Title:	Reconstruction of Daily Data - Precipitation
Version:	3.0.0
Description:	Applies quality control to daily precipitation observations; reconstructs the original series by estimating precipitation in missing values; and creates gridded datasets of daily precipitation.
License:	GPL-3
Encoding:	UTF-8
Imports:	terra, stats, foreach, doParallel, reshape, qmap, BreakPoints, Kendall, car, geosphere, gridExtra, lattice, pracma, xts, zoo, e1071, neuralnet, randomForest, reshape2, xgboost
RoxygenNote:	7.3.2
NeedsCompilation:	no
Packaged:	2025-04-12 20:02:10 UTC; rserrano
Author:	Roberto Serrano-Notivoli [aut, cre], Adrian Huerta [aut], Abel Centella [ctb]
Maintainer:	Roberto Serrano-Notivoli <roberto.serrano@unizar.es>
Repository:	CRAN
Date/Publication:	2025-04-24 11:30:02 UTC

Enhanced quality control plots for daily precipitation

Description

The function create a matrix plot of the enhanced quality control tests for daily precipitation

Usage

eqc_Plot(xts_obj)

Arguments

Details

Six plots are built based on daily precipitation: time series, truncation, time series (threshold = 5 mm), small gaps, precision and rounding patterns and weekly cycle. These provide a visual inspection (Hunziker et al., 2017) but also how the automatic application (Huerta et al., 2020) of the enhanced quality control is applied.

References

Hunziker, S., Gubler, S., Calle, J., Moreno, I., Andrade, M., Velarde, F., ... & Brönnimann, S. (2017). Identifying, attributing, and overcoming common data quality issues of manned station observations. https://doi.org/10.1002/joc.5037

Huerta, A., Serrano-Notivoli, R., & Brönnimann, S. (2024). SC-PREC4SA: A serially complete daily precipitation dataset for South America. https://doi.org/10.31223/X57D8R

Examples

## Not run: 
set.seed(123)

# creating fake daily precipitation data
dates_t <- seq(as.Date("1980-01-01"), as.Date("2015-12-31"), by = "day")
prec <- round(rnorm(length(dates_t), mean = 1.2, sd = 6), 1)
prec[prec<0] <- 0
xts_obj <- xts::xts(prec, dates_t)
names(xts_obj) <- "prec"

# enhanced qc plots
eqc_Plot(xts_obj)

# it also work if there is some empty data (but not if all is NA)
xts_obj["1990/2010"] <- NA
eqc_Plot(xts_obj)

## End(Not run)

Precision and rounding patterns of daily precipitation

Description

The function determine the level (0, 1 or 2) of the precision and rounding patterns test

Usage

eqc_PrecisionRounding(
  xts_obj,
  lmn_yday = 365 * 80/100,
  ths = list(lv0 = c(percent = 70), lv1to2 = c(percent = 50))
)

Arguments

Details

Precision and rounding patterns depict inconsistencies in the frequency of decimal values in the time series. As there is no absolute correct frequency of decimals, we decided to measure how similar the decimal patterns are in the time series. A decimal pattern is interpreted as the list of unique decimal values observed, sorted in descending order. In this way, the decimal pattern for each year is computed first, followed by the selection of the most dominating pattern (mode). Based on how much (in percentage) this dominating pattern represents the time series, it is defined: Level 0: coherent precision and rounding pattern (similar decimal pattern in more (or equal) than 70% of the time series). Level 1: a similar decimal pattern in less than 70% but more than (or equal) 50% of the time series. Level 2: different decimal patterns (no dominant pattern). Argument ths present the above default thresholds for the levels definition.

References

Hunziker, S., Gubler, S., Calle, J., Moreno, I., Andrade, M., Velarde, F., ... & Brönnimann, S. (2017). Identifying, attributing, and overcoming common data quality issues of manned station observations. https://doi.org/10.1002/joc.5037

Huerta, A., Serrano-Notivoli, R., & Brönnimann, S. (2024). SC-PREC4SA: A serially complete daily precipitation dataset for South America. https://doi.org/10.31223/X57D8R

Examples

## Not run: 
set.seed(123)

# creating fake daily precipitation data
dates_t <- seq(as.Date("1980-01-01"), as.Date("2015-12-31"), by = "day")
prec <- round(rnorm(length(dates_t), mean = 1.2, sd = 6), 1)
prec[prec<0] <- 0
xts_obj <- xts::xts(prec, dates_t)
names(xts_obj) <- "prec"

# Comparison with visual inspection
eqc_Plot(xts_obj)
eqc_PrecisionRounding(xts_obj)

# it also work if there is some empty data (but not if all is NA)
xts_obj["1990/2010"] <- NA
eqc_Plot(xts_obj)
eqc_PrecisionRounding(xts_obj)

# creating some rounding values
xts_obj["1980/1990"] <- round(xts_obj["1980/1990"], 0) 
eqc_Plot(xts_obj)
eqc_PrecisionRounding(xts_obj)


## End(Not run)

Small gaps level of daily precipitation

Description

The function determine the level (0, 1 or 2) of the small gaps test

Usage

eqc_SmallGaps(
  xts_obj,
  lmn_yday = 365 * 80/100,
  ths = list(lv0 = c(percent = 0), lv1to2 = c(percent = 20))
)

Arguments

Details

Small gaps can be seen as unreported precipitation events that result in a gap or a frequency reduction in values below a specific threshold. To define the small gaps, we calculated the total count of values in five precipitation ranges from 0-1, 1-2, 2-3, 3-4, and 4-5 mm (not including the values in the limits) for each year. Therefore, considering the percentage of years with zero counts: Level 0: no small gaps (0%; years with at least one value in any of the precipitation ranges). Level 1: small gaps in at least 20% of years. Level 2: small gaps in more than 20% of years Argument ths present the above default thresholds for the levels definition.

References

Hunziker, S., Gubler, S., Calle, J., Moreno, I., Andrade, M., Velarde, F., ... & Brönnimann, S. (2017). Identifying, attributing, and overcoming common data quality issues of manned station observations. https://doi.org/10.1002/joc.5037

Huerta, A., Serrano-Notivoli, R., & Brönnimann, S. (2024). SC-PREC4SA: A serially complete daily precipitation dataset for South America. https://doi.org/10.31223/X57D8R

Examples

## Not run: 
set.seed(123)

# creating fake daily precipitation data
dates_t <- seq(as.Date("1980-01-01"), as.Date("2015-12-31"), by = "day")
prec <- round(rnorm(length(dates_t), mean = 1.2, sd = 6), 1)
prec[prec<0] <- 0
prec[prec<3 & prec>2] <- 0
xts_obj <- xts::xts(prec, dates_t)
names(xts_obj) <- "prec"

# Comparison with visual inspection
eqc_Plot(xts_obj)
eqc_SmallGaps(xts_obj)

# it also work if there is some empty data (but not if all is NA)
xts_obj["1990/2010"] <- NA
eqc_Plot(xts_obj)
eqc_SmallGaps(xts_obj)


## End(Not run)

Truncation level of daily precipitation

Description

The function determine the level (0, 1 or 2) of the truncation test

Usage

eqc_Truncation(
  xts_obj,
  ths = list(lv0 = c(n_years = 2), lv1to2 = c(n_years_l = 3, n_years_h = 5))
)

Arguments

Details

Truncation is when heavy precipitation episodes are truncated or noticeably reduced in frequency above a given threshold. Because there is no preceding algorithm for truncation, it is defined here as when the maximum boundary of a time series lasts for a set length of time (years). The maximum boundary is computed as the daily precipitation's maximum moving window value. Thus, based on the length of years: Level 0: no truncation (a constant maximum boundary lasts at least 2 years). Level 1: a constant maximum boundary lasts longer equal (or above) 3 years but less than 5 years Level 2: a constant maximum boundary lasts more than 5 years. Argument ths present the above default thresholds for the levels definition.

References

Hunziker, S., Gubler, S., Calle, J., Moreno, I., Andrade, M., Velarde, F., ... & Brönnimann, S. (2017). Identifying, attributing, and overcoming common data quality issues of manned station observations. https://doi.org/10.1002/joc.5037

Huerta, A., Serrano-Notivoli, R., & Brönnimann, S. (2024). SC-PREC4SA: A serially complete daily precipitation dataset for South America. https://doi.org/10.31223/X57D8R

Examples

## Not run: 
set.seed(123)

# creating fake daily precipitation data
dates_t <- seq(as.Date("1980-01-01"), as.Date("2015-12-31"), by = "day")
prec <- round(rnorm(length(dates_t), mean = 1.2, sd = 50), 1)
prec[prec<0] <- 0
prec[prec<3 & prec>2] <- 0
xts_obj <- xts::xts(prec, dates_t)
names(xts_obj) <- "prec"

# Comparison with visual inspection
eqc_Plot(xts_obj)
eqc_Truncation(xts_obj)

# it also work if there is some empty data (but not if all is NA)
xts_obj["1990/2010"] <- NA
eqc_Plot(xts_obj)
eqc_Truncation(xts_obj)


## End(Not run)

Automatic enhanced quality control for daily precipitation time series

Description

The function determine the level (0, 1 or 2) of the enhanced quality control tests

Usage

eqc_Ts(
  prec,
  sts,
  lmn_yday = 365 * 80/100,
  ths_trc = list(lv0 = c(n_years = 2), lv1to2 = c(n_years_l = 3, n_years_h = 5)),
  ths_sgs = list(lv0 = c(percent = 0), lv1to2 = c(percent = 20)),
  ths_wcc = list(lv0 = c(n_days = 0), lv1to2 = c(n_days = 2, percent = 10)),
  ths_prp = list(lv0 = c(percent = 70), lv1to2 = c(percent = 50)),
  ncpu = 1
)

Arguments

Details

Truncation is when heavy precipitation episodes are truncated or noticeably reduced in frequency above a given threshold. Because there is no preceding algorithm for truncation, it is defined here as when the maximum boundary of a time series lasts for a set length of time (years). The maximum boundary is computed as the daily precipitation's maximum moving window value. Thus, based on the length of years: Level 0: no truncation (a constant maximum boundary lasts at least 2 years). Level 1: a constant maximum boundary lasts longer equal (or above) 3 years but less than 5 years Level 2: a constant maximum boundary lasts more than 5 years. Small gaps can be seen as unreported precipitation events that result in a gap or a frequency reduction in values below a specific threshold. To define the small gaps, we calculated the total count of values in five precipitation ranges from 0-1, 1-2, 2-3, 3-4, and 4-5 mm (not including the values in the limits) for each year. Therefore, considering the percentage of years with zero counts: Level 0: no small gaps (0%; years with at least one value in any of the precipitation ranges). Level 1: small gaps in at least 20% of years. Level 2: small gaps in more than 20% of years Weekly cycles are characterized by the occurrence of wet days that significantly differ between the days of the week. To compute the weekly cycles, first, for each day of the week, the probability of precipitation is calculated by dividing the total number of wet days by the total counts of values. Later, the number of wet days is tested by a two-sided binomial test (95 % confidence level). Based on how many days were significant, it was defined: Level 0: no atypical weekly cycle (similar probability between the days of the week). Level 1: at least two days present an atypical probability (significant test). Level 2: more than two days present an atypical probability (significant test) or one day presents an extremely different probability (more than 10%). Precision and rounding patterns depict inconsistencies in the frequency of decimal values in the time series. As there is no absolute correct frequency of decimals, we decided to measure how similar the decimal patterns are in the time series. A decimal pattern is interpreted as the list of unique decimal values observed, sorted in descending order. In this way, the decimal pattern for each year is computed first, followed by the selection of the most dominating pattern (mode). Based on how much (in percentage) this dominating pattern represents the time series, it is defined: Level 0: coherent precision and rounding pattern (similar decimal pattern in more (or equal) than 70% of the time series). Level 1: a similar decimal pattern in less than 70% but more than (or equal) 50% of the time series. Level 2: different decimal patterns (no dominant pattern).

References

Hunziker, S., Gubler, S., Calle, J., Moreno, I., Andrade, M., Velarde, F., ... & Brönnimann, S. (2017). Identifying, attributing, and overcoming common data quality issues of manned station observations. https://doi.org/10.1002/joc.5037

Huerta, A., Serrano-Notivoli, R., & Brönnimann, S. (2024). SC-PREC4SA: A serially complete daily precipitation dataset for South America. https://doi.org/10.31223/X57D8R

Weekly cycle level of daily precipitation

Description

The function determine the level (0, 1 or 2) of the weekly cycle test

Usage

eqc_WeeklyCycle(
  xts_obj,
  ths = list(lv0 = c(n_days = 0), lv1to2 = c(n_days = 2, percent = 10))
)

Arguments

Details

Weekly cycles are characterized by the occurrence of wet days that significantly differ between the days of the week. To compute the weekly cycles, first, for each day of the week, the probability of precipitation is calculated by dividing the total number of wet days by the total counts of values. Later, the number of wet days is tested by a two-sided binomial test (95 % confidence level). Based on how many days were significant, it was defined: Level 0: no atypical weekly cycle (similar probability between the days of the week). Level 1: at least two days present an atypical probability (significant test). Level 2: more than two days present an atypical probability (significant test) or one day presents an extremely different probability (more than 10%). Argument ths present the above default thresholds for the levels definition.

References

Hunziker, S., Gubler, S., Calle, J., Moreno, I., Andrade, M., Velarde, F., ... & Brönnimann, S. (2017). Identifying, attributing, and overcoming common data quality issues of manned station observations. https://doi.org/10.1002/joc.5037

Huerta, A., Serrano-Notivoli, R., & Brönnimann, S. (2024). SC-PREC4SA: A serially complete daily precipitation dataset for South America. https://doi.org/10.31223/X57D8R

Examples

## Not run: 
set.seed(123)

# creating fake daily precipitation data
dates_t <- seq(as.Date("1980-01-01"), as.Date("2015-12-31"), by = "day")
prec <- round(rnorm(length(dates_t), mean = 1.2, sd = 6), 1)
prec[prec<0] <- 0
xts_obj <- xts::xts(prec, dates_t)
names(xts_obj) <- "prec"

# Comparison with visual inspection
eqc_Plot(xts_obj)
eqc_WeeklyCycle(xts_obj)

# it also work if there is some empty data (but not if all is NA)
xts_obj["1990/2010"] <- NA
eqc_Plot(xts_obj)
eqc_WeeklyCycle(xts_obj)


## End(Not run)

Estimating new values in original missing values data series of daily precipitation

Description

This function uses the neighboring observations to estimate new precipitation values in those days and locations where no records exist.

Usage

gapFilling(
  prec,
  sts,
  model_fun = learner_glm,
  dates,
  stmethod = NULL,
  thres = NA,
  neibs = 10,
  coords,
  crs,
  coords_as_preds = TRUE,
  window,
  ncpu = 2
)

Arguments

Details

After the gap filling, "stmethod" allows for an standardization of the predictions based on the observations. It only works for daily data. For other timescales (monthly, annual) use "stmethod=NULL". The "window" parameter is a daily-moving centered window from which data is collected for each year (i.e. a 15-day window on 16th January will take all predictions from 1st to 30th January of all years to standardize them with their corresponding observations. Only standardized prediction of 16th January is returned. Process is repeated for all days).

Examples

## Not run: 
set.seed(123)
prec <- round(matrix(rnorm(30*50, mean = 1.2, sd = 6), 30, 50), 1)
prec[prec<0] <- 0
prec <- apply(prec, 2, FUN = function(x){x[sample(length(x),5)] <- NA; x})
colnames(prec) <- paste0('sts_',1:50)
sts <- data.frame(ID = paste0('sts_',1:50), lon = rnorm(50,0,1), 
                  lat = rnorm(50,40,1), dcoast = rnorm(50,200,50))
filled <- gapFilling(prec, sts, 
                    dates = seq.Date(as.Date('2023-04-01'),
                    as.Date('2023-04-30'),by='day'), 
                    stmethod = "ratio", thres = NA, coords = c('lon','lat'),
                    coords_as_preds = TRUE, crs = 'EPSG:4326', neibs = 10, 
                    window = 11, ncpu = 2)
str(filled)
summary(filled)

## End(Not run)

Gridded dataset creation

Description

This function creates a gridded precipitation dataset from a station-based dataset.

Usage

gridPcp(
  prec,
  grid,
  dyncovars = NULL,
  sts,
  model_fun,
  dates,
  ncpu,
  thres,
  neibs,
  coords,
  crs,
  coords_as_preds,
  dir_name = NA
)

Arguments

Details

All the rasters provided in "grid" and "dyncovars" must have the same spatial characteristics (resolution, crs, extent, etc.). The function estimates precipitation based on the nearest observations ("sts" and "prec") using as covariates the predictors contained in "grid" and "dyncovars". Predictors of "grid" are used for all days while those on "dyncovars" are selected depending on the day. For instance, to model the first day the algorithm considers all rasters in "grid" and only those corresponding to the first day of each variable.

Examples

## Not run: 
# fixed covariates (elevation, latitude, longitude)
alt <- terra::rast(volcano, crs = 'EPSG:4326')
terra::ext(alt) <- c(-1,3,38,42)
lon <- terra::rast(cbind(terra::crds(alt),terra::crds(alt)[,1]),type='xyz',crs='EPSG:4326')
lat <- terra::rast(cbind(terra::crds(alt),terra::crds(alt)[,2]),type='xyz',crs='EPSG:4326')
dcoast <- terra::costDist(alt,target=min(terra::values(alt)))/1000
grid <- c(alt, lon, lat, dcoast)
names(grid) <- c('alt', 'lon', 'lat', 'dcoast')

# Dynamic covariates (Variable 1, Variable 2, Variable 3)
foo <- alt
terra::values(foo) <- runif(length(terra::values(alt)))
dyncovars1 <- rep(foo, 7)
names(dyncovars1) <- paste('dynvar1.day',1:terra::nlyr(dyncovars1), 
                      sep = "_") # not use blank space!
dyncovars2 <- dyncovars1*0.0234
names(dyncovars2) <- paste('dynvar2.day',1:terra::nlyr(dyncovars2), 
                      sep = "_") # not use blank space!
dyncovars3 <- dyncovars1*10502
names(dyncovars3) <- paste('dynvar3.day',1:terra::nlyr(dyncovars3), 
                      sep = "_") # not use blank space!
dyncovars <- terra::sds(dyncovars1, dyncovars2, dyncovars3)
 
# precipitation and stations generation
set.seed(123)
prec <- round(matrix(rnorm(7*25, mean = 1.2, sd = 4), 7, 25), 1)+1
prec[prec<0] <- 0
colnames(prec) <- paste0('sts_',1:25)
sts <- data.frame(ID = paste0('sts_',1:25), as.data.frame(terra::spatSample(grid, 25)))


gridPcp(prec = prec, 
        grid = grid, 
        sts = sts, 
        model_fun = learner_glm, 
        dates = seq.Date(as.Date('2023-04-01'),as.Date('2023-04-07'),by='day'), 
        ncpu = 4, 
        thres = NA, 
        neibs = 15,
        coords = c('lon','lat'), 
        crs = 'EPSG:4326', 
        coords_as_preds = TRUE)

r <- terra::rast(c('./pred/20230401.tif','./err/20230401.tif'))
terra::plot(r)

## End(Not run)

Homogenization of daily precipitation time series

Description

The function applies the homogenization procedure (detection and adjustment) for daily precipitation based on Huerta et al. (2024).

Usage

hmg_Ts(
  prec,
  sts,
  neibs_max = 8,
  neibs_min = 3,
  thres = 1e+06,
  cor_neibs = 0.5,
  cleaning = FALSE,
  perc_break = 7,
  wet_day = 0,
  window_c = 15,
  apply_qc = 1,
  mm_apply_qc = 0,
  ncpu = 2
)

Arguments

Details

The homogenization procedure uses an automatic algorithm for both detection and adjustment without metadata information. In addition, relative and absolute approaches are combined for situations in which relative homogenization can not be performed. The absolute test, which has a lower power of detection than the relative tests, is thus intended as a backup test for when a relative test is hardly possible. To ensure high confidence in breakpoint detection, a combination of different statistical tests and intercomparison of their results was used. Five univariate breakpoint tests were applied: Student's, Mann-Whitney, Buishand-R, Pettit, and the Standard Normal Homogeneity Test. Depending on the availability of nearby stations for a target time series: relative and absolute. For the relative approach, the algorithm searches for up to neibs_max well-correlated (> cor_neibs) nearby stations within a three radius. Later, the five tests are applied to difference series (target minus nearby) created with three different temporal aggregations. Finally, the breakpoint is set to a certain year if it is found in at least perc_break (%) of the number of difference time series that are significant (p-value < 0.05), using a tolerance of +/- 1 year. The absolute approach is used if the algorithm detects fewer than neibs_min nearby stations or none at all. In the adjustment, it was adapted to the quantile-matching technique outlined in Squintu et al. (2018). It should be mentioned that this algorithm was created for temperature data; therefore, we made some changes to be used for precipitation. Dry values (wet_day parameter) cannot be corrected, and wet values were transformed twice (square root and log) before the algorithm execution to force a normal distribution. Based on this consideration, the correction was applied in two ways: relative and absolute. For the relative approach, the adjustment factor was computed using the target and nearby time series of the detection stage. It is assumed that the data after the break is correct; thus, the correction is backward. For the absolute case, the adjustment factor was computed using the target time series. This can be seen as an application of quantile mapping as there are no nearby stations. Adjustment of daily precipitation can influence the extreme tails. Therefore, adjusted values can be set to not exceed a limit (apply_qc - difference of the root cubic) with the raw data. The apply_qc can be applied to all precipitation values or above a specific threshold: mm_apply_qc. This procedure still keeps the extreme adjustment while preventing the creation of extremely excessive values. The output of the function is a list that contains: the detection results and the adjusted time series for each station.

References

Huerta, A., Serrano-Notivoli, R., & Brönnimann, S. (2024). SC-PREC4SA: A serially complete daily precipitation dataset for South America. https://doi.org/10.31223/X57D8R

Lund, R. B., Beaulieu, C., Killick, R., Lu, Q., & Shi, X. (2023). Good practices and common pitfalls in climate time series changepoint techniques: A review. Journal of Climate, 36(23), 8041-8057.

Squintu, A. A., van der Schrier, G., Brugnara, Y., & Klein Tank, A. (2018). Homogenization of daily ECA&D temperature series. International journal of climatology, 39(3), 1243-1261.

Function of the generalized linear model (glm)

Description

The function of the generalized linear model (glm) is used to predict wet/dry (classification) days and the amount of rain on wet days (regression). This function employs the hybrid model of Serrano-Notivoli et al. (2017).

Usage

learner_glm(ref, can, covars)

Arguments

Details

The function should not be used directly, as its main purpose is to be passed as an argument in the core functions (qcPrec, gapFilling and gridPcp) of the package. The core functions can handle whatever condition that avoids the execution of this function. Therefore, there is no need to test different situations on the data (empty values, among others). However, users who want to build their functions should test first before being passed as an argument in the core functions (you can check the other models (learner) functions available in the package). This function encapsulates two types of models: classification and regression, and its output is a single numeric vector of three elements: probability of wet day, amount of wet day, and uncertainty of the amount of wet day.

References

Serrano-Notivoli, R., Beguería, S., Saz, M. Á., Longares, L. A., & de Luis, M. (2017). SPREAD: a high-resolution daily gridded precipitation dataset for Spain–an extreme events frequency and intensity overview. Earth System Science Data, 9(2), 721-738.

Examples

## Not run: 
set.seed(123)
# creating random data (three predictors)
lon = rnorm(50,0,1)
lat = rnorm(50,40,1)
dcoast = rnorm(50,200,50)
prec = rnorm(1*50, mean = 1.2, sd = 6)
prec[prec<0] <- 0

# precipitation column should be call as "val"
data_full <- data.frame(lon = lon, lat = lat, dcoast = dcoast, val = prec)

# parameters for learner_glm
ref = data_full[-1, ]
can = data_full[1, ]
covars = c("lon", "lat", "dcoast")

learner_glm(ref = ref, can = can, covars = covars)

# case when "prec" is full wet
prec = rnorm(1*50, mean = 1.2, sd = 6)
prec[prec<0] <- 1

data_full <- data.frame(lon = lon, lat = lat, dcoast = dcoast, val = prec)

# parameters for learner_glm
ref = data_full[-1, ]
can = data_full[1, ]
covars = c("lon", "lat", "dcoast")

learner_glm(ref = ref, can = can, covars = covars)

## End(Not run)

Function of the neural network model (nn)

Description

The function of the neural network model (nn) is used to predict wet/dry (classification) days and the amount of rain on wet days (regression). This function employs the hybrid model of Serrano-Notivoli et al. (2017).

Usage

learner_nn(ref, can, covars)

Arguments

Details

The function should not be used directly, as its main purpose is to be passed as an argument in the core functions (qcPrec, gapFilling and gridPcp) of the package. The core functions can handle whatever condition that avoids the execution of this function. Therefore, there is no need to test different situations on the data (empty values, among others). However, users who want to build their functions should test first before being passed as an argument in the core functions (you can check the other models (learner) functions available in the package). This function encapsulates two types of models: classification and regression, and its output is a single numeric vector of three elements: probability of wet day, amount of wet day, and uncertainty of the amount of wet day. To use this function, the neuralnet package should have been installed previously

References

Huerta, A., Serrano-Notivoli, R., & Brönnimann, S. (2024). SC-PREC4SA: A serially complete daily precipitation dataset for South America. https://doi.org/10.31223/X57D8R

Examples

## Not run: 
set.seed(123)
# creating random data (three predictors)
lon = rnorm(50,0,1)
lat = rnorm(50,40,1)
dcoast = rnorm(50,200,50)
prec = rnorm(1*50, mean = 1.2, sd = 6)
prec[prec<0] <- 0

# precipitation column should be call as "val"
data_full <- data.frame(lon = lon, lat = lat, dcoast = dcoast, val = prec)

# parameters for learner_nn
ref = data_full[-1, ]
can = data_full[1, ]
covars = c("lon", "lat", "dcoast")

learner_nn(ref = ref, can = can, covars = covars)

# case when "prec" is full wet
prec = rnorm(1*50, mean = 1.2, sd = 6)
prec[prec<0] <- 1

data_full <- data.frame(lon = lon, lat = lat, dcoast = dcoast, val = prec)

# parameters for learner_nn
ref = data_full[-1, ]
can = data_full[1, ]
covars = c("lon", "lat", "dcoast")

learner_nn(ref = ref, can = can, covars = covars)

## End(Not run)

Function of the random forest model (rf)

Description

The function of the random forest model (rf) is used to predict wet/dry (classification) days and the amount of rain on wet days (regression). This function employs the hybrid model of Huerta et al. (2024).

Usage

learner_rf(ref, can, covars)

Arguments

Details

The function should not be used directly, as its main purpose is to be passed as an argument in the core functions (qcPrec, gapFilling and gridPcp) of the package. The core functions can handle whatever condition that avoids the execution of this function. Therefore, there is no need to test different situations on the data (empty values, among others). However, users who want to build their functions should test first before being passed as an argument in the core functions (you can check the other models (learner) functions available in the package). This function encapsulates two types of models: classification and regression, and its output is a single numeric vector of three elements: probability of wet day, amount of wet day, and uncertainty of the amount of wet day. To use this function, the randomForest package should have been installed previously

References

Huerta, A., Serrano-Notivoli, R., & Brönnimann, S. (2024). SC-PREC4SA: A serially complete daily precipitation dataset for South America. https://doi.org/10.31223/X57D8R

Examples

## Not run: 
set.seed(123)
# creating random data (three predictors)
lon = rnorm(50,0,1)
lat = rnorm(50,40,1)
dcoast = rnorm(50,200,50)
prec = rnorm(1*50, mean = 1.2, sd = 6)
prec[prec<0] <- 0

# precipitation column should be call as "val"
data_full <- data.frame(lon = lon, lat = lat, dcoast = dcoast, val = prec)

# parameters for learner_rf
ref = data_full[-1, ]
can = data_full[1, ]
covars = c("lon", "lat", "dcoast")

learner_rf(ref = ref, can = can, covars = covars)

# case when "prec" is full wet
prec = rnorm(1*50, mean = 1.2, sd = 6)
prec[prec<0] <- 1

data_full <- data.frame(lon = lon, lat = lat, dcoast = dcoast, val = prec)

# parameters for learner_rf
ref = data_full[-1, ]
can = data_full[1, ]
covars = c("lon", "lat", "dcoast")

learner_rf(ref = ref, can = can, covars = covars)

## End(Not run)

Function of the support vector machine model (svm)

Description

The function of the support vector machine model (svm) is used to predict wet/dry (classification) days and the amount of rain on wet days (regression). This function employs the hybrid model of Serrano-Notivoli et al. (2017).

Usage

learner_svm(ref, can, covars)

Arguments

Details

The function should not be used directly, as its main purpose is to be passed as an argument in the core functions (qcPrec, gapFilling and gridPcp) of the package. The core functions can handle whatever condition that avoids the execution of this function. Therefore, there is no need to test different situations on the data (empty values, among others). However, users who want to build their functions should test first before being passed as an argument in the core functions (you can check the other models (learner) functions available in the package). This function encapsulates two types of models: classification and regression, and its output is a single numeric vector of three elements: probability of wet day, amount of wet day, and uncertainty of the amount of wet day. To use this function, the e1071 package should have been installed previously

References

Huerta, A., Serrano-Notivoli, R., & Brönnimann, S. (2024). SC-PREC4SA: A serially complete daily precipitation dataset for South America. https://doi.org/10.31223/X57D8R

Examples

## Not run: 
set.seed(123)
# creating random data (three predictors)
lon = rnorm(50,0,1)
lat = rnorm(50,40,1)
dcoast = rnorm(50,200,50)
prec = rnorm(1*50, mean = 1.2, sd = 6)
prec[prec<0] <- 0

# precipitation column should be call as "val"
data_full <- data.frame(lon = lon, lat = lat, dcoast = dcoast, val = prec)

# parameters for learner_svm
ref = data_full[-1, ]
can = data_full[1, ]
covars = c("lon", "lat", "dcoast")

learner_svm(ref = ref, can = can, covars = covars)

# case when "prec" is full wet
prec = rnorm(1*50, mean = 1.2, sd = 6)
prec[prec<0] <- 1

data_full <- data.frame(lon = lon, lat = lat, dcoast = dcoast, val = prec)

# parameters for learner_svm
ref = data_full[-1, ]
can = data_full[1, ]
covars = c("lon", "lat", "dcoast")

learner_svm(ref = ref, can = can, covars = covars)

## End(Not run)

Function of the eXtreme Gradient Boosting model (xgboost)

Description

The function of the eXtreme Gradient Boosting model (xgboost) is used to predict wet/dry (classification) days and the amount of rain on wet days (regression). This function employs the hybrid model of Huerta et al. (2024).

Usage

learner_xgboost(ref, can, covars)

Arguments

Details

The function should not be used directly, as its main purpose is to be passed as an argument in the core functions (qcPrec, gapFilling and gridPcp) of the package. The core functions can handle whatever condition that avoids the execution of this function. Therefore, there is no need to test different situations on the data (empty values, among others). However, users who want to build their functions should test first before being passed as an argument in the core functions (you can check the other models (learner) functions available in the package). This function encapsulates two types of models: classification and regression, and its output is a single numeric vector of three elements: probability of wet day, amount of wet day, and uncertainty of the amount of wet day. To use this function, the xgboost package should have been installed previously

References

Huerta, A., Serrano-Notivoli, R., & Brönnimann, S. (2024). SC-PREC4SA: A serially complete daily precipitation dataset for South America. https://doi.org/10.31223/X57D8R

Examples

## Not run: 
set.seed(123)
# creating random data (three predictors)
lon = rnorm(50,0,1)
lat = rnorm(50,40,1)
dcoast = rnorm(50,200,50)
prec = rnorm(1*50, mean = 1.2, sd = 6)
prec[prec<0] <- 0

# precipitation column should be call as "val"
data_full <- data.frame(lon = lon, lat = lat, dcoast = dcoast, val = prec)

# parameters for learner_xgboost
ref = data_full[-1, ]
can = data_full[1, ]
covars = c("lon", "lat", "dcoast")

learner_xgboost(ref = ref, can = can, covars = covars)

# case when "prec" is full wet
prec = rnorm(1*50, mean = 1.2, sd = 6)
prec[prec<0] <- 1

data_full <- data.frame(lon = lon, lat = lat, dcoast = dcoast, val = prec)

# parameters for learner_xgboost
ref = data_full[-1, ]
can = data_full[1, ]
covars = c("lon", "lat", "dcoast")

learner_xgboost(ref = ref, can = can, covars = covars)

## End(Not run)

Quality Control of daily precipitation observations

Description

This function apply several threshold-based criteria to filter original observations of daily precipitation.

Usage

qcPrec(
  prec,
  sts,
  model_fun = learner_glm,
  crs,
  coords,
  coords_as_preds = TRUE,
  neibs = 10,
  thres = NA,
  qc = "all",
  qc3 = 10,
  qc4 = c(0.99, 5),
  qc5 = c(0.01, 0.1, 5),
  ncpu = 1
)

Arguments

Details

Parameter "sts" must have an "ID" field containing unique identifiers of the stations.

"qc" can be "all" (all criteria are applied) or a vector of strings (e.g.: c("1","2","4")) indicating the QC criteria to apply to observations: "1" (suspect value): obs==0 & all(neibs>0); "2" (suspect zero): obs>0 & all(neibs==0); "3" (suspect outlier): obs is "qc3" times higher or lower than the estimate; "4" (suspect wet): obs==0 & wet probability > "qc4[1]" & estimate > "qc4[2]"; "5" (suspect dry): obs>"qc5[3]" & dry probability < "qc5[1]" & estimate < "qc5[2]"

Examples

## Not run: 
set.seed(123)
prec <- round(matrix(rnorm(30*50, mean = 1.2, sd = 6), 30, 50), 1)
prec[prec<0] <- 0
colnames(prec) <- paste0('sts_',1:50) 
sts <- data.frame(ID = paste0('sts_',1:50), lon = rnorm(50,0,1), 
                  lat = rnorm(50,40,1), dcoast = rnorm(50,200,50))
qcdata <- qcPrec(prec, sts, crs = 'EPSG:4326', coords = c('lon','lat'),
                 coords_as_preds = TRUE, neibs = 10, thres = NA,
                 qc = 'all', qc3 = 10, qc4 = c(0.99, 5), qc5 = c(0.01, 0.1, 5),
                 ncpu=2)
str(qcdata)

## End(Not run)