Seasonal Forecast of Wet Season Calendar (WSC) from Tercile Rainfall Probabilities using Quantile Bin Resampling (QBR)

Generates seasonal forecasts for Wet Season Calendar (WSC) variables (onset or cessation) using tercile seasonal rainfall probabilities

Usage

seasFcstQBR(
  hisYearStart,
  hisYearEnd,
  rainTerc,
  seasRain,
  hisWSCvar,
  fcstVarName,
  tercileMethod
)

Arguments

hisYearStart: Starting year (YYYY) for historical resampling.
hisYearEnd: Ending year (YYYY) for historical resampling.
rainTerc: A data frame with tercile probabilities for rainfall. Columns should be named "T1" (below normal), "T2" (normal), and "T3" (above normal).
seasRain: A data frame containing seasonal rainfall data with columns
hisWSCvar: A data frame containing historical WSC simulations. This can be the output from the calcSeasCal function or a similar data frame with required columns.
fcstVarName: A character string indicating the WSC variable to forecast ("Onset" or "Cessation").
tercileMethod: options are "quantiles" or "fixedValues"

Value

A data frame containing the tercile probabilities for the WSC variable ("BelowNormal", "Normal", and "AboveNormal").

Details

Uses QBR (Quantile Bin Resampling) to produce forecasts for onset or cessation of the rainy season. It first categorizes historical WSC simulations based on seasonal rainfall terciles and then resamples based on given rainfall probabilities to generate ensemble forecasts.

References

MacLeod D, Quichimbo EA, Michaelides K, Asfaw DT, Rosolem R, Cuthbert MO, et al. (2023) Translating seasonal climate forecasts into water balance forecasts for decision making. PLOS Clim 2(3): e0000138. https://doi.org/10.1371/journal.pclm.0000138
van den Dool HM. A New Look at Weather Forecasting through Analogues. Monthly Weather Review. 1989; 117(10):2230–2247. https://doi.org/10.1175/1520-0493(1989)117%3C2230:ANLAWF%3E2.0.CO;2

Examples

# \donttest{
library(dplyr)
#> 
#> Attaching package: ‘dplyr’
#> The following objects are masked from ‘package:stats’:
#> 
#>     filter, lag
#> The following objects are masked from ‘package:base’:
#> 
#>     intersect, setdiff, setequal, union

## Load example data:
data(AgroClimateData)

## Estimate daily PET:
PET <- calcEto(AgroClimateData, method = "PM", Zh = 10)
#> Adjusted 0 instances where Tmin was equal to or greater than Tmax.
#> Penman-Monteith FAO56 Reference Crop ET
#> Evaporative surface: FAO-56 hypothetical short grass, albedo = 0.23 ; surface resistance = 70 sm^-1; crop height = 0.12 m; roughness height = 0.02 m
#> Timestep: daily
#> Units: mm
#> Time duration: 1982-01-01 to 2022-12-31

## Add the estimated PET 'ET.Daily' to a new column in AgroClimateData:
AgroClimateData$Eto <- PET$ET.Daily

## Estimate daily water balance for the soil having 100mm of WHC:
watBal.list <- calcWatBal(data = AgroClimateData, soilWHC = 100)
watBal <- watBal.list$data

## seasonal calendar is estimated for the onset window ranges from
## 01 September to 31 January having a soil with 100mm of WHC:

soilWHC <- 100
onsetWind.start <- "09-01"
onsetWind.end <- "01-31"
cessaWind.end <- "06-30"

seasCal.dF <- calcSeasCal(
  data = watBal, onsetWind.start, onsetWind.end,
  cessaWind.end, soilWHC
)

## Tercile Rainfall Probabilities of seasonal Forecast for OND, 2023:
rainTerc <- data.frame(T1 = 0.15, T2 = 0.10, T3 = 0.75)

## Summarize rainfall data for October to December:
seasRain <- AgroClimateData %>%
  filter(Month %in% c(10, 11, 12)) %>%
  group_by(Year) %>%
  summarize(sRain = sum(Rain))

## Start of the historical resampling year
hisYearStart <- 1991

## End of the historical resampling year
hisYearEnd <- 2022

## Historical WSC Simulations:
hisWSCvar <- seasCal.dF

## WSC variable to forecast:
fcstVarName <- "Onset"
tercileMethod <- "quantiles"

SeasFcst.dF <- seasFcstQBR(
  hisYearStart, hisYearEnd, rainTerc,
  seasRain, hisWSCvar, fcstVarName,
  tercileMethod
)
# }