Steps for fitting an mcgf object

Introduction

The mcgf package contains useful functions to simulate Markov chain Gaussian fields (MCGF) and regime-switching Markov chain Gaussian fields (RS-MCGF) with covariance structures of the Gneiting class (Gneiting 2002). It also provides useful tools to estimate the parameters by weighted least squares (WLS) and maximum likelihood estimation (MLE). The mcgf function can be used to fit covariance models and obtain Kriging forecasts. A typical workflow for fitting a non-regime-switching mcgf is given below.

  1. Create an mcgf object by providing a dataset and the corresponding locations/distances
  2. Calculate auto- and cross-correlations.
  3. Fit the base covariance model which is a fully symmetric model.
  4. Fit the Lagrangian model to account for asymmetry, if necessary.
  5. Obtain Kriging forecasts.
  6. Obtain Kriging forecasts for new locations given their coordinates.

We will demonstrate the use of mcgf by an example below.

Irish Wind

The Ireland wind data contains daily wind speeds for 1961-1978 at 11 synoptic meteorological stations in the Republic of Ireland. It is available in the gstat package and is also imported to mcgf. To view the details on this dataset, run help(wind). The object wind contains the wind speeds data as well as the locations of the stations.

library(mcgf)
data(wind)
head(wind$data)
#>         time      VAL      BEL       CLA      SHA      ROC      BIR      MUL
#> 1 1961-01-01 7.696089 9.517222 5.2730556 7.181644 7.737244 5.077567 5.571433
#> 2 1961-01-02 8.683822 9.023356 5.1650222 6.492289 7.567478 3.945789 5.036411
#> 3 1961-01-03 8.683822 6.559167 4.1361333 5.746344 9.517222 3.174122 4.372778
#> 4 1961-01-04 3.410767 2.808867 0.9208556 2.335578 5.442822 1.481600 2.999211
#> 5 1961-01-05 6.816389 6.646622 3.3644667 5.509700 6.857544 4.223589 5.617733
#> 6 1961-01-06 4.177289 4.177289 2.2738444 2.762567 6.795811 2.315000 3.688567
#>        MAL      KIL      CLO      DUB
#> 1 7.737244 4.779189 6.471711 7.032456
#> 2 7.114767 3.343889 4.974678 5.916111
#> 3 6.538589 5.211322 3.945789 5.787500
#> 4 5.597156 2.356156 3.024933 4.439656
#> 5 6.085878 3.174122 5.319356 6.132178
#> 6 6.775233 3.431344 3.858333 5.489122
wind$locations
#>            lon      lat
#> VAL -10.250000 51.93333
#> BEL -10.000000 54.23333
#> CLA  -8.983333 53.71667
#> SHA  -8.916667 52.70000
#> ROC  -8.250000 51.80000
#> BIR  -7.883333 53.08333
#> MUL  -7.366667 53.53333
#> MAL  -7.333333 55.36667
#> KIL  -7.266667 52.66667
#> CLO  -7.233333 54.18333
#> DUB  -6.250000 53.43333

Data De-treding

To fit covariance models on wind, the first step is to load the data set and de-trend the data to have a mean-zero time series. We will follow the data de-trending procedure in Gneiting, Genton, and Guttorp (2006).

  1. We start with removing the leap day and take the square-root transformation. To do this we need to use the lubridate package.
# install.packages("lubridate")
library(lubridate)
ind_leap <- month(wind$data$time) == 2 & day(wind$data$time) == 29
wind_de <- wind$data[!ind_leap, ]
wind_de[, -1] <- sqrt(wind_de[, -1])
  1. Next, we split the data into training and test datasets, where training contains data in 1961-1970 and test spans 1971-1978.
is_train <- year(wind_de$time) <= 1970
wind_train <- wind_de[is_train, ]
wind_test <- wind_de[!is_train, ]
  1. We will estimate and remove the annual trend for the training dataset. The annual trend is the daily averages across the training years. We will use the dplyr package for this task.
# install.packages("dplyr")
library(dplyr)
# Estimate annual trend
avg_train <- tibble(
    month = month(wind_train$time),
    day = day(wind_train$time),
    ws = rowMeans(wind_train[, -1])
)
trend_train <- avg_train %>%
    group_by(month, day) %>%
    summarise(trend = mean(ws))

# Subtract annual trend
trend_train2 <- left_join(avg_train, trend_train, by = c("month", "day"))$trend
wind_train[, -1] <- wind_train[, -1] - trend_train2
  1. To obtain a mean-zero time series, we will subtract the station-wise mean for each station.
# Subtract station-wise mean
mean_train <- colMeans(wind_train[, -1])
wind_train[, -1] <- sweep(wind_train[, -1], 2, mean_train)
wind_trend <- list(
    annual = as.data.frame(trend_train),
    mean = mean_train
)
  1. Finally, we will subtract the annual trend and station-wise mean from the test dataset.
avg_test <- tibble(
    month = month(wind_test$time),
    day = day(wind_test$time)
)
trend_train3 <-
    left_join(avg_test, trend_train, by = c("month", "day"))$trend
wind_test[, -1] <- wind_test[, -1] - trend_train3
wind_test[, -1] <- sweep(wind_test[, -1], 2, mean_train)

Fitting Covariance Models

We will first fit pure spatial and temporal models, then the fully symmetric model, and finally the general stationary model. First, we will create an mcgf object and calculate auto- and cross- correlations.

wind_mcgf <- mcgf(wind_train[, -1], locations = wind$locations, longlat = TRUE)
#> `time` is not provided, assuming rows are equally spaced temporally.
wind_mcgf <- add_acfs(wind_mcgf, lag_max = 3)
wind_mcgf <- add_ccfs(wind_mcgf, lag_max = 3)

Here acfs actually refers to the mean auto-correlations across the stations for each time lag. To view the calculated acfs, we can run:

acfs(wind_mcgf)
#>      lag0      lag1      lag2      lag3 
#> 1.0000000 0.5072739 0.2382621 0.1643998

Similarly, we can view the ccfs by:

ccfs(wind_mcgf)
#> , , lag0
#> 
#>           VAL       BEL       CLA       SHA       ROC       BIR       MUL
#> VAL 1.0000000 0.7076428 0.7588797 0.8299603 0.7962082 0.7719814 0.6688609
#> BEL 0.7076428 1.0000000 0.8471918 0.7432194 0.5795413 0.7650007 0.7371354
#> CLA 0.7588797 0.8471918 1.0000000 0.8520358 0.6994680 0.8777462 0.8410719
#> SHA 0.8299603 0.7432194 0.8520358 1.0000000 0.7960579 0.8865147 0.8239009
#> ROC 0.7962082 0.5795413 0.6994680 0.7960579 1.0000000 0.7592625 0.7305102
#> BIR 0.7719814 0.7650007 0.8777462 0.8865147 0.7592625 1.0000000 0.8841425
#> MUL 0.6688609 0.7371354 0.8410719 0.8239009 0.7305102 0.8841425 1.0000000
#> MAL 0.5153690 0.7087588 0.6818106 0.6121929 0.5357199 0.6411302 0.7375539
#> KIL 0.7333467 0.6487696 0.7772730 0.8236153 0.8338255 0.8406781 0.8414494
#> CLO 0.6718720 0.7697673 0.8511102 0.7759129 0.6876764 0.8436560 0.8699373
#> DUB 0.5967529 0.6498491 0.7420593 0.7361857 0.6862624 0.7812147 0.8609285
#>           MAL       KIL       CLO       DUB
#> VAL 0.5153690 0.7333467 0.6718720 0.5967529
#> BEL 0.7087588 0.6487696 0.7697673 0.6498491
#> CLA 0.6818106 0.7772730 0.8511102 0.7420593
#> SHA 0.6121929 0.8236153 0.7759129 0.7361857
#> ROC 0.5357199 0.8338255 0.6876764 0.6862624
#> BIR 0.6411302 0.8406781 0.8436560 0.7812147
#> MUL 0.7375539 0.8414494 0.8699373 0.8609285
#> MAL 1.0000000 0.5892218 0.7568132 0.7145366
#> KIL 0.5892218 1.0000000 0.7934818 0.7770512
#> CLO 0.7568132 0.7934818 1.0000000 0.8161496
#> DUB 0.7145366 0.7770512 0.8161496 1.0000000
#> 
#> , , lag1
#> 
#>           VAL       BEL       CLA       SHA       ROC       BIR       MUL
#> VAL 0.4739681 0.4078097 0.3734029 0.4028977 0.3406302 0.3871317 0.3171416
#> BEL 0.3843919 0.5310138 0.4208376 0.3899411 0.2819817 0.4000168 0.3695143
#> CLA 0.4164508 0.4814516 0.4993218 0.4450647 0.3244728 0.4521809 0.4020938
#> SHA 0.4719522 0.4646048 0.4689832 0.5137855 0.3883962 0.4677375 0.4100365
#> ROC 0.4706594 0.3793417 0.3852120 0.4285917 0.4442973 0.4113073 0.3590478
#> BIR 0.4725125 0.4964189 0.4986726 0.4927316 0.3919890 0.5409202 0.4557624
#> MUL 0.4438395 0.5082231 0.5039360 0.4898975 0.4056061 0.5097809 0.5228395
#> MAL 0.3550724 0.4616666 0.4118390 0.3918386 0.3360198 0.3792364 0.4182625
#> KIL 0.4867586 0.4607927 0.4866086 0.4964729 0.4299541 0.4930438 0.4430179
#> CLO 0.4390459 0.5117641 0.5057839 0.4686427 0.3853159 0.4904330 0.4549374
#> DUB 0.4333070 0.4698080 0.4829445 0.4819004 0.4349093 0.4860929 0.4858079
#>           MAL       KIL       CLO       DUB
#> VAL 0.2712570 0.2917876 0.3248601 0.2900024
#> BEL 0.3554572 0.3012227 0.3806926 0.3179568
#> CLA 0.3289562 0.3458138 0.4112028 0.3569920
#> SHA 0.3203410 0.3834180 0.4088611 0.3736715
#> ROC 0.2950458 0.3540329 0.3479210 0.3484134
#> BIR 0.3408605 0.3973215 0.4505508 0.3984836
#> MUL 0.4151149 0.4232129 0.4664284 0.4434980
#> MAL 0.5296408 0.3105146 0.4036000 0.4002747
#> KIL 0.3227371 0.4642266 0.4387080 0.3981810
#> CLO 0.4032414 0.4076724 0.5190745 0.4289406
#> DUB 0.4428008 0.4300411 0.4765407 0.5409250
#> 
#> , , lag2
#> 
#>           VAL       BEL       CLA       SHA       ROC       BIR       MUL
#> VAL 0.2011319 0.1973159 0.1630851 0.1871285 0.1489858 0.1943355 0.1492425
#> BEL 0.1713197 0.2588323 0.1891480 0.1936055 0.1240906 0.2020276 0.1806373
#> CLA 0.1585236 0.2132206 0.2247423 0.2000905 0.1184259 0.2115274 0.1763905
#> SHA 0.1938118 0.2298367 0.2157028 0.2391773 0.1589136 0.2265611 0.1844087
#> ROC 0.1765714 0.1716184 0.1421018 0.1764547 0.1696931 0.1742268 0.1451217
#> BIR 0.2020194 0.2401048 0.2278565 0.2262318 0.1628422 0.2765692 0.2136481
#> MUL 0.1862142 0.2487854 0.2182151 0.2159567 0.1694341 0.2448788 0.2510983
#> MAL 0.1555252 0.2110852 0.1623937 0.1691266 0.1462008 0.1585537 0.1853651
#> KIL 0.1743715 0.2068322 0.1970328 0.2091868 0.1495429 0.2192173 0.1882474
#> CLO 0.1835787 0.2349159 0.2190671 0.2138391 0.1511675 0.2355708 0.2042911
#> DUB 0.1856084 0.2176080 0.2022041 0.2148639 0.1950841 0.2234817 0.2150465
#>           MAL       KIL       CLO       DUB
#> VAL 0.1427551 0.1382869 0.1666310 0.1535865
#> BEL 0.1754907 0.1388237 0.1941650 0.1629718
#> CLA 0.1419690 0.1537041 0.1979937 0.1674888
#> SHA 0.1503395 0.1803397 0.2022438 0.1764521
#> ROC 0.1458883 0.1373046 0.1488279 0.1654739
#> BIR 0.1438303 0.1844782 0.2197392 0.1920556
#> MUL 0.1950942 0.1905402 0.2183714 0.2084112
#> MAL 0.2636079 0.1107242 0.1729375 0.1921164
#> KIL 0.1334084 0.1931797 0.1982144 0.1790683
#> CLO 0.1715671 0.1793731 0.2624348 0.2135477
#> DUB 0.2180694 0.1899728 0.2270066 0.2804169
#> 
#> , , lag3
#> 
#>            VAL       BEL        CLA        SHA        ROC        BIR        MUL
#> VAL 0.15186704 0.1424697 0.12263507 0.12486035 0.11217575 0.14386635 0.09991865
#> BEL 0.12223738 0.1854410 0.13490665 0.13169403 0.08728794 0.14207057 0.12397596
#> CLA 0.10549441 0.1408343 0.16084299 0.13066846 0.07914442 0.13859706 0.10865229
#> SHA 0.12004043 0.1511578 0.14232764 0.14961877 0.09888111 0.14670946 0.11247087
#> ROC 0.11051322 0.1022188 0.08007433 0.09282562 0.10856180 0.10063755 0.07957425
#> BIR 0.13309457 0.1594322 0.15267585 0.13773090 0.10595023 0.19194636 0.13771463
#> MUL 0.11364183 0.1682815 0.14195183 0.12747872 0.10407473 0.15372629 0.16630054
#> MAL 0.09065312 0.1431289 0.10244646 0.10197080 0.09770836 0.08625211 0.11386343
#> KIL 0.09511038 0.1179299 0.11535092 0.11479319 0.07924395 0.12369722 0.10034132
#> CLO 0.12004487 0.1584521 0.15581710 0.13825387 0.10650927 0.15601087 0.13152273
#> DUB 0.10526163 0.1331954 0.11965370 0.11622550 0.11602725 0.12380225 0.11590933
#>            MAL        KIL        CLO       DUB
#> VAL 0.11038186 0.09840820 0.12881956 0.1131093
#> BEL 0.11895940 0.09539470 0.13930407 0.1065024
#> CLA 0.08566429 0.10250903 0.13807158 0.1059626
#> SHA 0.10015324 0.11889068 0.14364630 0.1110210
#> ROC 0.10700118 0.08017778 0.09975703 0.1043980
#> BIR 0.09283539 0.12122167 0.15906405 0.1304590
#> MUL 0.13033344 0.11347989 0.14872192 0.1286318
#> MAL 0.19778972 0.05899522 0.10990283 0.1253038
#> KIL 0.07627083 0.11333371 0.12609113 0.1024267
#> CLO 0.10926382 0.12014806 0.20136056 0.1459052
#> DUB 0.14141118 0.10085759 0.14688397 0.1813353

Pure Spatial Model

The pure spatial model can be fitted using the fit_base function. The results are actually obtained from the optimization function nlminb.

fit_spatial <- fit_base(
    wind_mcgf,
    model = "spatial",
    lag = 3,
    par_init = c(nugget = 0.1, c = 0.001),
    par_fixed = c(gamma = 0.5)
)
fit_spatial$fit
#> $par
#>           c      nugget 
#> 0.001326192 0.049136042 
#> 
#> $objective
#> [1] 1.742162
#> 
#> $convergence
#> [1] 0
#> 
#> $iterations
#> [1] 9
#> 
#> $evaluations
#> function gradient 
#>       23       24 
#> 
#> $message
#> [1] "relative convergence (4)"

Here we set gamma to be 0.5 and it is not estimated along with c or nugget. By default mcgf provides two optimization functions: nlminb and optim. Other optimization functions are also supported as long as their first two arguments are initial values for the parameters and a function to be minimized respectively (same as that of optim and nlminb). Also, if the argument names for upper and lower bounds are not upper or lower, we can create a simple wrapper to “change” them.

library(Rsolnp)
solnp2 <- function(pars, fun, lower, upper, ...) {
    solnp(pars, fun, LB = lower, UB = upper, ...)
}
fit_spatial2 <- fit_base(
    wind_mcgf,
    model = "spatial",
    lag = 3,
    par_init = c(nugget = 0.1, c = 0.001),
    par_fixed = c(gamma = 0.5),
    optim_fn = "other",
    other_optim_fn = "solnp2"
)
#> 
#> Iter: 1 fn: 1.7422    Pars:  0.001326 0.049164
#> Iter: 2 fn: 1.7422    Pars:  0.001326 0.049164
#> solnp--> Completed in 2 iterations
fit_spatial2$fit
#> $pars
#>           c      nugget 
#> 0.001325958 0.049163635 
#> 
#> $convergence
#> [1] 0
#> 
#> $values
#> [1] 2.539010 1.742162 1.742162
#> 
#> $lagrange
#> [1] 0
#> 
#> $hessian
#>            [,1]       [,2]
#> [1,] 58504086.2 391216.053
#> [2,]   391216.1   3319.584
#> 
#> $ineqx0
#> NULL
#> 
#> $nfuneval
#> [1] 62
#> 
#> $outer.iter
#> [1] 2
#> 
#> $elapsed
#> Time difference of 0.02971911 secs
#> 
#> $vscale
#> [1] 1 1 1

Pure Temporal Model

The pure temporal can also be fitted by fit_base:

fit_temporal <- fit_base(
    wind_mcgf,
    model = "temporal",
    lag = 3,
    par_init = c(a = 0.5, alpha = 0.5)
)
fit_temporal$fit
#> $par
#>         a     alpha 
#> 0.9774157 0.8053363 
#> 
#> $objective
#> [1] 0.0006466064
#> 
#> $convergence
#> [1] 0
#> 
#> $iterations
#> [1] 12
#> 
#> $evaluations
#> function gradient 
#>       15       31 
#> 
#> $message
#> [1] "both X-convergence and relative convergence (5)"

Before fitting the fully symmetric model, we need to store the fitted spatial and temporal models to wind_mcgf using add_base:

wind_mcgf <- add_base(wind_mcgf,
    fit_s = fit_spatial,
    fit_t = fit_temporal,
    sep = T
)

Separable Model

We can also fit the pure spatial and temporal models all at once by fitting a separable model:

fit_sep <- fit_base(
    wind_mcgf,
    model = "sep",
    lag = 3,
    par_init = c(nugget = 0.1, c = 0.001, a = 0.5, alpha = 0.5),
    par_fixed = c(gamma = 0.5)
)
fit_sep$fit
#> $par
#>           c      nugget           a       alpha 
#> 0.001309486 0.050562930 0.877171418 0.869117992 
#> 
#> $objective
#> [1] 2.838695
#> 
#> $convergence
#> [1] 0
#> 
#> $iterations
#> [1] 25
#> 
#> $evaluations
#> function gradient 
#>       54      121 
#> 
#> $message
#> [1] "relative convergence (4)"

Once can also store this model to wind_mcgf, but to be consistent with Gneiting, Genton, and Guttorp (2006), we will just use the estimates from the pure spatial and temporal models to estimate the interaction parameter beta.

Fully Symmetric Model

When holding other parameters constant, the parameter beta can be estimated by:

par_sep <- c(fit_spatial$fit$par, fit_temporal$fit$par, gamma = 0.5)
fit_fs <-
    fit_base(
        wind_mcgf,
        model = "fs",
        lag = 3,
        par_init = c(beta = 0),
        par_fixed = par_sep
    )
fit_fs$fit
#> $par
#>      beta 
#> 0.6232458 
#> 
#> $objective
#> [1] 2.858384
#> 
#> $convergence
#> [1] 0
#> 
#> $iterations
#> [1] 5
#> 
#> $evaluations
#> function gradient 
#>        6        7 
#> 
#> $message
#> [1] "relative convergence (4)"

At this stage, we have fitted the base model, and we will store the fully symmetric model as the base model and print the base model:

wind_mcgf <- add_base(wind_mcgf, fit_base = fit_fs, old = TRUE)
model(wind_mcgf, model = "base", old = TRUE)
#> ----------------------------------------
#>                  Model
#> ----------------------------------------
#> - Time lag: 3 
#> - Scale of time lag: 1 
#> - Forecast horizon: 1 
#> ----------------------------------------
#>             Old - not in use
#> ----------------------------------------
#> - Base-old model: sep 
#> - Parameters:
#>           c      nugget       gamma           a       alpha 
#> 0.001326192 0.049136042 0.500000000 0.977415717 0.805336266 
#> 
#> - Fixed parameters:
#> gamma 
#>   0.5 
#> 
#> - Parameter estimation method: wls wls 
#> 
#> - Optimization function: nlminb nlminb 
#> ----------------------------------------
#>                  Base
#> ----------------------------------------
#> - Base model: fs 
#> - Parameters:
#>        beta           c      nugget           a       alpha       gamma 
#> 0.623245823 0.001326192 0.049136042 0.977415717 0.805336266 0.500000000 
#> 
#> - Fixed parameters:
#>           c      nugget           a       alpha       gamma 
#> 0.001326192 0.049136042 0.977415717 0.805336266 0.500000000 
#> 
#> - Parameter estimation method: wls 
#> 
#> - Optimization function: nlminb

The old = TRUE in add_base keeps the fitted pure spatial and temporal models for records, and they are not used for any further steps. It is recommended to keep the old model not only for reproducibility, but to keep a history of fitted models.

Lagrangian Model

We will fit a Lagrangian correlation function to model the westerly wind:

fit_lagr <- fit_lagr(wind_mcgf,
    model = "lagr_tri",
    par_init = c(v1 = 200, lambda = 0.1),
    par_fixed = c(v2 = 0, k = 2)
)
fit_lagr$fit
#> $par
#>       lambda           v1 
#>   0.05595016 212.29485855 
#> 
#> $objective
#> [1] 2.662327
#> 
#> $convergence
#> [1] 0
#> 
#> $iterations
#> [1] 18
#> 
#> $evaluations
#> function gradient 
#>       21       54 
#> 
#> $message
#> [1] "relative convergence (4)"

Finally we will store this model to wind_mcgf using add_lagr and then print the final model:

wind_mcgf <- add_lagr(wind_mcgf, fit_lagr = fit_lagr)
model(wind_mcgf, old = TRUE)
#> ----------------------------------------
#>                  Model
#> ----------------------------------------
#> - Time lag: 3 
#> - Scale of time lag: 1 
#> - Forecast horizon: 1 
#> ----------------------------------------
#>             Old - not in use
#> ----------------------------------------
#> - Base-old model: sep 
#> - Parameters:
#>           c      nugget       gamma           a       alpha 
#> 0.001326192 0.049136042 0.500000000 0.977415717 0.805336266 
#> 
#> - Fixed parameters:
#> gamma 
#>   0.5 
#> 
#> - Parameter estimation method: wls wls 
#> 
#> - Optimization function: nlminb nlminb 
#> ----------------------------------------
#>                  Base
#> ----------------------------------------
#> - Base model: fs 
#> - Parameters:
#>        beta           c      nugget           a       alpha       gamma 
#> 0.623245823 0.001326192 0.049136042 0.977415717 0.805336266 0.500000000 
#> 
#> - Fixed parameters:
#>           c      nugget           a       alpha       gamma 
#> 0.001326192 0.049136042 0.977415717 0.805336266 0.500000000 
#> 
#> - Parameter estimation method: wls 
#> 
#> - Optimization function: nlminb 
#> ----------------------------------------
#>               Lagrangian
#> ----------------------------------------
#> - Lagrangian model: lagr_tri 
#> - Parameters:
#>       lambda           v1           v2            k 
#>   0.05595016 212.29485855   0.00000000   2.00000000 
#> 
#> - Fixed parameters:
#> v2  k 
#>  0  2 
#> 
#> - Parameter estimation method: wls 
#> 
#> - Optimization function: nlminb

Simple Kriging

For the test dataset, we can obtain the simple kriging forecasts and intervals for the empirical model, base model, and general stationary model using the krige function:

krige_emp <- krige(
    x = wind_mcgf,
    newdata = wind_test[, -1],
    model = "empirical",
    interval = TRUE
)
krige_base <- krige(
    x = wind_mcgf,
    newdata = wind_test[, -1],
    model = "base",
    interval = TRUE
)
krige_stat <- krige(
    x = wind_mcgf,
    newdata = wind_test[, -1],
    model = "all",
    interval = TRUE
)

Next, we can compute the root mean square error (RMSE), mean absolute error (MAE), and the realized percentage of observations falling outside the 95% PI (POPI) for these models on the test dataset.

RMSE

# RMSE
rmse_emp <- sqrt(colMeans((wind_test[, -1] - krige_emp$fit)^2, na.rm = T))
rmse_base <- sqrt(colMeans((wind_test[, -1] - krige_base$fit)^2, na.rm = T))
rmse_stat <- sqrt(colMeans((wind_test[, -1] - krige_stat$fit)^2, na.rm = T))
rmse <- c(
    "Empirical" = mean(rmse_emp),
    "Base" = mean(rmse_base),
    "STAT" = mean(rmse_stat)
)
rmse
#> Empirical      Base      STAT 
#> 0.4778542 0.4890462 0.4852207

MAE

mae_emp <- colMeans(abs(wind_test[, -1] - krige_emp$fit), na.rm = T)
mae_base <- colMeans(abs(wind_test[, -1] - krige_base$fit), na.rm = T)
mae_stat <- colMeans(abs(wind_test[, -1] - krige_stat$fit), na.rm = T)
mae <- c(
    "Empirical" = mean(mae_emp),
    "Base" = mean(mae_base),
    "STAT" = mean(mae_stat)
)
mae
#> Empirical      Base      STAT 
#> 0.3776108 0.3890038 0.3851617

POPI

# POPI
popi_emp <- colMeans(
    wind_test[, -1] < krige_emp$lower | wind_test[, -1] > krige_emp$upper,
    na.rm = T
)
popi_base <- colMeans(
    wind_test[, -1] < krige_base$lower | wind_test[, -1] > krige_base$upper,
    na.rm = T
)
popi_stat <- colMeans(
    wind_test[, -1] < krige_stat$lower | wind_test[, -1] > krige_stat$upper,
    na.rm = T
)
popi <- c(
    "Empirical" = mean(popi_emp),
    "Base" = mean(popi_base),
    "STAT" = mean(popi_stat)
)
popi
#>  Empirical       Base       STAT 
#> 0.06962321 0.06473026 0.06326550

Simple Kriging for new locations

We provide functionalists for computing simple Kriging forecasts for new locations. The associated function is krige_new, and users can either supply the coordinates for the new locations or the distance matrices for all locations. Hypotheticaly, suppose a new location is at (\(9^\circ\)W, \(52^\circ\)N), then its kriging forecasts along with forecasts for the rest of the stations for the general stationary model can be obtained as follows.

krige_stat_new <- krige_new(
    x = wind_mcgf,
    newdata = wind_test[, -1],
    locations_new = c(-9, 52),
    model = "all",
    interval = TRUE
)
head(krige_stat_new$fit)
#>             VAL        BEL        CLA        SHA        ROC        BIR
#> 3653         NA         NA         NA         NA         NA         NA
#> 3654         NA         NA         NA         NA         NA         NA
#> 3655         NA         NA         NA         NA         NA         NA
#> 3656 -0.1447040 -0.3018214 -0.5758966 -0.4693314 -0.1046231 -0.5499500
#> 3657  0.5307650  0.1531176  0.2150980  0.3479844  0.4910715  0.3157901
#> 3658  0.1632995  0.1648983  0.1734723  0.1751080  0.1567910  0.3091076
#>             MUL        MAL        KIL        CLO        DUB      New_1
#> 3653         NA         NA         NA         NA         NA         NA
#> 3654         NA         NA         NA         NA         NA         NA
#> 3655         NA         NA         NA         NA         NA         NA
#> 3656 -0.5196208 -0.2799845 -0.4627520 -0.7397430 -0.4240240 -0.4152704
#> 3657  0.2576929  0.1404444  0.1013024  0.1157068  0.2114298  0.8410688
#> 3658  0.3274626  0.3504309  0.1426461  0.1006568  0.4521153  0.2792891

Below is a time-series plot for the first 100 forecasts for New_1:

plot.ts(krige_stat_new$fit[1:100, 12], ylab = "Wind Speed for New_1")

References

Gneiting, Tilmann. 2002. “Nonseparable, Stationary Covariance Functions for Space–Time Data.” Journal of the American Statistical Association 97 (458): 590–600. https://doi.org/10.1198/016214502760047113.
Gneiting, Tilmann, Marc G. Genton, and Peter Guttorp. 2006. “Geostatistical Space-Time Models, Stationarity, Separability and Full Symmetry.” Department of Statistics, University of Washington; https://stat.uw.edu/sites/default/files/files/reports/2005/tr475.pdf.