Merge pull request #121 from Burke-Lauenroth-Lab/expose_calc_of_evapc…

…oeff

Expose and fix calculation of evapcoeff as funtion

- pull request/commit passes all CI builds by travis and appveyor

- pull request passes all test projects (except 6 which is currently not set up)
>       elapsed_s has_run has_problems made_new_refs deleted_output
> Test1  1191.355    TRUE         TRUE         FALSE          FALSE
> Test2   942.577    TRUE        FALSE         FALSE          FALSE
> Test3   174.627    TRUE        FALSE         FALSE          FALSE
> Test4    42.126    TRUE        FALSE         FALSE          FALSE
> Test5    46.085    TRUE        FALSE         FALSE          FALSE
> Test6     0.000   FALSE         TRUE         FALSE          FALSE
> Test7   305.494    TRUE        FALSE         FALSE          FALSE
> Test8   248.107    TRUE        FALSE         FALSE          FALSE
>                                                     referenceDB
> Test1        dbTables_Test1_downscaling_overhaul_v2.1.1.sqlite3
> Test2        dbTables_Test2_LookupWeatherFolders_v2.1.1.sqlite3
> Test3         dbTables_Test3_OnlyMeanDailyOutput_v2.1.1.sqlite3
> Test4 dbTables_Test4_AllOverallAggregations_snow_v2.1.1.sqlite3
> Test5  dbTables_Test5_AllOverallAggregations_mpi_v2.1.1.sqlite3
> Test6                                                          
> Test7           dbTables_Test7_livneh_extraction_v2.1.1.sqlite3
> Test8          dbTables_Test8_ExperimentalDesign_v2.1.1.sqlite3

- pull request passes all unit tests when run locally with devtools::test()
> Testing rSFSW2
> dbOutput: action: check: .............
> dbWork: runIDs organization: 
> Delta-hybrid (mod3) downscaling: add_delta_to_PPT: ..............
> Delta-hybrid (mod3) downscaling: applyPPTdelta_simple and applyPPTdelta_detailed: .........................
> Delta-hybrid (mod3) downscaling: fix_PPTdata_length: ...........
> GISSM: germination_wait_times: .............
> GISSM: get_KilledBySoilLayers: ...................
> GISSM: setFALSE_SeedlingSurvival_1stSeason: .........
> Indices: .........................................................................................................
> Input/Output of rSFSW2 datafiles: ...........
> lints: S
> Match for appending data: ....
> Pedotransfer functions: SWP <-> VWC: ...................................................................................................................................................
> Soil functions: ..............................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................
> Manipulate soil layers: .............................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................
> 
> Skipped -------------------------------------------------
> 1. Package Style (@test_lint.R#4) - rSFSW2 is not ready for 'lint' testing...
> 
> DONE ====================================================

.travis.yml

@@ -1,13 +1,21 @@
 # R for travis: see documentation at https://docs.travis-ci.com/user/languages/r
 
 language: R
+env:
+  - _R_CHECK_FORCE_SUGGESTS_=false # because of Rmpi, ncdf4, hydromat, and weathergen
 sudo: false
-cache: packages
-warnings_are_errors: false # this is only needed for rSOILWAT2
+cache:
+  - packages
+warnings_are_errors: false # turn this back on for more severe testing
+
 before_install:
+  # install dependencies of rSOILWAT2 (because rSOILWAT2 is installed 'before install' and
+  #   thus dependencies are not installed and r_packages has not taken effect yet)
+  - Rscript -e 'utils::install.packages(c("blob", "DBI", "RSQLite"))'
+  # obtain source code of rSOILWAT2 from github
   - git clone -b master --single-branch --recursive https://github.com/Burke-Lauenroth-Lab/rSOILWAT2.git /tmp/rSOILWAT2
-  - cd /tmp/rSOILWAT2
-  - ./tools/TarAndInstall_31n.sh
+  # install rSOILWAT2
+  - R CMD INSTALL /tmp/rSOILWAT2
 
 after_success:
   - Rscript -e 'covr::codecov()'

DESCRIPTION

@@ -1,7 +1,7 @@
 Package: rSFSW2
 Title: Simulation Framework for SOILWAT2
-Version: 2.2.0
-Date: 2017-07-20
+Version: 2.3.0
+Date: 2017-07-23
 Authors@R: c(person("Daniel", "Schlaepfer", email = "daniel.schlaepfer@yale.edu", role = c("aut", "cre")),
       person("Caitlin", "Andrews", role = "ctb"),
       person("Zach", "Kramer", role = "ctb"),
@@ -14,7 +14,7 @@ Imports:
     rSOILWAT2 (>= 1.3.4),
     RSQLite (>= 1.1),
     DBI (>= 0.5),
-    Rcpp (>= 0.12.7),
+    Rcpp (>= 0.12.12),
     raster (>= 2.5.8),
     sp (>= 1.2.3),
     methods

NAMESPACE

@@ -25,7 +25,7 @@ export(TranspCoeffByVegType)
 export(VWCtoSWP)
 export(align_with_target_grid)
 export(align_with_target_res)
-export(calc_CalculateBareSoilEvaporationCoefficientsFromSoilTexture)
+export(calc_BareSoilEvaporationCoefficientsFromSoilTexture)
 export(calc_ExtendSoilDatafileToRequestedSoilLayers)
 export(calc_SiteClimate)
 export(check_lock_content)
@@ -40,6 +40,7 @@ export(dbOutput_ListDesignTables)
 export(dbOutput_ListOutputTables)
 export(dbOutput_Tables_have_SoilLayers)
 export(dbW_has_missingClimScens)
+export(dbW_missingAmbient_siteIDs)
 export(dbWork_check)
 export(dbWork_redo)
 export(dbWork_timing)
@@ -136,4 +137,4 @@ importFrom(methods,setGeneric)
 importFrom(methods,setMethod)
 importFrom(methods,signature)
 importFrom(methods,slot)
-useDynLib(rSFSW2, .registration = TRUE, .fixes = "C_")
+useDynLib(rSFSW2, .registration = TRUE)

R/ExtractData_ClimateDownscaling.R

@@ -2938,9 +2938,10 @@ PrepareClimateScenarios <- function(SFSW2_prj_meta, SFSW2_prj_inputs, opt_parall
     SFSW2_prj_meta[["sim_scens"]][["sources"]])
 
   if (resume) {
-    todos <- dbW_has_missingClimScens(fdbWeather = SFSW2_prj_meta[["fnames_in"]][["fdbWeather"]],
-      SFSW2_prj_inputs, req_scenN = SFSW2_prj_meta[["sim_scens"]][["N"]], opt_verbosity,
-      opt_chunks)
+    todos <- dbW_has_missingClimScens(
+      fdbWeather = SFSW2_prj_meta[["fnames_in"]][["fdbWeather"]], SFSW2_prj_inputs,
+      runIDs_sites = SFSW2_prj_meta[["sim_size"]][["runIDs_sites"]],
+      req_scenN = SFSW2_prj_meta[["sim_scens"]][["N"]], opt_verbosity, opt_chunks)
 
   } else {
     todos <- SFSW2_prj_inputs[["include_YN"]] &

R/ExtractData_Soils.R

@@ -486,7 +486,7 @@ ISRICWISE_try_weightedMeanForSimulationCell <- function(i, sim_cells_SUIDs,
 #'   modelling (WISE30sec) with estimates of global soil carbon stocks. Geoderma 269,
 #'   61-68 (http://dx.doi.org/10.1016/j.geoderma.2016.01.034).
 #' @references Batjes N.H. 2015. World soil property estimates for broad-scale modelling
-#'   (WISE30sec, ver. 1.0). Report 2015/01, ISRIC—World Soil Information, Wageningen
+#'   (WISE30sec, ver. 1.0). Report 2015/01, ISRIC-World Soil Information, Wageningen
 #'   [available at ISRIC Soil Data Hub](http://geonode.isric.org/search/?title__icontains=World%20soil%20property%20estimates%20for%20broad-scale%20modelling%20(WISE30sec)&limit=100&offset=0),
 #'   with addendum and corrigendum.
 #'
@@ -625,7 +625,7 @@ do_ExtractSoilDataFromISRICWISE_Global <- function(MMC, sim_size, sim_space,
           res = cell_res_wise, grid = grid_wise, sp_sites = run_sites_wise,
           att = rat_att)
       }
-      
+
       sim_cells_SUIDs <- do.call(rbind, sim_cells_SUIDs)
     }
 

R/Miscellaneous_Functions.R

@@ -1026,7 +1026,23 @@ tabulate_values_in_bins <- function(x, method = c("duration", "values"),
   list(eventsPerYear = eventsPerYear, freq.summary = freq.summary)
 }
 
-
+#' NAs present but only in deepest soil layers
+#'
+#' Checks that NAs are present and that NAs occur only grouped together
+#' in the right-most columns per row (e.g., deepest soil layers if columns represent
+#' soil layers and rows represent sites).
+#'
+#' @param x A data.frame, matrix, or array with at least two dimensions.
+#'
+#' @return A logical vector of length equal to the first dimension of \code{x} with
+#'  \code{TRUE} if there are n[k] \code{NA}s in the k-th row and they occupy the k
+#'  rightmost columns.
+#'
+has_NAs_pooled_at_depth <- function(x) {
+  stopifnot(!is.null(dim(x)))
+  temp <- apply(x, 1, function(dat) rle(is.na(dat)))
+  sapply(temp, function(dat) length(dat$values) <= 2 && dat$values[length(dat$values)])
+}
 
 
 benchmark_BLAS <- function(platform, seed = NA) {

R/PriorCalculations.R

@@ -111,10 +111,10 @@ calc_ExtendSoilDatafileToRequestedSoilLayers <- function(SFSW2_prj_meta, SFSW2_p
 }
 
 
-#' Calculate potential bare soil evaporation coefficients per soil layer
+#' Calculate potential bare-soil evaporation coefficients
 #'
 #' Soil texture influence based on re-analysis of data from Wythers et al. 1999.
-#' Default of depth_max_bs_evap = 15 cm from Torres et al. 2010.
+#' Default of \code{depth_max_bs_evap} = 15 cm from Torres et al. 2010.
 #'
 #' @references Torres EA, Calera A (2010) Bare soil evaporation under high evaporation
 #'  demand: a proposed modification to the FAO-56 model. Hydrological Sciences Journal-
@@ -123,8 +123,111 @@ calc_ExtendSoilDatafileToRequestedSoilLayers <- function(SFSW2_prj_meta, SFSW2_p
 #' @references Wythers KR, Lauenroth WK, Paruelo JM (1999) Bare-Soil Evaporation Under
 #'  Semiarid Field Conditions. Soil Science Society of America Journal, 63, 1341-1349.
 #'
+#' @param layers_depth A numeric vector, matrix, or data.frame. Values describe the lower
+#'  soil layer depths in units of centimeters.
+#' @param sand A numeric vector, matrix, or data.frame. Values are sand contents in units
+#'  of mass-percentage / 100.
+#' @param clay A numeric vector, matrix, or data.frame. Values are clay contents in units
+#'  of mass-percentage / 100.
+#' @param depth_max_bs_evap_cm A numeric value. The maximal soil depth in centimeters from
+#'  which bare-soil evaporation is potentially drawing moisture.
+#'
+#' @section Notes: Rows of soil input arguments \code{layers_depth}, \code{sand}, and
+#'  \code{clay} correspond to sites and columns to soil layers. If \code{sand} and/or
+#'  \code{clay} are vectors, then they are converted to 1-row matrices.
+#'  If \code{layers_depth} is a vector, then it is converted to a matrix with as many
+#'  sites/rows as \code{sand} and \code{clay} have. That is the code assumes identical
+#'  soil layer depths for each site. All soil input arguments must have a the same number
+#'  of sites and of soil layers, i.e., identical matrix dimensions.
+#'
+#' @return A numeric matrix with potential bare-soil evaporation coefficients where rows
+#'  correspond to sites and columns to soil layers.
+#'
 #' @export
-calc_CalculateBareSoilEvaporationCoefficientsFromSoilTexture <- function(SFSW2_prj_meta,
+calc_BareSoilEvaporationCoefficientsFromSoilTexture <- function(layers_depth, sand, clay,
+  depth_max_bs_evap_cm = 15) {
+
+  #--- If inputs are not site x layers, then convert them into 1 site x layers table
+  if (is.null(dim(sand))) {
+    sand <- matrix(sand, nrow = 1, ncol = length(sand))
+  }
+  if (is.null(dim(clay))) {
+    clay <- matrix(clay, nrow = 1, ncol = length(clay))
+  }
+  if (is.null(dim(layers_depth))) {
+    layers_depth <- matrix(layers_depth, nrow = dim(sand)[1], ncol = length(layers_depth),
+      byrow = TRUE)
+  }
+
+  #--- Test inputs
+  # - sand and clay have identical number of sites and layers
+  # - all soil inputs have identical number of sites and at least as many layers as depths
+  # - soil layer depths are numeric and positive -- or NA, if all deeper layers are NA
+  # - sand and clay are numeric and values between 0 and 1 -- or NA, if all deeper
+  #   layers are NA as well
+  # - the sum of sand and clay is less or equal to 1
+  sand_and_clay <- sand + clay
+  stopifnot(
+    identical(dim(sand), dim(clay)),
+    identical(dim(sand)[1], dim(layers_depth)[1]), dim(sand)[2] >= dim(layers_depth)[2],
+    is.numeric(layers_depth), layers_depth > 0 | has_NAs_pooled_at_depth(layers_depth),
+    is.numeric(unlist(sand)), sand >= 0 & sand <= 1 | has_NAs_pooled_at_depth(sand),
+    is.numeric(unlist(clay)), clay >= 0 & clay <= 1 | has_NAs_pooled_at_depth(clay),
+    sand_and_clay <= 1 | has_NAs_pooled_at_depth(sand_and_clay),
+    is.finite(depth_max_bs_evap_cm) & depth_max_bs_evap_cm >= 0)
+
+
+  #--- Calculate
+
+  depth_min_bs_evap <- min(layers_depth[, 1], na.rm = TRUE)
+  if (depth_min_bs_evap > depth_max_bs_evap_cm) {
+    # all sites have first layer with coeff = 1
+    res <- array(1, dim = dim(sand))
+    res[, -1] <- 0
+    return(res)
+  }
+
+  lyrs_max_bs_evap <- t(apply(layers_depth, 1, function(x) {
+    xdm <- depth_max_bs_evap_cm - x
+    i0 <- abs(xdm) < SFSW2_glovars[["tol"]]
+    ld <- if (any(i0, na.rm = TRUE)) {
+      which(i0)
+    } else {
+      temp <- which(xdm < 0)
+      if (length(temp) > 0) temp[1] else length(x)
+    }
+    c(diff(c(0, x))[seq_len(ld)], rep(0L, length(x) - ld))
+  }))
+  ldepth_max_bs_evap <- rowSums(lyrs_max_bs_evap)
+
+  sand_mean <- rowSums(lyrs_max_bs_evap * sand, na.rm = TRUE) / ldepth_max_bs_evap
+  clay_mean <- rowSums(lyrs_max_bs_evap * clay, na.rm = TRUE) / ldepth_max_bs_evap
+
+  temp_depth <- 4.1984 + 0.6695 * sand_mean ^ 2 + 168.7603 * clay_mean ^ 2 # equation from re-analysis
+  depth_bs_evap <- pmin(pmax(temp_depth, depth_min_bs_evap, na.rm = TRUE),
+    depth_max_bs_evap_cm, na.rm = TRUE)
+  lyrs_bs_evap0 <- t(apply(depth_bs_evap - layers_depth, 1, function(x) {
+    i0 <- abs(x) < SFSW2_glovars[["tol"]]
+    ld <- if (any(i0, na.rm = TRUE)) {
+      which(i0)
+    } else {
+      temp <- which(x < 0)
+      if (length(temp) > 0) temp[1] else sum(!is.na(x))
+    }
+    ld0 <- max(0, ld - 1)
+
+    c(rep(TRUE, ld0), rep(FALSE, length(x) - ld0))
+  }))
+
+  # function made up to match previous cummulative distributions
+  temp_coeff <- 1 - exp(- 5 * layers_depth / depth_bs_evap)
+  temp_coeff[!lyrs_bs_evap0 | is.na(temp_coeff)] <- 1
+  coeff_bs_evap <- round(t(apply(cbind(0, temp_coeff), 1, diff)), 4)
+  coeff_bs_evap / rowSums(coeff_bs_evap, na.rm = TRUE)
+}
+
+
+get_BareSoilEvaporationCoefficientsForSoilInputFile <- function(SFSW2_prj_meta,
   SFSW2_prj_inputs, runIDs_adjust, resume = TRUE, verbose = FALSE) {
 
   if (verbose) {
@@ -158,47 +261,14 @@ calc_CalculateBareSoilEvaporationCoefficientsFromSoilTexture <- function(SFSW2_p
     temp <- SFSW2_prj_inputs[["sw_input_soillayers"]][runIDs_adjust, grep("depth_L",
       names(SFSW2_prj_inputs[["sw_input_soillayers"]]))[use_layers], drop = FALSE]
     layers_depth <- as.matrix(temp)
-    depth_min_bs_evap <- min(layers_depth[, 1])
-    stopifnot(stats::na.exclude(depth_min_bs_evap < SFSW2_prj_meta[["opt_sim"]][["depth_max_bs_evap_cm"]]))
-
-    lyrs_max_bs_evap <- t(apply(layers_depth, 1, function(x) {
-      xdm <- SFSW2_prj_meta[["opt_sim"]][["depth_max_bs_evap_cm"]] - x
-      i0 <- abs(xdm) < SFSW2_glovars[["tol"]]
-      ld <- if (any(i0, na.rm = TRUE)) {
-        which(i0)
-      } else {
-        temp <- which(xdm < 0)
-        if (length(temp) > 0) temp[1] else length(x)
-      }
-      c(diff(c(0, x))[seq_len(ld)], rep(0L, length(x) - ld))
-    }))
-    ldepth_max_bs_evap <- rowSums(lyrs_max_bs_evap)
 
     #TODO: add influence of gravel
     sand <- SFSW2_prj_inputs[["sw_input_soils"]][runIDs_adjust, icol_sand, drop = FALSE]
     clay <- SFSW2_prj_inputs[["sw_input_soils"]][runIDs_adjust, icol_clay, drop = FALSE]
-    sand_mean <- rowSums(lyrs_max_bs_evap * sand, na.rm = TRUE) / ldepth_max_bs_evap
-    clay_mean <- rowSums(lyrs_max_bs_evap * clay, na.rm = TRUE) / ldepth_max_bs_evap
-
-    temp_depth <- 4.1984 + 0.6695 * sand_mean ^ 2 + 168.7603 * clay_mean ^ 2 # equation from re-analysis
-    depth_bs_evap <- pmin(pmax(temp_depth, depth_min_bs_evap, na.rm = TRUE),
-      SFSW2_prj_meta[["opt_sim"]][["depth_max_bs_evap_cm"]], na.rm = TRUE)
-    lyrs_bs_evap <- t(apply(depth_bs_evap - layers_depth, 1, function(x) {
-      i0 <- abs(x) < SFSW2_glovars[["tol"]]
-      ld <- if (any(i0, na.rm = TRUE)) {
-        which(i0)
-      } else {
-        temp <- which(x < 0)
-        if (length(temp) > 0) temp[1] else sum(!is.na(x))
-      }
-      c(rep(TRUE, ld), rep(FALSE, length(x) - ld))
-    }))
-
-    # function made up to match previous cummulative distributions
-    temp_coeff <- 1 - exp(- 5 * layers_depth / depth_bs_evap)
-    temp_coeff[!lyrs_bs_evap | is.na(temp_coeff)] <- 1
-    coeff_bs_evap <- round(t(apply(cbind(0, temp_coeff), 1, diff)), 4)
-    coeff_bs_evap <- coeff_bs_evap / rowSums(coeff_bs_evap, na.rm = TRUE)
+
+    coeff_bs_evap <- calc_BareSoilEvaporationCoefficientsFromSoilTexture(layers_depth, sand, clay,
+      depth_max_bs_evap_cm = SFSW2_prj_meta[["opt_sim"]][["depth_max_bs_evap_cm"]])
+
 
     #add data to sw_input_soils and set the use flags
     icol <- seq_len(sum(apply(coeff_bs_evap, 2, function(x) any(x > SFSW2_glovars[["tol"]]))))

R/RcppExports.R

@@ -30,7 +30,7 @@
 #'
 #' @export
 germination_wait_times <- function(time_to_germinate, duration_fave_cond) {
-    .Call(C_rSFSW2_germination_wait_times, time_to_germinate, duration_fave_cond, PACKAGE = "rSFSW2")
+    .Call(`_rSFSW2_germination_wait_times`, time_to_germinate, duration_fave_cond)
 }
 
 #' Determine if all conditions across rooted soil layers are deadly
@@ -74,7 +74,7 @@ germination_wait_times <- function(time_to_germinate, duration_fave_cond) {
 #'
 #' @export
 get_KilledBySoilLayers <- function(relevantLayers, kill_conditions) {
-    .Call(C_rSFSW2_get_KilledBySoilLayers, relevantLayers, kill_conditions, PACKAGE = "rSFSW2")
+    .Call(`_rSFSW2_get_KilledBySoilLayers`, relevantLayers, kill_conditions)
 }
 
 #' Determine seedling survival in the first season (ss1s)
@@ -100,6 +100,6 @@ get_KilledBySoilLayers <- function(relevantLayers, kill_conditions) {
 #'
 #' @export
 setFALSE_SeedlingSurvival_1stSeason <- function(ss1s, ry_year_day, ry_useyrs, y, doy) {
-    .Call(C_rSFSW2_setFALSE_SeedlingSurvival_1stSeason, ss1s, ry_year_day, ry_useyrs, y, doy, PACKAGE = "rSFSW2")
+    .Call(`_rSFSW2_setFALSE_SeedlingSurvival_1stSeason`, ss1s, ry_year_day, ry_useyrs, y, doy)
 }
 

R/Simulation_Project.R

@@ -494,7 +494,7 @@ populate_rSFSW2_project_with_data <- function(SFSW2_prj_meta, opt_behave, opt_pa
   if (SFSW2_prj_meta[["pcalcs"]][["CalculateBareSoilEvaporationCoefficientsFromSoilTexture"]]) {
     if (todo_intracker(SFSW2_prj_meta, "calc_bsevap", "prepared")) {
 
-      SFSW2_prj_inputs <- calc_CalculateBareSoilEvaporationCoefficientsFromSoilTexture(
+      SFSW2_prj_inputs <- get_BareSoilEvaporationCoefficientsForSoilInputFile(
         SFSW2_prj_meta, SFSW2_prj_inputs, runIDs_adjust, resume = opt_behave[["resume"]],
         verbose = opt_verbosity[["verbose"]])
 

R/WeatherDB_Check.R

@@ -63,15 +63,15 @@ dbW_missingAmbient_siteIDs <- function(req_sites, fdbWeather, dbW_iSiteTable = N
 #'  all data are in the weather database, then \code{FALSE}.
 #'
 #' @export
-dbW_has_missingClimScens <- function(fdbWeather, SFSW2_prj_inputs, req_scenN,
+dbW_has_missingClimScens <- function(fdbWeather, SFSW2_prj_inputs, runIDs_sites, req_scenN,
   opt_verbosity, opt_chunks) {
   # Init: set every included site as having missing climate scenario data. If a site_id
   #   is found to have all climate scenario data, then set its todo element to FALSE
   todos <- SFSW2_prj_inputs[["include_YN"]]
 
   # Determine 'site_id's for which the dbWeather should contain daily weather data for
   # every climate scenario
-  req_siteIDs <- SFSW2_prj_meta[["sim_size"]][["runIDs_sites"]] # equal to SFSW2_prj_inputs[["SWRunInformation"]][todos, "site_id"]
+  req_siteIDs <- runIDs_sites # runIDs_sites equal to SFSW2_prj_meta[["sim_size"]][["runIDs_sites"]] and equal to SFSW2_prj_inputs[["SWRunInformation"]][todos, "site_id"]
 
   # Check presence of records with climate scenario data in dbWeather
   con <- DBI::dbConnect(RSQLite::SQLite(), dbname = fdbWeather, flags = RSQLite::SQLITE_RO)

R/rSFSW2-package.R

@@ -38,7 +38,7 @@
 #'    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 #'    GNU General Public License for more details.
 #'
-#' @useDynLib rSFSW2, .registration = TRUE, .fixes = "C_"
+#' @useDynLib rSFSW2, .registration = TRUE
 #' @docType package
 #' @name rSFSW2
 "_PACKAGE"

README.md

@@ -61,12 +61,19 @@ system2(command = "git", args = "clone -b master --single-branch https://github.
 tools::Rcmd(args = paste("INSTALL rSFSW2"))
 ```
 
+### Binary package version
 If you want a binary version of the 'rSFSW2' package (e.g., to distribute to someone
 without development tools) for a platform to which you do not have access, then you may
 consider using one of the cloud services (no endorsements):
 - https://builder.r-hub.io offers different Linux, Windows, and mac OS flavors as targets
 - http://win-builder.r-project.org/ offers Windows OS as target
 
+Alternatively, you may access the previous binary package version for Windows OS from our
+CI appveyor service if the build was successful and an artifact was generated for the
+binary package (this would be named 'rSWSF2_X.Y.Z.zip' with version number X.Y.Z) at
+- https://ci.appveyor.com/project/dschlaep/soilwat-r-wrapper/build/artifacts
+If the latest build should have failed, then you may want to check out the 'History' tab
+for binaries of older versions.
 
 # Use rSFSW2 for your simulation project