How to make seperate legend in ggplot

I want to plot single scatter plot showing a different color for two station months rainfall. for example A station Jan rainfall value in red, Feb in yellow and B station Jan rainfall value in blue, Feb in green and so on which appears on the legend. Also, I want to include a smooth line for the both stations data that also appear on the legend like the red smooth line for A station and blue for station B. In this link, you can find the both stations CSV data: https://drive.google.com/file/d/0B3fQ9_46L-O0TjJwYmF6UThNSGs/view?usp=sharing https://drive.google.com/file/d/0B3fQ9_46L-O0ZXVYb3lzZDBZaHM/view?usp=sharing

Below is the code I tried but could not succeed.

#reading csv file of ramoili station of rautahat[Scatterplot of two stations][1]
ram = read.csv('preci_ramoili.csv',header=TRUE, stringsAsFactors=FALSE)
#reading CSV file of gaur station of rautahat
gaur= read.csv('preci_Gaur.csv',header=TRUE, stringsAsFactors=FALSE)
#gaur rainfall
rain <- data.frame(index(agg),stack(as.data.frame(coredata(agg))))
rain
head(rain)
tail(rain)
names(rain)[1] <- "Year"
names(rain)[2] <- "Rainfall"
names(rain)[3] <- "Month"
#ramoili rainfall
rain1<-data.frame(index(core),stack(as.data.frame(coredata(core))))
rain1
head(rain1)
names(rain1)[1] <- "Year"
names(rain1)[2] <- "Rainfall"
names(rain1)[3] <- "Month"
head(rain1)
#ramoili premonsoon rainfall
rain1_pre<-data.frame(index(core[,3:5]),stack(as.data.frame(coredata(core[,3:5]))))
head(rain_pre)
tail(rain1_pre)
names(rain1_pre)[1] <- "Year"
names(rain1_pre)[2] <- "Rainfall"
names(rain1_pre)[3] <- "Month"
#ggplot of two stations gaur and ramoili yearly rainfall of rautahat in same plot
p9 <- ggplot(rain, aes(x =Year, y=Rainfall, size=Rainfall)) + geom_point(shape = 21,color = "#000000", fill = "#40b8d0") + 
  geom_smooth(aes(fill="Gaur"), colour="darkblue", size=1)

p10 <- p9 + geom_point(data=rain1, aes(x =Year, y=Rainfall, color=Month )) + 
  geom_smooth(data=rain1, aes(fill="Ramoili"), colour="red", size=1)+ 
  ggtitle(" Yearly rainfall at two stations of Rautahat")+
  scale_fill_manual(name="Stations", values=c("blue", "red"))
print(p10)

Solution

In the absence of complete sample data, and using the sample data you provided, I've illustrated an approach.

I'm uncertain as to why you want the different month colors for each station, and I think that difference would be illustrated better using facets. If not, I would still recommend keeping the month colors consistent and remove the facet grid.

You would need to modify the axes, titles, scales, etc. to fit your liking.

library(dplyr)
library(ggplot2)
library(purrr)
library(tidyr)
library(trend)

gaur <- list.files("~/Desktop", pattern = "Gaur", full.names = TRUE)  %>% 
  read.csv() %>% 
  mutate(station = "Gaur")

ramoili <- list.files("~/Desktop", pattern = "ramoili", full.names = TRUE)  %>% 
  read.csv() %>% 
  mutate(station = "Ramoli")

plot_data <- bind_rows(gaur, ramoili) %>% 
  gather(month, rainfall, -Year, -station) 


ggplot(plot_data, aes(x = Year, y = rainfall)) +
  geom_line(aes(color = month)) +
  geom_point(aes(color = month), show.legend = FALSE) +
  geom_smooth(aes(fill = station), size = 0.1) +
  scale_x_continuous(breaks = scales::pretty_breaks()) +
  scale_fill_manual(name = "Stations", values = c("blue", "red")) +
  facet_grid(month ~ station) +
  theme_minimal()

And for models:

models <- bind_rows(gaur, ramoili) %>% 
      select(-Year) %>% 
      nest(-station) %>% 
      mutate(ts_data = map(data, ~ts(.x, frequency = 1, start = c(1984,1)))) %>% 
      mutate(mk_model = map(ts_data, mk.test),
             sens_slope = map(ts_data, sens.slope))

> models$mk_model[1]
[[1]]
Mann-Kendall Test

two-sided homogeinity test
H0: S = 0 (no trend)
HA: S != 0 (monotonic trend)

Statistics for total series
      S     varS    Z    tau     pvalue
1 -1637 39765.67 -8.2 -0.275 2.2289e-16

> models$sens_slope[1]
[[1]]

Sen's slope and intercept


slope:  0
95 percent confidence intervall for slope
0 0

intercept: 14.9
nr. of observations: 384