R help - Nov 2023 - Problem in R code

Good afternoon,
I have been working on my thesis project on the topic "Urban Heat Island
Pattern in India". To achieve the results I am applying a* two-dimensional
Gaussian fit* on an LST raster of 1 km spatial resolution but I am facing
two errors in the following code.

library(raster)
LST <- raster("D:/Celsius_Day/MOD_01.tif")
gaussian2d <- function(x, y, mu_x, mu_y, sigma_x, sigma_y, amp)
{
  exp(-((x - mu_x)^2/(2*sigma_x^2) + (y - mu_y)^2/(2*sigma_y^2)))*amp
}

#define a function for the sum of squared errors between the data and the
Gaussian
sse <- function(p)
{ mu_x <- p
mu_y <- p
sigma_x <- p
sigma_y <- p
amp <- p[5]
fitted <- gaussian2d(x, y, mu_x, mu_y, sigma_x, sigma_y, amp)
sum((z - fitted)^2)
}

#loop over 8 cities
cities <-
c("Delhi","Jaipur","Kolkata","Mumbai","Pune","Hyderabad","Bangalore","Chennai")
lon <-
c(77.219934,75.793261,88.365394,72.900361,73.875199,78.47476,77.602114,80.192181)
lat <-
c(28.589256,26.892024,22.619754,19.110629,18.50269,17.422973,12.974087,13.044415)
results <- data.frame() #A data frame to store the results
for(i in 1:8)
{
  LST_city <- extract(LST, c(lon[i],lat[i]), fun = mean, buffer = 10000,
na.rm = TRUE)   #error
}

# Fit a 2D Gaussian surface to the LST data
x <- coordinates(LST)[,1]
y <- coordinates(LST)[,2]
z <- values(LST)
mu_x0 <- mean(x)
mu_y0 <- mean(y)
sigma_x0 <- sd(x)/2
sigma_y0 <- sd(y)/2
amp0 <- max(z)
opt <- optim(c(mu_x0, mu_y0, sigma_x0, sigma_y0, amp0), sse)     #error 2

#Calculate the footprint of SUHI effect (FP) by the Gaussian surface
FP <- which(gaussian2d(x, y, opt$par, opt$par, opt$par, opt$par,
opt$par[5]) >= threshold)

#store the results for each city in the data frame
results <- rbind(results, data.frame(city=cities[i], FP=mean(FP)))

#print the results
results


The two errors are in the row which are defining the variables
"LST_city"
and "opt".
The first error is:
Error in .cellValues(x, y, ...) : unused arguments (fun
new("standardGeneric", .Data = function (x, ...)
standardGeneric("mean"),
generic = "mean", package = "base", group = list(),
valueClass character(0), signature = "x", default =
new("derivedDefaultMethod", .Data
= function (x, ...) UseMethod("mean"), target =
new("signature", .Data "ANY", names = "x", package
= "methods"), defined = new("signature", .Data
= "ANY", names = "x", package = "methods"),
generic = "mean"), skeleton (new("derivedDefaultMethod",
.Data = function (x, ...) UseMethod("mean"),
target = new("signature", .Data = "ANY", names =
"x", package = "methods"),
defined = new("signature", .Data = "ANY", names =
"x", package "methods"), generic = "mean"))(x,
...)), buffer = 20000, na.rm = TRUE)


The second error is:

Error in optim(c(mu_x0, mu_y0, sigma_x0, sigma_y0, amp0), sse) :
  non-finite value supplied by optim



What could be the possible reason behind these errors and most
importantly how can I get rid of these errors?



Thank you


Regards

DD

	[[alternative HTML version deleted]]

? Wed, 8 Nov 2023 16:03:15 +0530
Crown Flame <crown11flame at gmail.com> ?????:
> for(i in 1:8)
> {
>   LST_city <- extract(LST, c(lon[i],lat[i]), fun = mean, buffer >
10000, na.rm = TRUE)   #error
> }
Three things you might need to change:

1. You are trying to assign the output of extract() to the same
variable LST_city in all 8 iterations of the loop. You will probably
need to make it a list and assign to the list elements, e.g.
LST_city[[i]] <- ...

It will also help to learn about lapply() and other functions that
encapsulate loops, although that is more of a matter of taste.

2. You are giving a single vector c(lon[i], lat[i]) to extract() as the
y=... argument. According to help(extract), this is interpreted as
_numbers_ of cells inside x, not coordinates. You should probably
construct a spPolygons() object and use that as the `y` argument to
extract().

3. The description of the 'raster' package says that it's been
superseded by the 'terra' package. You don't have to rewrite all
your
code now, but it may be beneficial to check whether 'terra' can
accomplish what you want.
> Error in optim(c(mu_x0, mu_y0, sigma_x0, sigma_y0, amp0), sse) :
>   non-finite value supplied by optim
This must be self-explanatory: your `sse` function returns something
that is not a real number. Try options(error = recover) to see which
arguments it is given when it fails in such a manner. See the free book
"The R Inferno" for more R debugging tips
<https://www.burns-stat.com/documents/books/the-r-inferno/>. Consider
using the 'optimx' package which contains optimisation algorithms that
might work better for you.

-- 
Best regards,
Ivan