# ShowGeotiff.r
# Have a look at a downloaded satellite image and check if stats are correctly calculated
# I downloaded and calculated the stats for the polygon defined in C:\git\FarmMapsApiClient_WURtest\FarmmapsDataDownload\DataDownloadInput.json
# in which I set "SatelliteBand": "wdvi" and in which in the console I requested the image for date '2022-03-08'

library(raster)
library(sf)
library(rgdal)
setwd("C:/git/FarmMapsApiClient_WURtest/FarmmapsDataDownload/bin/Debug/netcoreapp3.1/Downloads")

# FarmmapsDataDownload
fileGeotiff <- "sentinelhub_test_BvdTFieldlabG92_20220308.tif"
lenfilename <- nchar(fileGeotiff)
year <- substr(fileGeotiff,lenfilename-11,lenfilename-8)
imgdate <- substr(fileGeotiff,lenfilename-11,lenfilename-4)

stk.sentinelhub <- stack(x=fileGeotiff)
# plot(stk.sentinelhub) shows 6 plots (6 bands)
# 1. ndvi
# 2. wdvi Note wdvi-red
# 3. ci-red
# 4. natural: red
# 5. natural: green
# 6. natural: blue
names(stk.sentinelhub) <- c("ndvi","wdvired","ci-red","red","green","blue")
plot(stk.sentinelhub)
crs(stk.sentinelhub)
# CRS arguments: +proj=longlat +datum=WGS84 +no_defs
stk.sentinelhub.rd <- projectRaster(stk.sentinelhub, crs = CRS('+init=EPSG:28992'))
crs(stk.sentinelhub)

r.sentinelhub.rd.wdvi <- subset(stk.sentinelhub.rd,2)
dev.off()
plot(r.sentinelhub.rd.wdvi,main=paste("wdvi",imgdate),xlab="RDX",ylab="RDY")
cellStats(r.sentinelhub.rd.wdvi,'mean')	# 0.2252725

# Convert the .rd.wdvi raster to WGS84
r.sentinelhub.wgs84.wdvi <- projectRaster(r.sentinelhub.rd.wdvi, crs = CRS('+init=EPSG:4326'))

# Draw a polygon on top of the raster
# Polygon pol from C:\git\FarmMapsApiClient_WURtest\FarmmapsDataDownload\DataDownloadInput.json
pol <- data.frame(id = 1, wkt = gsub("\n","",'POLYGON((
      5.563472073408009 52.547554398144172,
      5.567425915520115 52.547725375100377,
      5.567917474269188 52.540608459298582,
      5.563878143678981 52.54048022658143,
      5.563472073408009 52.547554398144172
      ))'))


pol.wgs84 <- st_as_sf(pol, wkt = 'wkt', crs = CRS('+init=EPSG:4326'))
pol.rd <- st_transform(pol.wgs84, "+init=epsg:28992")

#Calculate approximate middle of polygon
res <- as.data.frame(do.call("rbind", lapply(st_geometry(pol.wgs84), st_bbox)))
res$latmid <- (res$ymax+res$ymin)/2.0
res$lonmid <- (res$xmax+res$xmin)/2.0
res
      # xmin     ymin     xmax     ymax  latmid   lonmid
# 1 5.563472 52.54048 5.567917 52.54773 52.5441 5.565695

# Have a look at both polygons
# wg84
dev.off()
plot(r.sentinelhub.wgs84.wdvi,main=paste("wdvi",imgdate),xlab="LON",ylab="LAT")
plot(pol.wgs84,add=TRUE, col="transparent",border="red")
# RD
dev.off()
plot(r.sentinelhub.rd.wdvi,main=paste("wdvi",imgdate),xlab="RDX",ylab="RDY")
plot(pol.rd,add=TRUE, col="transparent",border="red")

# Clip the polygon from the full rectangle figure
r.sentinelhub.rd.wdvi.pol <- mask(r.sentinelhub.rd.wdvi,pol.rd)
r.sentinelhub.wgs84.wdvi.pol <- mask(r.sentinelhub.wgs84.wdvi,pol.wgs84)
dev.off()
plot(r.sentinelhub.wgs84.wdvi.pol,main=paste("wdvi",imgdate),xlab="LON",ylab="LAT")
plot(pol.wgs84,add=TRUE, col="transparent",border="red")
#That's what we want! 

# Now compare the stats
cellStats(r.sentinelhub.wgs84.wdvi,'mean')	# [1] 0.2250987 # Stats from rectangle, WGS84
cellStats(r.sentinelhub.rd.wdvi,'mean')	# [1] 0.2252725 # Stats from rectangle, RD. Almost but not exactly same as above
cellStats(r.sentinelhub.wgs84.wdvi.pol,'mean')	# [1] 0.2275067 # Stats from raster clipped by polygon, WGS84
cellStats(r.sentinelhub.rd.wdvi.pol,'mean')	# [1] 0.2275073 # Stats from raster clipped by polygon, RD. Almost but not exactly same as above
# file satelliteStats_test_BvdTFieldlabG92_20220308.csv
# "wdvi" "mean": 0.22744397204465
# Mean in csv corresponds with cellStats calculated from clipped tif!
# So while the tif returned is a non-clipped image, the downloaded statistics are from the clipped image
# Exactly as we wanted.
cellStats(r.sentinelhub.wgs84.wdvi.pol,'median')	# Error in .local(x, stat, ...) : invalid 'stat'. Should be sum, min, max, sd, mean, or 'countNA'
r.sentinelhub.wgs84.wdvi.vals <- values(r.sentinelhub.wgs84.wdvi)
median(r.sentinelhub.wgs84.wdvi.vals)	# [1] NA
median(r.sentinelhub.wgs84.wdvi.vals,na.rm=TRUE)	# [1] 0.2318
r.sentinelhub.wgs84.wdvi.pol.vals <- values(r.sentinelhub.wgs84.wdvi.pol)
median(r.sentinelhub.wgs84.wdvi.pol.vals)	# [1] NA
median(r.sentinelhub.wgs84.wdvi.pol.vals,na.rm=TRUE)	# [1] 0.2338
# file satelliteStats_test_BvdTFieldlabG92_20220308.csv
# "wdvi" "mean": 0.233799993991851
# Median is same as for median(r.sentinelhub.wgs84.wdvi.pol.vals,na.rm=TRUE)
#   in csv corresponds with cellStats calculated from clipped tif!
# So while the tif returned is a non-clipped image, the downloaded statistics are from the clipped image
# Exactly as we wanted.
cellStats(r.sentinelhub.wgs84.wdvi,'countNA')	# [1] 27896
ncell(r.sentinelhub.wgs84.wdvi)	# [1] 272718
cellStats(r.sentinelhub.wgs84.wdvi,'countNA') / ncell(r.sentinelhub.wgs84.wdvi)	# [1] 0.1022888	# 10% no data? doesn't show in the plot?
cellStats(r.sentinelhub.wgs84.wdvi.pol,'countNA')	# [1] 57625
summary(r.sentinelhub.wgs84.wdvi.pol.vals)	# shows the same: NA's: 57625
ncell(r.sentinelhub.wgs84.wdvi.pol)	# [1] 272718
populationCount = ncell(r.sentinelhub.wgs84.wdvi.pol) - cellStats(r.sentinelhub.wgs84.wdvi.pol,'countNA')
populationCount # [1] 215093
# file satelliteStats_test_BvdTFieldlabG92_20220308.csv
# "wdvi" "populationCount": 214688
# similar but not same