Climate Analogues for La Hilda

This is a small demostration of the tool Climate Analogues developed by CIAT for the use on RStudio.

Their motto is “finding solutions today, for tomorrow’s climate adaptation”. It allows us:

• 1. To find analogues (similar weathers) for the weather of our location today. This is called Present Analysis
• 2. To find analogues in the present, for the future weather of our location. This is also called Backward Analysis (Future -> Present). It answers the question: Where is my weather of tomorrow?
• 3. To find analogues in the future, for the present weather of our location. This is also called Forward Analysis (Present -> Forward). It answers the question: Where will my weather be tomorrow?

This is a small demostration of the tool, using a beautiful coffee farm - Finca La Hilda - located on the foothills of the Poas Volcano, in Alajuela, Costa Rica.

Present and future data.

For present data, we use the database “WorldClim”, with accumulated data from the 1970-2000 period.

For future projections, we use data from the CMIP5 projections for 2050. We also choose model RCP 8.5, the most pessimistic.

For now, we only we use monthly precipitation and mean temperature.

The underlying code can be seen by clicking the button on the right (same is true for all sections).

Present analysis.

We check for locations with similar weather to ours.

Climate Analogues in the Present for La Hilda Farm

params_pres <- createParameters(x=lon, y= lat,     #Coordinates 
                                vars=c("prec","tmean"),  #Variabiles
                                weights=c(0.5,0.5),      #"weigth per variable
                                ndivisions=c(12,12),     # Number of months
                                growing.season=c(1,12),  # Months to inclue
                                rotation="tmean",        
                                threshold=0.5,           
                                env.data.ref=list(prec_presente, temp_presente),       
                                env.data.targ=list(prec_presente, temp_presente),  
                                outfile="~/.",fname=NA,writefile=FALSE) 

######### Calculare la similarita 

simulation_pres <- calc_similarity(params_pres)

#Set the colors
pal <- colorNumeric(c( "#FFFFCC", "#41B6C4", "#0C2C84"), values(simulation_pres),
  na.color = "transparent")

#Draw the map
leaflet() %>%
            addTiles() %>%
            addRasterImage(simulation_pres, colors= pal, opacity = 0.5)  %>%
  addRasterImage(simulation_pres, colors= pal, opacity = 0.6)  %>%
            setView(lng= lon, lat= lat, zoom= 8)  %>%
            addLegend( pal= pal, values = values(simulation_pres), title = "Similarity Index")%>%
            addMarkers(lng= lon, lat= lat, label= "La Hilda Farm")

Unsurprisingly, we see that the weather from La Hilda shares the most similarities with that of locations at similar altitudes within the central mountain range of Costa Rica.

You can explore the similarities in other countries, but don’t go to far because the map has a finite size.

Backward analysis: Where is my weather of tomorrow?

We first plot a climatogram, to visualize the changes in weather.

#Extract values for rain present
rasvalue = extract(prec_presente, coordinates[,c(2,3)])
rasvalue= cbind(coordinates[1],rasvalue)
rasvalue = melt(rasvalue, id = c("farm"))
rasvalue = cbind(rasvalue, month= rep(1:12, each = 2))
rain_pres = subset(rasvalue, farm== "La Hilda")


#Extract values for rain future
rasvalue = extract(c5_prec, coordinates[,c(2,3)])
rasvalue= cbind(coordinates[1],rasvalue)
rasvalue = melt(rasvalue, id = c("farm"))
rasvalue = cbind(rasvalue, month= rep(1:12, each = 2))
rain_fut = subset(rasvalue, farm== "La Hilda")

#Extract values for tmean present
rasvalue = extract(temp_presente, coordinates[,c(2,3)])
rasvalue= cbind(coordinates[1],rasvalue)
rasvalue = melt(rasvalue, id = c("farm"))
rasvalue = cbind(rasvalue, month= rep(1:12, each = 2))
temp_pres = subset(rasvalue, farm== "La Hilda")

#Extract values for tmean future
rasvalue = extract(c5_temp, coordinates[,c(2,3)])
rasvalue= cbind(coordinates[1],rasvalue)
rasvalue = melt(rasvalue, id = c("farm"))
rasvalue = cbind(rasvalue, month= rep(1:12, each = 2))
temp_fut = subset(rasvalue, farm== "La Hilda")



##Bind all

alldata= cbind(rain_pres, rain_fut= rain_fut$value, temp_pres = temp_pres$value/10, temp_fut = temp_fut$value/10)
names= c("name",'x','rain_pres', 'month','rain_fut', 'temp_pres', 'temp_fut')
colnames(alldata)= names
monthsinenglish= c("Jan","Feb","Mar","Apr","May","Jun","Jul", "Aug","Sep","Oct", "Nov", "Dec")
months <- as.data.frame(cbind(seq(1:12),  monthsinenglish))
colnames(months) = c('month', 'monthname')
alldata <- merge(alldata,months)
pres = alldata[,c(2,4,6,8,1)]
pres$Time <- "Present (1970-2000)"
fut=alldata[,c(2,5,7,8,1)]
fut$Time <- "Future (2050 w/ RCP 8.5)"
names= c("Farm", "prec", "temp", "month", "monthnumber","Time")
names(pres)= names
names(fut) = names
mydata = rbind(pres, fut)

mydata$Time =  factor(mydata$Time , levels= c("Present (1970-2000)","Future (2050 w/ RCP 8.5)"))

#now plot
ggplot(data = mydata, mapping = aes(x = monthnumber, y = prec, fill = Time)) +
  ggtitle(label=  "Climatogram for La Hilda Farm")+
  scale_y_continuous("Rainfall [mm]", 
                     sec.axis = sec_axis(~ . /5 , 
                                         name = "Mean Temperature [°C]"))+
  geom_bar(stat= "identity", position=position_dodge())+
  geom_line(mapping = aes(y = temp*5, x= monthnumber), col ="black", size=1.2)+
  geom_line(mapping = aes(y = temp*5, x= monthnumber, col =Time), size=0.8)+
  scale_color_manual(values =c("darkolivegreen3","firebrick3"))+
  scale_fill_manual(values =c("darkolivegreen3","firebrick3"))+
  theme(text= element_text( size = 14),
        plot.title = element_text(hjust = 0.5),
        axis.text =  element_text( size = 14),
        legend.position="bottom")+
  scale_x_continuous(breaks=seq(1:12), labels= monthsinenglish, name= "Month")

We then make the map

Present Analogues for La Hilda’s weather in the Future.

params_pres_fut<- createParameters(x=lon, y= lat,  #Coordinates  
                           vars=c("prec","tmean"),  #Variabiles
                           weights=c(0.5,0.5),      #"WEIGHT per variabe
                           ndivisions=c(12,12),  # Number of months
                           growing.season=c(1,12),  # Months to include
                           rotation="tmean",        
                           threshold=0.5,         #minimum similarity to plot
                       env.data.targ=list(prec_presente, temp_presente),      
                           env.data.ref=list(c5_prec, c5_temp), 
                           outfile="~/.",fname=NA,writefile=FALSE)  

simulation_back <- calc_similarity(params_pres_fut)


pal <- colorNumeric(c( "#FFFFCC", "#41B6C4", "#0C2C84"), values(simulation_back),
  na.color = "transparent")


leaflet() %>%
            addTiles() %>%
            addRasterImage(simulation_back, colors= pal, opacity = 0.5)  %>%
            setView(lng= lon, lat= lat, zoom= 8)  %>%
            addLegend( pal= pal, values = values(simulation_back), title = "Similarity Index")%>%
            addMarkers(lng= lon, lat= lat, label= "La Hilda Farm")

Surprisingly, for 2050 (and according to the worst of predictions), the weather from Finca La Hilda will be more similar to those that can be found, for example, in the areas of Naranjo and Palmares. In comparison to the last map, we see that all the most similar areas have “stepped down” in the altitude range.

That is all. Hope you enjoyed.

# Forward Analysis: Where will my weather be tomorrow? 
#NOT DONE YET!

#This is how I did it, but I had to tag it away from the Markdown code because it was getting to heavy, i just decided to read the previously done raster

params_forw <- createParameters(x=lon, y= lat,  #Coordinates   
                              vars=c("prec","tmean"),  #Variabiles
                             weights=c(0.5,0.5),       #weight per variable
                              ndivisions=c(12,12),     # Number of months
                              growing.season=c(1,12),  # Months to include
                              rotation="tmean",        
                             threshold=0.5,           # THreshold
                              env.data.targ=list(c5_prec, c5_temp),      
                              env.data.ref=list(prec_presente, temp_presente),
                              outfile="~/.",fname=NA,writefile=FALSE)   

simulation_forw <- calc_similarity(params_forw)

writeRaster(simulation_forw, "simulation_forw_LaHilda", overwrite= T, format= "GTiff")

#Reading the raster from my disk
simulation_forw <- Raster('simulation_forw_LaHilda.tif')

#Draw the map
pal <- colorNumeric(c( "#FFFFCC", "#41B6C4", "#0C2C84"), values(simulation_forw),
  na.color = "transparent")

leaflet() %>%
            addTiles() %>%
            addRasterImage(simulation_forw, colors= pal, opacity = 0.6)  %>%
            setView(lng= lon, lat= lat, zoom= 4)  %>%
            addLegend( pal= pal, values = values(simulation_pres), title = "Similarity Index")%>%
            addMarkers(lng= lon, lat= lat, label= "La Hilda Farm")