README.md

In laubblatt/cleaRskyQuantileRegression: Calculate monthly fractional clear-sky shortwave transmission from global radiation records

R package cleaRskyQuantileRegression

By Maik Renner, Max-Planck Institute for Biogeochemistry, Jena, Germany

2019-08-26

This package provides the functions to calculate the fractional clear-sky shortwave transmission and shortwave clear-sky fluxes at the surface using (half-)hourly global radiation records. The method is validated and described in

Renner, M., M. Wild, M. Schwarz, and A. Kleidon. Estimating Shortwave Clear-Sky Fluxes from Hourly Global Radiation Records by Quantile Regression. Earth and Space Science, 2019. https://doi.org/10.1029/2019EA000686.

Main functionality is to perform a Quantile Regression on data of observed global radiation (total incoming shortwave radiation) records with potential solar radiation as a function of location and date and time. The package also provides a function to perfrom the regression for different time windows, which can be required when there is persistent cloud cover. The derived fractional transmission of clear-sky solar radiation can be used to calculate the corresponding clear sky fluxes and the shortwave cloud radiative effect. The method is intended for observational sites where no data on diffuse and direct radiation is available. The method only requires global radiation in hourly or half-hourly resolution.

The package also provide functions to calculate the potential shortwave radiation at the surface, i.e without any clouds and aerosols. Therefore Location (lat,lon) and Date and Time are required. The function is based on an code of the R packages REddyProc and solartime maintained by Thomas Wutzler (MPI-BGC)

Public release:

Instructions

To install the R package use:

library(devtools)
install_github("laubblatt/cleaRskyQuantileRegression")
 ```

## Working Example and illustration of the method
Load a dataset of half-hourly solar radiation from the site Lindenberg, Germany for the year 2003

```R
library(cleaRskyQuantileRegression)
data(LIN2003)
 ```

Extract data for August 2003
```R
(dat = LIN2003[year(Date) == 2003 & month(Date) == 8, ])

Calculate potential surface solar radiation

dat[ , Rsdpot_12 := calc_PotRadiation_CosineResponsePower(doy = yday(Date), hour = Time/3600 + 0.25,
                                                          latDeg = 52.21 ,
                                                          longDeg = 14.122,
                                                          timeZone = 0, isCorrectSolartime = TRUE,
                                                          cosineResponsePower = 1.2 )]

Apply Qunatile regression and plot for one month

library(latticeExtra)
lab_Rsdpot12_name = expression(bold("Potential Shortwave Radiation ")(W*m^-2))
lab_Rsd      = expression(bold("Observed Shortwave Radiation ")(W*m^-2))
xyplot(IncomingShortwave ~  Rsdpot_12, data = dat,
       xlab = list(lab_Rsdpot12_name,cex = 1.3), ylab = list(label=lab_Rsd, cex=1.3),
       type = c("p","g"), pch = ".", cex = 3,
       # main = paste0(dtyrmon[SiteCode == sico, unique(SiteName)]," ,Quantile Regression, tau = 0.90" ),
       panel = function(x, y, ...) {
         panel.xyplot(x, y, ...)
         panel.abline(rq(y~x, tau = 0.85), col=1, lwd = 2)
         # panel.abline(0,1, col=1, lty = 2, lwd = 2)
         panel.ablineq(0,1, col=1, lty = 2, lwd = 2, at = 0.89, label = "1:1", rotate = TRUE, fontfamily = "sans", cex = 1.5,pos = 3)
         panel.ablineq(rq(y~x, tau = 0.85), col=1, at = 0.7, rotate =TRUE,
                       pos = 3, label = "Slope of 85% Quantile", cex = 1.5, font = "Helvetica", fontface = 2 )
         panel.text(700,-20, "half hourly data of one month", fontfamily = "Helvetica")
       },
       grid = TRUE)

| | |:--:| | Illustration of Quantile Regression approach. The scatterplot shows a monthly sample of 30min observations of incoming shortwave radiation versus the corresponding potential shortwave radiation (Lindenberg, Germany, August 2003). The scatter forms a well-defined orthogonal triangle with clear-sky conditions close to the upper boundary. The quantile regression method (Koenker, 2005 ) allows to estimate this upper slope for a given quantile here 85%. This slope represents the fractional transmission of shortwave radiation under clear-sky conditions of that month. |

Perform regression for all months

R (rqmw = LIN2003[ , calc_ClearSky_QuantileRegression_MonthlyTimeWindow(Date,Time,IncomingShortwave, tau = 0.85, lat = 52.21, lon = 14.122, hourshift = 0.5,timeZone = 0)])

The result is a monthly data table of fractional shortwave transmission (ftau), the resulting mean flux of shortwave radiation without clouds (IncomingShortwaveClearSky) and the length of the sampling period (Window) which was derived internally by the goodness of fit (R1) and deviations of the monthly ftau from the site mean ftau. Resulting fluxes can be plotted: R plot(IncomingShortwavePotential ~ month, data = rqmw, type = "l", col = 4, ylab = "Monthly mean Shortwave Radiation (W/m2)", ylim = c(0,500)) lines(IncomingShortwaveClearSky ~ month, data = rqmw, type = "l", col =2) lines(IncomingShortwave ~ month, data = rqmw, type = "b") legend("topright", c("Potential", "Clear Sky flux", "Observed at surface"), col = c(4,2,1), lty = 1)

laubblatt/cleaRskyQuantileRegression documentation built on Nov. 27, 2019, 12:26 p.m.