patch_daily_temps {chillR}

Patch gaps in daily weather records - updated

Description

This is the successor function of patch_daily_temperatures, which will no longer be updated.

Usage

patch_daily_temps(
  weather,
  patch_weather,
  vars = c("Tmin", "Tmax"),
  max_mean_bias = NA,
  max_stdev_bias = NA,
  time_interval = "month"
)

Arguments

weather

chillR-compatible weather record to be patched

patch_weather

list of chillR-compatible weather records to be used for patching holes in weather. They are used sequentially, until all have been used or until there are no holes left.

vars

vector of column names to be considered in patching. Defaults to c("Tmin","Tmax"), the most common variables in chillR applications.

max_mean_bias

maximum mean bias of auxiliary data compared to the original dataset (applied to all variables in vars). If this threshold is exceeded, the respective variable from that particular dataset will not be used. Defaults to NA, meaning no records are excluded.

max_stdev_bias

maximum standard deviation of the bias in the auxiliary data compared to the original dataset (applied to all variables in vars). If this threshold is exceeded, the respective variable from that particular dataset will not be used. Defaults to NA, meaning no records are excluded.

time_interval

time interval for which mean bias and standard deviation of the bias are to be evaluated. This defaults to "month", which means that the function looks for overlapping days between weather and patch_weather for each calendar month. Bias correction is then also done on a monthly basis. 'time_interval' can also assume other values, such as 'week' or '2 weeks'.

Details

The patch_daily_temps function uses auxiliary data sources to fill gaps in daily weather data. It can accommodate multiple sources of auxiliary information, which are used in the user-specified sequence. There have to be some overlapping records for this to work, because without bias correction, this procedure could produce erroneous records. Bias correction is done by computing the mean difference between main and auxiliary data for each variable and adjusting for it in filling the gaps. You can specify a maximum mean bias and a maximum standard deviation of the bias to exclude unsuitable records that aren't similar enough to the original data. When patching records, the function breaks the calendar year down into smaller intervals that can be specified with the 'time_interval' parameter (this was not possible in [chillR::patch_daily_temperatures], but is recommended for accurate results).

Value

list of two elements: weather (the patched weather record, with additional columns specifying the data source for each value) and statistics (containing data.frames for each element of patch_weather that indicate the mean bias, the number of values that were filled from this source and the number of missing records that remained after exhausting this auxiliary data source.)

Author(s)

Eike Luedeling

Examples

gap_weather<-KA_weather[1:100,]
gap_weather[c(3,4,7:15,20,22:25,27:28,35:45,55,67,70:75,80:88,95:97),"Tmin"]<-NA
gap_weather[c(10:25,30,36:44,50,57,65,70:80,86,91:94),"Tmax"]<-NA
p1<-KA_weather[65:95,]
p1$Tmin<-p1$Tmin-2
p2<-KA_weather[c(15:40,60:80),]
p2$Tmax<-p2$Tmax+3
p3<-KA_weather[12:35,]
p3$Tmax<-p3$Tmax-2
p4<-KA_weather
p4$Tmax<-p4$Tmax+0.5
patch_weather<-list(stat1=p1,st2=p2,home=p3,last=p4)

patch_daily_temps(gap_weather,patch_weather)

patch_daily_temps(gap_weather,patch_weather,max_mean_bias=0.1,time_interval="2 weeks")