This is a collection of R code to process raw (ASCII) format data on climatic variables obtained from the Department of Hydrology and Meteorology (DHM), Babarmahal, Kathmandu, Nepal. While the raw data can be opened in any text editor and copied to a spreadsheet, it can be quite a hassle when dealing with decades of data from multiple stations.
These are the code that I am using for my own research to read the raw data in R, batch combine records for each station and output the data in csv file. This is a work in progress and I will continue to add the code for analysis as my research progresses.
No sample data is presented here as the data obtained from DHM cannot be shared publicly. The format for rainfall and temperature data is presented in the sections below with code sniplets to read and process the records.
{list of publications using parts of the code for climate data analysis}
For rainfall data, each file in the dataset consists of one year of rainfall records with each line represented by day of year (DOY) and rainfall for the corresponding DOY in millimetres (mm). The DOY ranges from 1 to 365 during non-leap years and from 1 to 366 in leap years. Missing data is denoted by string 'DNA' and 'trace amount of rainfall' is marked by string 'T'.
Format for rainfall data:
1 3.0
2 3.0
3 4.0
4 0.0
5 0.0
6 T
7 0.0
8 0.2
9 6.1
10 0.8
and so on
The following code reads rainfall record files of multiple stations and creates a continuous record for each station as csv file. Each station's record should be in a separate folder. The daily rainfall records obtained from DHM is named in the AS####YY.yy where #### is a four digit station code (leading 0 required for three digit codes), YY represents the first two digits of year and yy represents the last two digits for the year. For instance, a file AS142120.14 is a daily rainfall record file for year 2014 for station 1421 which is Gaida Kankai in Eastern Nepal.
The year number is extracted from the file. The output file name is extracted based on the folder name which are named in the format Rain#### where #### is a four digit station code. E.g. Rain0105 would be a folder containing daily rainfall records for Mahendranagar station (Station no. 105)
dirData <- "x:/metro_data/Rain" #directory which has folder(s)
dirList <- list.dirs(dirData, full.names=TRUE, recursive=FALSE) #list of folders
#* Function to read and append rainfall files
readRain <- function(xdir) {
xfilelist <- list.files(xdir, full.names=TRUE, pattern="AS")
rainrecx <- data.frame(do.call("rbind", lapply(xfilelist,
function(xfile) {
tempdata <- read.table(xfile, header=FALSE)
dayNos <- tempdata$V1
yearNos <- paste0(substr(xfile, nchar(xfile)-4, nchar(xfile)-3),
substr(xfile, nchar(xfile)-1, nchar(xfile)))
originDate <- paste0(yearNos, "-01-01")
varDate <- as.Date(dayNos - 1, origin = originDate)
varDate <- as.character(strptime(varDate, "%Y-%m-%d"))
varRain <- as.character(tempdata$V2)
varDay <- format(as.Date(varDate), "%j")
cbind(varDate, varRain, varDay)
})))
outputFile <- paste0(dirData, "/", substr(xdir, nchar(xdir)-3, nchar(xdir)),
"rain.csv")
write.csv(rainrecx, file = outputFile, row.names=F)
return(outputFile)
}
lapply(dirList, readRain) #apply the function to foldersThe same code with for loop.
#set the directory which has folders for each station
dirData <- "X:/metro_data/Rainfall"
dirList <- list.dirs(dirData, full.names=TRUE, recursive=FALSE)
#function to read, process and append record in each folder
readRain <- function(xdir) {
xfilelist <- list.files(xdir, full.names=TRUE, pattern="AS",
recursive=TRUE, include.dirs=TRUE)
rainrecx <- data.frame(do.call("rbind", lapply(xfilelist,
function(xfile) {
tempdata <- read.table(xfile, header=FALSE)
dayNos <- tempdata$V1
yearNos <- paste0(substr(xfile, nchar(xfile)-4, nchar(xfile)-3),
substr(xfile, nchar(xfile)-1, nchar(xfile)))
originDate <- paste0(yearNos, "-01-01")
varDate <- as.Date(dayNos - 1, origin = originDate)
varDate <- as.character(strptime(varDate, "%Y-%m-%d"))
varRain <- as.character(tempdata$V2)
varDay <- format(as.Date(varDate), "%j")
cbind(varDate, varRain, varDay)
})))
outputFile <- paste0(dirData, "/", substr(xdir, nchar(xdir)-3, nchar(xdir)),
"rain.csv")
write.csv(rainrecx, file = outputFile, row.names=F)
return(outputFile)
}
#apply read and write functon to all folders
lapply(dirList, readRain)Format of converted rainfall data csv file:
varDate varRain varDay
01-01-00 0 1
02-01-00 0 2
03-01-00 0 3
04-01-00 0 4
05-01-00 0 5
06-01-00 0 6
07-01-00 0 7
08-01-00 0 8
and so on
The first line of each of the files for temperature data consists of the corresponding 4 digit year number and the station name. The second line consists of column headings for maximum and minimum temperatures. From the third line onwards, each text file in the dataset consists of one year of temperature records with each line represented by DOY, maximum temperature and minimum temperature for the corresponding DOY in Celsius (°C). The DOY ranges from 1 to 365 during non-leap years and from 1 to 366 in leap years.
Format for temperature data:
1985 Temperature for <STATION NAME>
Tmax Tmin
1 33.0 4.0
2 30.5 10.1
3 26.0 10.8
4 27.0 20.0
5 28.0 12.5
6 29.0 5.0
7 22.5 5.5
and so on
The following code reads temperature record files of multiple stations from a folder and creates a continuous record for each station in csv format. Each station's record should be in a separate folder. The daily temperature records obtained from DHM is named in the TA####YY.yy where #### is a four digit station code (leading 0 required for three digit codes), YY represents the first two digits of year and yy represents the last two digits for the year. For instance, a file TA131120.14 is a daily temperature record file for year 2014 for station 1311 which is Dharan Bazar in Eastern Nepal.
The year number is extracted from the file. The output file name is extracted based on the folder name which are in the Temp#### format where #### is a four digit station code. E.g. Temp0105 would be a folder containing daily temperature records for Mahendranagar station (Station no. 105).
#set the directory which has folders for each station
dirData <- "x:/metro_data/Run4/Temp"
dirList <- list.dirs(dirData, full.names=TRUE, recursive=FALSE)
#function to read, process and append record in each folder
readTemperature <- function(xdir) {
xfilelist <- list.files(xdir, full.names=TRUE, pattern="TA")
temprecx <- data.frame(do.call("rbind", lapply(xfilelist,
function(xfile) {
tempdata <- read.table(xfile, header=FALSE, skip=2)
dayNos <- tempdata$V1
yearNos <- paste0(substr(xfile, nchar(xfile)-4, nchar(xfile)-3),
substr(xfile, nchar(xfile)-1, nchar(xfile)))
originDate <- paste0(yearNos, "-01-01")
varDate <- as.Date(dayNos - 1, origin = originDate)
varDate <- as.character(strptime(varDate, "%Y-%m-%d"))
varMax <- as.character(tempdata$V2)
varMin <- as.character(tempdata$V3)
#varDay <- format(as.Date(varDate), "%j")
cbind(varDate, varMax, varMin)
})))
outputFile <- paste0(dirData, "/", substr(xdir, nchar(xdir)-3, nchar(xdir)),
"temp.csv")
write.csv(temprecx, file = outputFile, row.names=F)
return(outputFile)
}
#apply read and write functon to all folders
lapply(dirList, readTemperature)Format of converted temperature data in csv file:
Date Max Min
01-01-00 23 9
02-01-00 22 11.5
03-01-00 18 7.1
04-01-00 20.9 7.5
05-01-00 20 12.5
06-01-00 20.5 4
07-01-00 20 7
and so on
List of meteorological stations in Nepal can be found at http://www.dhm.gov.np/meteorological-station
See https://github.com/asheshwor/R-maps/blob/master/01_simple-map.R for R code to plot meterological stations of Nepal.
One of the first we do once we get daily records is to check for days with no records. While there are many ways to count NA values in R, here's an example using ddply function from plyr package.
require(plyr)
temprec <- read.table("Dharan_Bazar_temp1.csv", sep="," ,
stringsAsFactors = FALSE)
temprec$Year <- strftime(as.Date(temprec$Date), format='%Y')
temprecna <- ddply(temprec, c("Year"), function(df) c(MaxNA = sum(is.na(df$Max)),
MinNA = sum(is.na(df$Min)),
MeaNA = sum(is.na(df$Mean))))
head(temprecna)Which results in the summary of NA values as a dataframe.
Year MaxNA MinNA MeanNA
1 2000 0 0 0
2 2001 0 0 0
3 2002 0 1 1
4 2003 0 0 0
5 2004 0 0 0
6 2005 0 0 0
The days between the first occurrence of at least 6 consecutive days with mean temperature > 20 and the first occurrence after July 1 of at least 6 consecutive days with mean temperature < 20.
gslm.this <- function(xdf) {
xlist <- sapply(xdf$Mean, function(x) x > 20)
ylist <- sapply(xdf$Mean[182:length(xdf$Mean)], function(x) x < 20)
xrle <- rle(xlist)
yrle <- rle(ylist)
xdoy <- which((xrle$lengths >= 6) & xrle$values)
xsum <- cumsum(xrle$lengths)
xdoyref <- xsum[xdoy[1] - 1] + 1
ydoy <- which((yrle$lengths >= 6) & yrle$values)
ysum <- cumsum(yrle$lengths)
ydoyref <- ysum[ydoy[1] - 1] + 1 + 182
return(data.frame(GSL = ydoyref))
}
gslm <- ddply(temprec, c("Year"), gslm.this)For most analysis, the days with trace amount of rainfall is replaced with 0. Before replacing, it may be a good idea to count the number of NAs and Ts. Here's the code to do just that for each year again using ddply function from plyr package.
rainrec <- read.csv(paste0("x:/DHM_data/Rain/", c("Dharan_Bazar.csv")), stringsAsFactors = FALSE)
names(rainrec) <- c("Date", "Rainfall", "DOY")
rainrec$Year <- strftime(as.Date(rainrec$Date, "%Y-%m-%d"),
format='%Y') #add year column
rainrecna <- ddply(rainrec, c("Year"), function(df) c(RainfallNA = sum(is.na(df$Rainfall)),
Tcount = sum(sapply(df$Rainfall,
function(x) {
df$Rainfall[is.na(df$Rainfall)] <- -5
x=="T"}))))
head(rainrecna)Which results in the summary of NA count and T count values as a dataframe.
Year RainfallNA Tcount
1 2000 0 6
2 2001 0 9
3 2002 0 5
4 2003 0 11
5 2004 0 10
6 2005 0 1
The following code converts the rainfall data into RClimDex readable text format. The format for RClimDex is documented here (http://etccdi.pacificclimate.org/RClimDex/RClimDexUserManual.doc).
rainrec.rclimdex <- cbind(rainrec.export$Year, rainrec.export$Month,
as.numeric(strftime(as.Date(rainrec.export$Date), format='%d')),
rainrec.export$Rainfall2)
#add NA values for temp cols
rainrec.rclimdex <- cbind(rainrec.rclimdex, rep(-99.9, length(rainrec.export$Year)),
rep(-99.9, length(rainrec.export$Year)))
#replace NAs with -99.9
rainrec.df <- as.data.frame(rainrec.rclimdex)
rainrec.df$V4[is.na(rainrec.df$V4)] <- -99.9
write.table(rainrec.df, file = paste0(as.character(stationmerged$name[this.station]),
".txt"), sep="\t", col.names=FALSE,
row.names=FALSE)For seasonal and yearly statistics, the data can be converted into a time series. Here is an example using the zoo package.
require(zoo)
#replace Ts with 0
rainrec$Rainfall[rainrec$Rainfall == "T"] <- 0
rainrec$Rainfall <- as.numeric(rainrec$Rainfall)
rain.zoo <- zoo(rainrec$Rainfall, as.Date(rainrec$Date)) #create zoo object
rain.monthly <- apply.monthly(rain.zoo, "sum")
rain.yearly <- apply.yearly(rain.zoo, "sum")
#plot monthly and annual total rainfall
par(mfrow=c(1,2))
plot(rain.monthly, main="Total monthly rainfall in mm")
plot(rain.yearly, main= "Total annual rainfall in mm")
par(mfrow=c(1,1))The following code uses ddply to aggregate seasonal statistics. The year is divided into four seasons with Pre-monsoon for the months of March to May, Monsoon for the months of June to September, Post-monsoon for the months of October - November and Winter for the months of December to February. Although this is the general practice, there are alternative definition of seasons in Nepal (see [4] for example).
rainrec$Month <- as.numeric(strftime(as.Date(rainrec$Date), format='%m'))
rainrec$Season <- factor(rainrec$Season, levels=c(1:4),
labels=c("Pre-monsoon", "Monsoon",
"Post-monsoon", "Winter"))
seasonal <- ddply(rainrec, c("Year", "Season"),
function(df) c(Total = sum(df$Rainfall, na.rm=TRUE)))
#the following takes mean annual rainfall from earlier calculation
# to calculate perecntage of seasonal means
meanseasonal <- ddply(seasonal, c("Season"),
function(df) c(SMean = mean(df$Total, na.rm=TRUE),
SPercent = (mean(df$Total, na.rm=TRUE) /
mean(rain.yearly, na.rm=TRUE)) * 100))The meanseasonal dataframe summarises the mean toal seasonal rainfall calculated from all the records.
Season SMean SPercent
1 Pre-monsoon 304.6667 13.952974
2 Monsoon 1688.6583 77.336341
3 Post-monsoon 149.9250 6.866191
4 Winter 40.2750 1.844495
monthly <- ddply(rainrec, c("Year", "Month"),
function(df) c(Total = sum(as.numeric(df$Rainfall), na.rm=TRUE),
Mean = mean(as.numeric(df$Rainfall), na.rm=TRUE)))This gives the dataframe monthly in the following format.
Year Month Total Mean
1 2000 1 21.0 0.6774194
2 2000 2 17.7 0.6103448
3 2000 3 0.0 0.0000000
4 2000 4 122.5 4.0833333
5 2000 5 349.8 11.2838710
6 2000 6 740.6 24.6866667
Again with the help of ddply the monsoon onset date for each year is computed as shown in this example. Monsoon onset depends on various factors besides rainfall amount [1]. Since we are going to compute monsoon onset date only from rainfall data, the definition of monsoon onset is taken as any rainy day that occurs after June 1 with total rainfall of three consecutive days exceeding 30mm. See [2] & [3] for detailed explanation. A day is counted as a rainy day if there is a rainfall of at least 0.85 mm. There are other similar criteria for calculating monsoon onset which can be done with little modification to the following code. For official dates of monsoon onset and withdrawal for Nepal, see DHM official announcement on monsoon onset and withdrawal.
rainrec.mon <- rainrec[rainrec$Season == "Monsoon",] #isolate only monsoon days
#replace NAs with 0
rainrec.mon$Rainfall[is.na(rainrec.mon$Rainfall)] <- 0
monsoonOnset <- function(xdf) {
reclen <- length(xdf$Rainfall) -2
for (i in 1:reclen) {
if ((xdf$Rainfall[i] >= 0.85) &
(sum(xdf$Rainfall[i],
xdf$Rainfall[i+1],
xdf$Rainfall[i+2], na.rm=TRUE) > 30))
return(xdf$DOY[i])
}
}
monsoon <- ddply(rainrec.mon, c("Year"), monsoonOnset)This returns a dataframe with year and their corrosponding monsoon onset days in DOY. An example with first 6 rows is given below.
Year V1
1 2000 153
2 2001 153
3 2002 153
4 2003 156
5 2004 160
6 2005 160
For this analysis, a dry spell is defined as at least 7 consecutive days of no rainfall after commencement of monsoon in the next 30 days [2, pp 4]. A no rainfall day is defined as a day with less than 0.85 mm of rain. The following function first calculates monsoon onset day and checks the next 30 days for occurance of dry spells using rle function. The function outputs a dataframe which is parsed by ddply into columns.
is.rain <- function(x) x >= 0.85 #only days with >= 0.85mm of rain
drySpell2 <- function(xdf) {
#first get monsoon onset
monindex <- 0
reclen <- length(xdf$Rainfall) - 2
for (i in 1:reclen) {
if ((xdf$Rainfall[i] >= 0.85) &
(sum(xdf$Rainfall[i],
xdf$Rainfall[i+1],
xdf$Rainfall[i+2], na.rm=TRUE) > 30))
{monindex <- i; break}
}
monindex <- monindex + 1 #we only need to check from the day after monsoon onset
reclen2 <- length(xdf$Rainfall)
rainlist.F.count <- 0; rainlist.F.len <- 0
rle.rain <- rle(sapply(xdf$Rainfall[monindex:(monindex+30)],
is.rain)) #check next 30 days
rainlist.F <- rle.rain$lengths[!rle.rain$values]
#total number of dry spells
rainlist.F.count <- sum(sapply(rainlist.F, function(x) x >= 7))
#total dry spell days
rainlist.F.len <- 0
if (rainlist.F.count >= 1)
{ rainlist.F.len <- sum(sapply(rainlist.F[rainlist.F >= 7], sum)) }
return(data.frame(drycount = rainlist.F.count, drylength = rainlist.F.len))
}
drydate <- ddply(rainrec.mon, c("Year"), drySpell2)The output dataframe drydate lists the number of dry days and the total length of dry spell in days.
Year drycount drylength
1 2000 0 0
2 2001 0 0
3 2002 0 0
4 2003 0 0
5 2004 1 7
6 2005 0 0
7 2006 0 0
Alternative code for dry spell days withtout for loop with rollapply function from zoo package is given below.
is.rain <- function(x) x >= 0.85 #only days with >= 0.85mm of rain
drySpell3 <- function(xdf) {
#first get monsoon onset
monindex <- 0
#reclen <- length(xdf$Rainfall) - 2
ydf <- xdf
ydf$Rainfall2[ydf$Rainfall2<0.85] <- NA
## *** corrected 2014-05-26
monindex <- min(which(sapply(rollapply(ydf$Rainfall, 3, sum, partial=3), function(x) x>= 30)))
if (monindex == Inf) {monindex = NA;
return(data.frame(monsoon1 = NA,
monsoon1DOY = NA,
drycount = NA,
drylength = NA))}
rainlist.F.count <- 0
rainlist.F.len <- 0
rainlist <- sapply(xdf$Rainfall[(monindex + 1) :(monindex+30)], is.rain)
rle.rain <- rle(rainlist)
rainlist.F <- rle.rain$lengths[!rle.rain$values]
if (length(rainlist.F) == 0) {return(data.frame(monsoon1 = xdf$Date[monindex],
monsoon1DOY = xdf$DOY[monindex],
drycount = 0,
drylength = 0))}
rainlist.F.count <- sum(sapply(rainlist.F, function(x) x >= 7), na.rm=TRUE)
if (rainlist.F.count >= 1)
{ rainlist.F.len <- sum(sapply(rainlist.F[rainlist.F >= 7], sum)) }
return(data.frame(monsoon1 = xdf$Date[monindex],
monsoon1DOY = xdf$DOY[monindex],
drycount = rainlist.F.count,
drylength = rainlist.F.len))
}
drydate3 <- ddply(rainrec.mon, c("Year"), drySpell3)The output dataframe drydate3 lists the monsoon onset date, monsoon onset DOY, number of dry days and the total length of dry spell in days.
Year monsoon1 monsoon1DOY drycount drylength
1 2000 2000-06-02 155 0 0
2 2001 2001-06-03 154 0 0
3 2002 2002-06-03 155 0 0
4 2003 2003-06-06 157 0 0
5 2004 2004-06-09 161 1 7
6 2005 2005-06-10 163 0 0
7 2006 2006-06-11 162 0 0
As the spatial interpolation of precipitation data is not recommended for Nepali terrain [5], the missing rainfall is replaced by a proxy based on the value of the day in the previous year and the value for the day in the next year. The following code does not take into account of leap years. In the case when the first year's data is missing, only the value from the next year is used.
rainproxy <- function(xlist) {
nax <- which(is.na(xlist))
naxi <- nax + 365 #next year; does not take into a/c of leap year
naxo <- nax - 365 #previous year
nay <- which(naxo <= 0)
naxo[nay] <- naxi[nay] #if missing year is 1st, next year's data is repeated
xlist[nax] <- rowMeans(cbind(xlist[naxi], xlist[naxo]))
return(xlist)
}
rainrec <- transform(rainrec, Rainfall2 = rainproxy(Rainfall))The rainfall-viz.R takes the monsoon rainfall data and plots the daily rainfall and the monsoon onset datefor each year, shades the occurenes of dry spell days if there are any and marks a successful plantation date if there is one. The rainfall amount less than the threshold value (0.85mm per day by default) are not shown. The dry-spell days are marked in light red, and successful plantation date i.e. no dry-spell days in 30 days after the monsoon onset date is marked in light green.
The following code uses Kendall package for Kendall's test for trend the Kendall's tau value and p value are reported in a table.
trend.grain <- ddply(grain, c("Station"), function(xdf) {
x <- MannKendall(xdf$V1)
return(data.frame(grain_tau = x$tau[1],
grain_p_value = x$sl[1]))
})Code for creating table for linear regression statistics. The regression statistics are summarized in a dataframe. The output dataframe is presented below the code.
trend.smk.max.lm.av <- ddply(temprec, c("Station"), function(xdf) {
ann.min <- ddply(xdf, c("Year"), function(ydf) {max(ydf$Min2, na.rm=TRUE)} )
fit <- lm(ann.min$V1 ~ ann.min$Year)
x <- summary(fit)
p <- round(x$coefficients[7], 3)
r <- round(x$r.squared, 3)
#confidence interval
cf <- confint(fit, level=0.95)
return(data.frame(max_lm_r_squared = r,
max_lm_p_value = p,
low_95 = cf[2],
high_95 = cf[4]))
})
print(trend.smk.max.lm.av) Station max_lm_r_squared max_lm_p_value low_95 high_95
1 Dhangadhi_Atariya 0.046 0.073 -0.06810873 0.018720521
2 Nepalgunj_RO 0.011 0.618 -0.01903439 0.035816885
3 Nepalgunj_Airport 0.169 0.057 -0.17849776 0.017713447
4 Dharan_Bazar 0.026 0.746 -0.10166887 0.168701841
5 Tarahara 0.046 0.748 -0.01829748 0.057887225
6 Gaida_Kankai 0.264 0.038 -0.14710102 0.007540581
For calculation of annual precipation anomalies, mean annual rainfall for each station is first computed. The anomaly for each year is given by substracting the value for that year with the mean for that station.
annualmean <- ddply(annual, c("Station"),
function(dfx) c(AnnMean = mean(as.numeric(dfx$Total), na.rm=TRUE)))
annualmerged <- merge(annual, annualmean, by="Station")
annualmerged$Anomaly <- annualmerged$Total - annualmerged$AnnMeanThis outputs the dataframe in the following format
Station Year Total Mean AnnMean Anomaly
1 Anarma 2000 3038.1 8.300820 2704.283 333.81667
2 Anarma 2001 3115.7 8.536164 2704.283 411.41667
3 Anarma 2002 2603.1 7.131781 2704.283 -101.18333
4 Anarma 2003 2789.7 7.643014 2704.283 85.41667
5 Anarma 2004 2636.8 7.204372 2704.283 -67.48333
6 Anarma 2005 1755.3 4.809041 2704.283 -948.98333
For daily discharge data, each text file contains daily records in cubimc m per second for the year in columnar format as shown below. The first nine lines of the file consists of the information on the station and title information. The final three lines of the file also contains the monthly summary of the data.
Station number: 120
Location: Nayal Badi Latitude: 29 40 20
River: Chamelia Longitude: 80 33 30
Year: 2000
Mean daily discharge in m3/s
============================
Day Jan. Feb. Mar. Apr. May Jun. Jul. Aug. Sep. Oct. Nov. Dec. Year
01 25.6 21.6 21.5 29.3 34.7 40.3 99.0 283 230 75.1 39.6 27.5
02 25.9 24.1 20.8 29.8 32.4 38.7 98.2 259 200 72.7 39.4 27.3
03 25.6 22.7 21.0 79.1 34.0 38.9 98.0 187 202 69.0 39.1 27.3
04 25.7 22.7 21.0 22.5 40.9 47.4 86.2 172 181 65.1 39.1 24.9
... and so on ...
Min 21.8 20.8 20.5 25.6 32.4 38.7 85.5 130 76.7 40.1 28.4 21.8 20.5
Mean 23.7 22.0 22.0 30.8 43.9 94.3 175 199 135 53.6 33.4 24.2 71.4
Max 25.9 24.1 29.1 41.6 61.5 398 488 291 230 74.1 39.6 27.5 488
-
Devkota, LP 1984, 'Onset of summer monsoon in Nepal', The Himalayan Review, vol. 15, no. 1983-84, pp. 11-20.
-
Karmacharya, J 2010, Exploring daily rainfall data to investigate evidence of climate change in Kathmandu Valley and its implication in rice farming in the area, Ministry of Agriculture, Kathmandu, Nepal.
-
Upadhyay, S 2010, 'Monsoon variability analysis in Nepal from 1979 to 2008', Department of Environmental Science and Engineering, Bachelor of Science (Honors) thesis, Kathmandu University.
-
Nayava, JL 1980, 'Rainfall in Nepal', Himalayan Review, vol. 12, pp. 1-18.
-
Hormann K 1994, Computer based climatological maps for high mountain areas. International Centre for Integrated Mountain Development, ICIMOD, Kathmandu, Nepal, MEM Ser, 12.