diff --git a/.gitignore b/.gitignore
index 89a262feca684eb5c483695ed9261014aba892af..b2d7bbf988d665ea906815f3c6e84c94c628f08a 100644
--- a/.gitignore
+++ b/.gitignore
@@ -17,4 +17,9 @@ Rplots.pdf
 .nfs*
 *.RData
 !data/*.RData
+<<<<<<< HEAD
+.Rproj.user/
+
+=======
 .Rproj.user
+>>>>>>> d82bd2503734b41ca8f7bdd66deb934c3b202f03
diff --git a/R/CST_Calibration.R b/R/CST_Calibration.R
index 3ba22917679947eb71b1b6c2211b9e60bda804be..c4942eaade57a66320625ea5ad96f2aef4f0f1e7 100644
--- a/R/CST_Calibration.R
+++ b/R/CST_Calibration.R
@@ -149,7 +149,7 @@ CST_Calibration <- function(exp, obs, exp_cor = NULL, cal.method = "mse_min",
                             eval.method = "leave-one-out", multi.model = FALSE, 
                             na.fill = TRUE, na.rm = TRUE, apply_to = NULL, 
                             alpha = NULL, memb_dim = 'member', sdate_dim = 'sdate', 
-                            dat_dim = NULL, ncores = NULL) {
+                            dat_dim = NULL, ncores = NULL, k = 1, tail.out = TRUE) {
   # Check 's2dv_cube'
   if (!inherits(exp, "s2dv_cube") || !inherits(obs, "s2dv_cube")) {
         stop("Parameter 'exp' and 'obs' must be of the class 's2dv_cube'.")
@@ -165,7 +165,7 @@ CST_Calibration <- function(exp, obs, exp_cor = NULL, cal.method = "mse_min",
                              multi.model =  multi.model, na.fill = na.fill, 
                              na.rm = na.rm, apply_to = apply_to, alpha = alpha, 
                              memb_dim = memb_dim, sdate_dim = sdate_dim, 
-                             dat_dim = dat_dim, ncores = ncores)
+                             dat_dim = dat_dim, ncores = ncores, k = k, tail.out = tail.out)
 
   if (is.null(exp_cor)) {
     exp$data <- Calibration
@@ -305,7 +305,7 @@ Calibration <- function(exp, obs, exp_cor = NULL,
                         multi.model = FALSE, na.fill = TRUE, 
                         na.rm = TRUE, apply_to = NULL, alpha = NULL,
                         memb_dim = 'member', sdate_dim = 'sdate', dat_dim = NULL, 
-                        ncores = NULL) {
+                        ncores = NULL, k = 1, tail.out = TRUE) {
 
   # Check inputs
   ## exp, obs
@@ -478,10 +478,10 @@ Calibration <- function(exp, obs, exp_cor = NULL,
     }
   }
   ## eval.method
-  if (!any(eval.method %in% c('in-sample', 'leave-one-out', 'hindcast-vs-forecast'))) {
+  if (!any(eval.method %in% c('in-sample', 'leave-one-out', 'hindcast-vs-forecast', 'k-fold', 'retrospective'))) {  
     stop(paste0("Parameter 'eval.method' must be a character string indicating ", 
-                "the sampling method used ('in-sample', 'leave-one-out' or ", 
-                "'hindcast-vs-forecast')."))
+                "the sampling method used ('in-sample', 'leave-one-out', 'hindcast-vs-forecast', 'k-fold' or ", 
+                "'retrospective')."))
   }
   ## multi.model
   if (!inherits(multi.model, "logical")) {
@@ -513,7 +513,7 @@ Calibration <- function(exp, obs, exp_cor = NULL,
                         cal.method = cal.method, eval.method = eval.method, multi.model = multi.model,
                         na.fill = na.fill, na.rm = na.rm, apply_to = apply_to, alpha = alpha,
                         target_dims = list(exp = target_dims_exp, obs = target_dims_obs),
-                        ncores = ncores, fun = .cal)$output1
+                        ncores = ncores, k = k, tail.out = tail.out, fun = .cal)$output1
   } else {
     calibrated <- Apply(data = list(exp = exp, obs = obs, exp_cor = exp_cor),
                         dat_dim = dat_dim, cal.method = cal.method, eval.method = eval.method,
@@ -521,7 +521,7 @@ Calibration <- function(exp, obs, exp_cor = NULL,
                         apply_to = apply_to, alpha = alpha,
                         target_dims = list(exp = target_dims_exp, obs = target_dims_obs, 
                                            exp_cor = target_dims_cor),
-                        ncores = ncores, fun = .cal)$output1
+                        ncores = ncores, k = k, tail.out = tail.out, fun = .cal)$output1
   }
 
   if (!is.null(dat_dim)) {
@@ -567,10 +567,62 @@ Calibration <- function(exp, obs, exp_cor = NULL,
   }
 }
 
-.make.eval.train.dexes <- function(eval.method, amt.points, amt.points_cor) { 
+.make.eval.train.dexes <- function(eval.method, amt.points, amt.points_cor,
+                                   k = 1, tail.out = TRUE) {
+# eval.method: 'leave-one-out', 'k-fold', 'in-sample', "hindcast-vs-forecast", "retrospective"
+# amt.points: length of the sample
+# amt.points_cor: only needed for hindcast-vs-forecast method
+# k: the number of samples to leave-out
+# tail_out: boolean for 'k-fold method; TRUE to remove both extremes of the sample when k is 
+# in the extreme keeping the same sample size for all k-folds (e.g. amt.points=50, k=3, 
+# eval.dexes=1, train.dexes={3,49}). FALSE to remove only the corresponding tail (e.g..
+# amt.points=50, k=3, eval.dexes=1, train.dexes={3,50})
+
+  if (k >= amt.points && !is.null(k) | k < 1) { 
+    stop("k needs to be a positive integer less than the amt.points")
+  }
   if (eval.method == "leave-one-out") {
     dexes.lst <- lapply(seq(1, amt.points), function(x) return(list(eval.dexes = x,
                         train.dexes = seq(1, amt.points)[-x])))
+  } else if (eval.method == "k-fold") {
+  # k is a odd number
+    if (tail.out == TRUE){
+          dexes.lst <- lapply(seq(1, amt.points), function(x, kfold = k) {
+      if (x >= ((kfold-1)/2) + 1 && x + ((kfold-1)/2) <= amt.points) {
+        ind <- (x-((kfold-1)/2)):(x+((kfold-1)/2))
+      } else if (x < ((kfold-1)/2) + 1) {
+        ind <- c((amt.points - ((kfold-1)/2-x)):amt.points, 1:(x+(kfold-1)/2))
+      } else if ((x+((kfold-1)/2)) > amt.points) {
+        ind <- c((x-(kfold-1)/2):amt.points, 1:(((kfold-1)/2)-amt.points+x)) 
+      } else {
+        stop("Review make.eval.train.dexes function")
+      }
+      return(list(eval.dexes = x,  train.dexes = seq(1, amt.points)[-ind]))
+      })
+    } else {
+      if (k != 1){
+        k <- (k-1)/2
+      }
+      
+      dexes.lst <- lapply(seq(1, amt.points), function(x) {
+        start_idx <- max(1, x - k)
+        end_idx <- min(amt.points, x + k)
+        eval_range <- start_idx:end_idx
+        train_idx <- setdiff(seq(1, amt.points), eval_range)
+        return(list(eval.dexes = x, train.dexes = train_idx))
+      })
+    }
+
+  } else if (eval.method == "retrospective") { 
+  # k can be any integer indicating the when to start
+    dexes.lst <- base::Filter(length, lapply(seq(1, amt.points),
+                        function(x, mindata = k) {
+                          if (x > k) { 
+                            eval.dexes <- x
+                            train.dexes <- 1:(x-1)
+                            return(list(eval.dexes = x,
+                                        train.dexes = 1:(x-1)))
+                          }})) 
   } else if (eval.method == "in-sample") {
     dexes.lst <- list(list(eval.dexes = seq(1, amt.points), 
                            train.dexes = seq(1, amt.points)))
@@ -580,12 +632,13 @@ Calibration <- function(exp, obs, exp_cor = NULL,
   } else {
     stop(paste0("unknown sampling method: ", eval.method))
   }
+
   return(dexes.lst)
 }
 
 .cal <- function(exp, obs, exp_cor = NULL, dat_dim = NULL, cal.method = "mse_min", 
                  eval.method = "leave-one-out", multi.model = FALSE, na.fill = TRUE, 
-                 na.rm = TRUE, apply_to = NULL, alpha = NULL) {
+                 na.rm = TRUE, apply_to = NULL, alpha = NULL, k = 1, tail.out = TRUE) {
 
   # exp: [memb, sdate, (dat)]
   # obs: [sdate (dat)]
@@ -642,7 +695,8 @@ Calibration <- function(exp, obs, exp_cor = NULL,
 
         eval.train.dexeses <- .make.eval.train.dexes(eval.method = eval.method, 
                                                      amt.points = sdate, 
-                                                     amt.points_cor = sdate_cor)
+                                                     amt.points_cor = sdate_cor, 
+                                                     k = k, tail.out = tail.out)
         amt.resamples <- length(eval.train.dexeses)
         for (i.sample in seq(1, amt.resamples)) {
           # defining training (tr) and evaluation (ev) subsets
diff --git a/R/CST_CategoricalEnsCombination.R b/R/CST_CategoricalEnsCombination.R
index 3c92d587efb6aa4be76bf7137bf68aa557bc4e40..f973cde79ff9a590438be36cee50d5c16f55c234 100644
--- a/R/CST_CategoricalEnsCombination.R
+++ b/R/CST_CategoricalEnsCombination.R
@@ -97,7 +97,7 @@
 #'@export
 CST_CategoricalEnsCombination <- function(exp, obs, cat.method = "pool", 
                                           eval.method = "leave-one-out", 
-                                          amt.cat = 3, 
+                                          amt.cat = 3, k = 1, tail.out = TRUE,
                                           ...) {
   # Check 's2dv_cube'
   if (!inherits(exp, "s2dv_cube") || !inherits(exp, "s2dv_cube")) {
@@ -113,7 +113,8 @@ CST_CategoricalEnsCombination <- function(exp, obs, cat.method = "pool",
   exp$data <- CategoricalEnsCombination(fc = exp$data, obs = obs$data, 
                                         cat.method = cat.method,
                                         eval.method = eval.method, 
-                                        amt.cat = amt.cat, ...)
+                                        amt.cat = amt.cat, k = k, 
+                                        tail.out = tail.out, ...)
 
   names.dim.tmp[which(names.dim.tmp == "member")] <- "category"
   names(dim(exp$data)) <- names.dim.tmp
@@ -175,7 +176,7 @@ CST_CategoricalEnsCombination <- function(exp, obs, cat.method = "pool",
 #'@importFrom s2dv InsertDim
 #'@import abind
 #'@export
-CategoricalEnsCombination <- function (fc, obs, cat.method, eval.method, amt.cat, ...) {
+CategoricalEnsCombination <- function (fc, obs, cat.method, eval.method, amt.cat, k, tail.out, ...) {
   
   if (!all(c("member", "sdate") %in% names(dim(fc)))) {
     stop("Parameter 'exp' must have the dimensions 'member' and 'sdate'.")
@@ -205,6 +206,8 @@ CategoricalEnsCombination <- function (fc, obs, cat.method, eval.method, amt.cat
     cat.method = cat.method,
     eval.method = eval.method,
     amt.cat = amt.cat,
+    k = k,
+    tail.out = tail.out,
     ...)
   return(cat_fc_out)
 }
@@ -243,7 +246,7 @@ comb.dims <- function(arr.in, dims.to.combine){
 	return(arr.out)
 }
 
-.apply.obs.fc <- function(obs, fc, target.dims, FUN, return.feat, cat.method, eval.method, amt.cat, ...){
+.apply.obs.fc <- function(obs, fc, target.dims, FUN, return.feat, cat.method, eval.method, amt.cat, k, tail.out,  ...){
   dimnames.tmp <- dimnames(fc)
   fc.dims.tmp <- dim(fc)
   dims.out.tmp <- return.feat$dim
@@ -260,6 +263,8 @@ comb.dims <- function(arr.in, dims.to.combine){
     cat.method = cat.method,
     eval.method = eval.method,
     amt.cat = amt.cat,
+    k = k,
+    tail.out = tail.out,
     ...)
   dims.tmp <- dim(arr.out)
   names.dims.tmp <- names(dim(arr.out))
@@ -280,7 +285,7 @@ comb.dims <- function(arr.in, dims.to.combine){
 }
 
 
-.cat_fc <- function(obs.fc, amt.cat, cat.method, eval.method) {
+.cat_fc <- function(obs.fc, amt.cat, cat.method, eval.method, k, tail.out) {
 	
   dims.tmp=dim(obs.fc)
   amt.mbr <- dims.tmp["member"][]-1
@@ -299,7 +304,7 @@ comb.dims <- function(arr.in, dims.to.combine){
   amt.coeff <- amt.mdl + 1
   var.cat.fc <- array(NA, c(amt.cat, amt.sdate))
   
-  eval.train.dexeses <- .make.eval.train.dexes(eval.method = eval.method, amt.points = amt.sdate)
+  eval.train.dexeses <- .make.eval.train.dexes(eval.method = eval.method, amt.points = amt.sdate, k = k, tail.out = tail.out)
   amt.resamples <- length(eval.train.dexeses)	
   
   for (i.sample in seq(1, amt.resamples)) {
diff --git a/man/CST_Calibration.Rd b/man/CST_Calibration.Rd
index 491b727195638d98ef7f7a891825696fbbaa5f9e..6843f1c9a244212e8071a317c9fa0b0d994c0bb4 100644
--- a/man/CST_Calibration.Rd
+++ b/man/CST_Calibration.Rd
@@ -18,7 +18,9 @@ CST_Calibration(
   memb_dim = "member",
   sdate_dim = "sdate",
   dat_dim = NULL,
-  ncores = NULL
+  ncores = NULL,
+  k = 1, 
+  tail.out = TRUE
 )
 }
 \arguments{
diff --git a/man/CST_CategoricalEnsCombination.Rd b/man/CST_CategoricalEnsCombination.Rd
index 85ebb7f8f00c0c393418d341af002cfee464a726..416ad7f82e4172f0f32dfd96660104b53e76b6e0 100644
--- a/man/CST_CategoricalEnsCombination.Rd
+++ b/man/CST_CategoricalEnsCombination.Rd
@@ -11,6 +11,8 @@ CST_CategoricalEnsCombination(
   cat.method = "pool",
   eval.method = "leave-one-out",
   amt.cat = 3,
+  k = 1,
+  tail.out = TRUE,
   ...
 )
 }
diff --git a/tests/testthat/test-CST_Calibration.R b/tests/testthat/test-CST_Calibration.R
index 491aff29556fc0a67f4abc9fb5ad11500aca272d..285e81c8907cd457219d6af3ea881577683baff9 100644
--- a/tests/testthat/test-CST_Calibration.R
+++ b/tests/testthat/test-CST_Calibration.R
@@ -218,8 +218,8 @@ test_that("1. Input checks", {
   expect_error(
     Calibration(exp4, obs4, eval.method = 'biass'),
     paste0("Parameter 'eval.method' must be a character string indicating ", 
-           "the sampling method used ('in-sample', 'leave-one-out' or ", 
-           "'hindcast-vs-forecast')."), 
+           "the sampling method used ('in-sample', 'leave-one-out', 'hindcast-vs-forecast', 'k-fold' or ", 
+           "'retrospective')."), 
     fixed = TRUE
   )
   # multi.model
@@ -453,6 +453,18 @@ test_that("6. Output checks: dat4", {
     c(-0.7119142, 0.2626203, -0.9635483, 1.9607986, 0.4380930),
     tolerance = 0.0001
   )
+  expect_equal(
+    as.vector(Calibration(exp4, obs4, eval.method = "k-fold", k = 5, tail.out = T))[1:5],
+    c(-0.8149557,  0.2043942, -1.0781616,  1.9806657,  0.3879362),
+    tolerance = 0.0001
+  )
+
+  expect_equal(
+    as.vector(Calibration(exp4, obs4, eval.method = "retrospective", k = 3)[1,1:5]),
+    c(NA, NA, NA, 1.6267865, -0.1658674),
+    tolerance = 0.0001
+  )
+  
 })
 
 ##############################################
diff --git a/tests/testthat/test-make-eval-train-dexes.R b/tests/testthat/test-make-eval-train-dexes.R
new file mode 100644
index 0000000000000000000000000000000000000000..a74a06c0fa18506bf3c0c0751194f611543c711e
--- /dev/null
+++ b/tests/testthat/test-make-eval-train-dexes.R
@@ -0,0 +1,93 @@
+##############################################
+
+test_that("1. Input checks", {
+  # s2dv_cube
+  expect_error(
+    .make.eval.train.dexes(eval.method = 1, 2),
+    paste0("unknown sampling method: 1")
+  )
+  expect_error(
+   .make.eval.train.dexes(eval.method = 'k-fold', 10, NULL, 12),
+    paste0("k needs to be a positive integer less than the amt.points")
+  )
+  expect_error(
+   .make.eval.train.dexes(eval.method = 'retrospective', 10, NULL, 10),
+    paste0("k needs to be a positive integer less than the amt.points")
+  )
+})
+
+##############################################
+
+test_that("2. Output checks: k-fold", {
+  expect_equal(
+    is.list(.make.eval.train.dexes(eval.method = 'k-fold', 10, NULL, 5)),
+    TRUE
+  )
+  expect_equal(
+   is.list(.make.eval.train.dexes(eval.method = 'k-fold', 10, NULL, 5)[[1]]),
+    TRUE
+  )
+  expect_equal(
+   length(.make.eval.train.dexes(eval.method = 'k-fold', 10, NULL, 5)),
+    10
+  )
+  expect_equal(
+   names(.make.eval.train.dexes(eval.method = 'k-fold', 10, NULL, 5)[[1]]),
+    c("eval.dexes", "train.dexes")
+  )
+  expect_equal(
+   .make.eval.train.dexes(eval.method = 'k-fold', 10, NULL, 5)[[1]],
+    list(eval.dexes = 1,
+         train.dexes = 4:8)
+  )
+   expect_equal(
+   .make.eval.train.dexes(eval.method = 'k-fold', 10, NULL, 5)[[2]],
+    list(eval.dexes = 2, 
+         train.dexes = 5:9)
+  )
+   expect_equal(
+   .make.eval.train.dexes(eval.method = 'k-fold', 10, NULL, 5)[[3]],
+    list(eval.dexes = 3,
+         train.dexes = 6:10)
+  )
+   expect_equal(
+   .make.eval.train.dexes(eval.method = 'k-fold', 10, NULL, 5)[[9]],
+    list(eval.dexes = 9,
+         train.dexes = 2:6)
+  )
+   expect_equal(
+   .make.eval.train.dexes(eval.method = 'k-fold', 10, NULL, 5)[[10]],
+    list(eval.dexes = 10,
+         train.dexes = 3:7)
+  )
+  expect_equal(
+   .make.eval.train.dexes(eval.method = 'k-fold', 10, NULL, 1),
+   .make.eval.train.dexes(eval.method = 'leave-one-out', 10, NULL)
+  )
+  expect_equal(
+   .make.eval.train.dexes(eval.method = 'k-fold', 20, NULL, 1),
+   .make.eval.train.dexes(eval.method = 'leave-one-out', 20, NULL)
+  )
+})
+
+##############################################
+
+test_that("3. Output checks: retrospective", {
+  expect_equal(
+    length(.make.eval.train.dexes(eval.method = 'retrospective', 10, NULL, 3)),
+    7
+  )
+  expect_equal(
+    .make.eval.train.dexes(eval.method = 'retrospective', 10, NULL, 3)[[1]],
+    list(eval.dexes = 4, train.dexes = 1:3)
+  )
+  expect_equal(
+    .make.eval.train.dexes(eval.method = 'retrospective', 10, NULL, 3)[[2]],
+    list(eval.dexes = 5, train.dexes = 1:4)
+  )
+  expect_equal(
+    .make.eval.train.dexes(eval.method = 'retrospective', 10, NULL, 3)[[7]],
+    list(eval.dexes = 10, train.dexes = 1:9)
+  )
+})
+