Obtaining maximum wind speed during the time series with dplyr

We can use the rWind function called "tidy" to play with the "rWind_series" "list" object to obtain, for example, the maximum speed reported for each cell in the study area during the time series.

 # After tidy our wind data, we can use dplyr utilities to easily  
 # obtain several metrics from our wind data. Notice that wind average  
 # is a special case, and it will be discussed later  
   
 t_ww <- tidy(ww)  
   
 library(dplyr)  
 g_ww <- t_ww %>% group_by(lat, lon)  
   
 # Now, we select only the maximum speed values for each coordinate  
   
 max_ww <- g_ww %>% summarise(speed = max(speed))  
 maxw <- cbind(max_ww$lon, max_ww$lat, max_ww$speed)  
 head(maxw)  
   
 # We can also check the maximum speed value in the whole study area   
   
 max(maxw[,3])  
   
 # Finally, we can transform this data into a raster layer  
   
 rmax <- rasterFromXYZ(maxw)  
 acol <- colorRampPalette(c("white", "blue", "darkblue"))  
 plot(rmax, col=acol(1000), main= "Maximum wind speed reported",  
     xlab="Longitude", ylab="Lattitude")  
   
 library(rworldmap)  
 lines(getMap(resolution = "high"), lwd=2)  

Computing averages of wind speeds and directions

Since wind data is a vectorial magnitude given in two components (U and V), to compute speed and direction averages we should use both vectors. To make it easier, rWind includes wind.mean function. Notice that wind.mean is specially developed to work with wind.dl_2 outputs, so it will require "rWind_series" object as input.

1. We'll use wind2raster to obtain raster layers of wind direction and speed of each day in the time series.
2. Then, for each day, we'll compute the conductance matrix, a matrix with connectivity values between every single cell of the entire study area.
3. Later, with each conductance matrix, we'll compute the cost between our locationsand will store them in a cost_list object.
4. Finally, for each day, if the Cost is not Infinite, we'll get the shortest path between two selected locations, and we'llstore it into a “paths” object.

 # 1) Raster layers from wind time series  
   
 layers <- wind2raster(ww)  
   
 # 2) Conductance from wind direction and speed  
   
 Conductance <- flow.dispersion(layers, type="passive",  
                 output="transitionLayer")  
   
 # 3) and 4) Using a loop to compute cost between  
 # locations and least cost paths between location 11  
 # and Azores islands (as an example).  
   
 cost_list <- array(NA_real_, dim=c(15,15,30))  
 paths <- list(1:30)  
   
 library(gdistance)  
   
 for (i in 1:30){  
  cost_list[,,i] <- costDistance(Conductance[[i]],loc)  
  if (costDistance(Conductance[[i]],loc[11,] , loc[15,]) != Inf){  
   paths[[i]] <- shortestPath(Conductance[[i]], loc[11,], loc[15,],  
                 output="SpatialLines")  
   print(i)  
  }  
 }

 # Transform costs in connectivity  
   
 connectivity <- 1/cost_list  
   
 # Obtain connectivity averages  
   
 conn_avg <- apply(connectivity, c(1, 2), mean, na.rm = TRUE)  
   
 # Here we filter only connectivity from mainland to Azores  
 # As connectivity has arbitrary magnitude and units, we will   
 # multiply connectivity values 1000 times to avoid decimal zeros.  
   
 conn_azores <- conn_avg[,15]  
 conn_azores <- conn_azores *1000  
   
 # Finally, we can plot connectivity values relative to the width  
 # of the connections between mainland and Azores. You can power  
 # the values to increase the differences between them.  
   
 plot(r_mean$wind.speed, main= "Connectivity from mainland to Azores",
     col=acol(1000))  
 lines(getMap(resolution = "low"), lwd=1)  
 for (i in 1:14){  
  lines(rbind(loc[i,], loc[15,]), lwd = conn_azores[i]^3, col="red4")  
 }  
 points(loc, col="red", pch = 19, cex = 1.5)  
 text( loc+2, labels=c(round(conn_azores[1:14], 2),"Azores"),  
    col="white", cex= 1.5, font=2)  

 # We'll use spatstat library to convert the shortest  
 # paths lines into an psp object.  
   
 library(spatstat)  
   
 paths_clean <- paths[!sapply(paths, is.null)]  
 paths_merged <- paths_clean[[1]]  
 for (h in 2:length(paths_clean)) {  
  paths_merged <- rbind(paths_merged,paths_clean[[h]])  
 }  
 paths_psp <- as(paths_merged, "psp")  
   
 # Now, we can obtain a kernel density of the paths  
 # and convert it into a raster layer that can be plotted.  
 # You can play with sigma value to modify the smoothing.  
   
 lines_kernel <- density(paths_psp, sigma=1, dimyx=c(350,410))  
 kernel <- raster(lines_kernel)  
 #kernel[kernel<(maxValue(kernel)*0.1)]<-NA  
   
 ext <- extent(r_mean$wind.speed)  
 kernel <- extend(kernel, ext)  
 kernel[is.na(kernel)] <- minValue(kernel)  
   
 acol <- colorRampPalette(c("grey40","darkblue","red2",
                               "orange","yellow","white"))  
 plot(kernel, col=acol(1000), zlim=c(-0.01,5),  
       main = "Least cost paths density", xlab="Longitude",  
       ylab="Lattitude")  
 lines(getMap(resolution = "high"), lwd=2)