## Owt appplication

### Apply SAM to Rrs

require(data.table)
require(openxlsx)
require(hsdar)
require(dplyr)
require(data.table)

hyp = read.xlsx('Data/rrs.xlsx', detectDates = T)

references = fread('Data/OWTs/OWT_allwaters_mean_standardised.csv')

references = references[is.na(references$`Pahlevan et. al. 2021`)==F, -2]

# Remove colunas não utilizadas

# Transpõe a tabela de referências (exceto primeira coluna)
ref_names <- references[[1]]
references.t <- as.data.frame(t(references[,-1]))
colnames(references.t) <- ref_names
rownames(references.t) <- 400:800

# Prepara a tabela de saída
point <- data.frame(id = hyp$BR_ID)

# Renomeia colunas do dataframe de reflectância
names(hyp)[-c(1)] <- paste0("X", 400:900)

# Função para calcular o SAM
calc_sam <- function(target, refs) {
  apply(refs, 2, function(ref) {
    numerador <- sum(target * ref)
    denominador <- sqrt(sum(target^2)) * sqrt(sum(ref^2))
    theta <- acos(numerador / denominador)
    return(theta)
  })
}



# Loop sobre as amostras
for (i in 1:nrow(hyp)) {
  pt <- hyp[i, ] %>%
    dplyr::select(starts_with("X")) %>%
    dplyr::select(paste0("X", 400:750)) %>%
    t() %>% data.frame() %>% na.omit()
  
  pt <- as.numeric(as.character(pt[,1]))
  pt <- pt / sum(pt)
  
  if (length(pt) == 351) {
    distance <- calc_sam(pt[1:350], references.t[1:350,])
    point[i, 2:(1 + length(distance))] <- distance
  } else {
    point[i, 2:(1 + ncol(references.t))] <- NA
  }
  
  print(i)
}

# Nomeia as colunas de distância
names(point)[2:(1 + length(ref_names))] <- ref_names

# Classificação com base no menor ângulo
point$Class <- apply(point[,-1], 1, function(row) {
  if (all(is.na(row))) return(NA)
  return(which.min(row))
})

write.csv(point, 'Outputs/SAM_OWT_rrs.csv', row.names = FALSE)