can be handled with by.
As the object-oriented wrapper to tapply, by slices a dataframe by one or more factors (i.e., year) and returns a list of whatever object the defined function outputs, here being the max dataframe. At end, do.call() row binds all dataframes in by list into one dataframe.
threshold_find <- function(df) {
rl <- with(df, rle(threshold))
cs <- cumsum(rl$lengths)
index <- cbind(cs[rl$values == 1] - rl$length[rl$values == 1] + 1,
cs[rl$values == 1])
df <- data.frame(index)
df$X3 <- df$X2 - df$X1+1
max <- df[which.max(df$X3),]
max
}
finaldf <- do.call(rbind, by(df, df$year, FUN=threshold_find))
finaldf
# X1 X2 X3
# 2009 1 10 10
# 2010 9 31 23
# 2011 10 18 9
# 2012 12 30 19