Computing the means of gridded data over land only using Raster Stack
I am interested in calculating precipitation averages globally. However, I only want to isolate land and/or oceanic areas to compute the mean of those separately. What I would like to do is somehow isolate those grid cells whose centers overlap with either land or ocean and then compute the annual mean. I already first created a raster stack, called "RCP1pctCO2Mean", which contains the mean values of interest. There are 138 layers, with each layer representing one year. This raster stack has the following attributes:
class : RasterStack
dimensions : 64, 128, 8192, 138 (nrow, ncol, ncell, nlayers)
resolution : 2.8125, 2.789327 (x, y)
extent : -181.4062, 178.5938, -89.25846, 89.25846 (xmin, xmax, ymin,
coord. ref. : +proj=longlat +datum=WGS84 +ellps=WGS84 +towgs84=0,0,0
names : layer.1, layer.2, layer.3, layer.4, layer.5,
layer.6, layer.7, layer.8, layer.9, layer.10, layer.11,
layer.12, layer.13, layer.14, layer.15, ...
min values : 0.42964514, 0.43375653, 0.51749371, 0.50838983, 0.45366730,
0.53099146, 0.49757186, 0.45697752, 0.41382199, 0.46082401, 0.45516687,
0.51857087, 0.41005131, 0.45956529, 0.47497867, ...
max values : 96.30350, 104.08584, 88.92751, 97.49373, 89.57201,
90.58570, 86.67651, 88.33519, 96.94720, 101.58247, 96.07792,
93.21948, 99.59785, 94.26218, 90.62138, ...
Previously, I tried isolating a specific region by specifying a range of longitudes and latitudes to obtain the means and medians for that region, just like this:
>expansion1<-expand.grid(103:120, 3:15) #This is a range of longitudes and then latitudes
#Packages loaded library(raster)
However, with this approach, too much water can mix with land areas, and if I narrow the latitude/longitude range on land, I might miss too much land to compute the mean meaningfully.
Therefore, with this RasterStack, would it be possible to create a condition that tells R that if the "center point" or centroid of each grid cell (with each grid cell center representing a specific latitude/longitude coordinate) happens to fall on land, then it would be considered as land (i.e. that would be TRUE - if not, then FALSE, or maybe somehow use 0s or 1s)? Even if a grid cell happens to have water mixed with land, but the center point/centroid of the grid is on land, that would be considered as land. I would like to do this for specific countries, too.
I want the individual 138 layers/years to be retained, so that all the Year 1s can be averaged across all relevant grid cells, then all Year 2s, then all Year 3s, then all Year 4s, etc. (to create a time series later). I'm not sure if this is the correct way to do this, but what I did first was take the "RCP1pctCO2Mean" RasterStack and created a SpatialPolygonsDataframe using:
>trans <- raster::rasterToPolygon(RCP1pctCO2Mean) >trans
Could generating an id value for each of those land (or water) polygons, such that the center of those grid cells (i.e. latitude/longitude coordinates) on land are only counted, be a logical next step to do something like this? Is it possible to somehow just isolate an all-land polygon, and then somehow specify which cells are considered land, and then compute averages for each year across the grid cells? If so, there are a lot of cells to assign a weight individually, so I am not sure if there is a way to do this quickly?
Thank you, and I would greatly appreciate any assistance! I look forward to your response!
[[alternative HTML version deleted]]