<center>Tutorial for merging a satellite-based precipitation product with ground observations using `VARMER`</center>

In VARMER: Variational Merging

knitr::opts_chunk$set(echo = TRUE)

About

This vignette describes a basic application of the VARMER library, which provides methods to combine satellite-based or model-based gridded images with ground-based observed time-series [1].

We use part of the Valparaiso (Chile) case study provided by the package RFmerge [2,3]. This dataset includes products at daily temporal scale and at $0.05^\circ$ spatial resolution, from January to August 1983 (243 days). The Valparaiso dataset includes the following resources:

1) time series of rainfall observations,

2) metadata describing the spatial coordinates of the rain gauges,

3) the Climate Hazards Group InfraRed Precipitation with Station data version 2.0 (CHIRPSv2),

4) the Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks - Climate Data Record (PERSIANN-CDR), and

5) the Shuttle Radar Topography Mission (SRTM-v4) digital elevation model (DEM).

RFmerge requires 5 resources to create a final product. VARMER, on the other hand, requieres only 2. VARMER applies a Variational Merging Approach, which makes it a simpler approach requiring only two inputs: a gridded image (which provides the global spatial structure) and the ground-based observations (which provides the fidelity term).

Installation

Install the latest stable version (from CRAN):

install.packages("VARMER")

Setting up the environment

1. Load other packages that will be used in this analysis and VARMER:

library(zoo)
library(sf)
library(raster)
library(RFmerge)
library(tictoc)
library(cluster)
#library(ggplot2)
library(parallel)
library(VARMER)

Loading input data

First, daily time series of rainfall observations in 34 rain gauges located in Valparaiso will be used for this example, from 1983-01-01 to 1983-08-31, which are available in the ValparaisoPPts dataset provided in the RFmerge package (for your own application, this dataset might be read from a CSV file or a zoo file). In addition, the ValparaisoPPgis dataset contains information about the IDs and spatial coordinates of each one of the rain gauges (for your own application, this dataset might be read from a CSV file). Finally, ValparaisoSHP stores the sf polygon defining the outer borders of the study area (for your own application, this dataset might be read from an ESRI shapefile).

data(ValparaisoPPts)    
data(ValparaisoPPgis) 
data(ValparaisoSHP)  

Secondly, we need to load the satellite-based precipitation dataset. Following the RFmerge example, CHIRPSv2 [4] at a spatial resolution of $0.05^\circ$ is used to provide the global spatial structure.

chirps.fname   <- system.file("extdata/CHIRPS5km.tif", package="RFmerge")
CHIRPS5km <- brick(chirps.fname)

Basic exploratory data analysis

The multi-band geotiff (CHIRPS) provided in the RFmerge package do not store the date of the precipitation estimate as name of the corresponding layer. Therefore, before any exploratory analysis, we would give meaningful names to each band (layer) in CHIRPS5km:

ldates                <- seq(from=as.Date("1983-01-01"), to=as.Date("1983-08-31"), by="days")
names(CHIRPS5km)      <- ldates

Then, we want to visualize the first six rows of the spatial metadata:

head(ValparaisoPPgis)

Plotting the daily precipitation time series for the first station (code: P5101005).

main <- paste("Daily precipitation for the station", ValparaisoPPgis$Code[1])
ylab <- "Precipitation [mm]"
x.ts <- ValparaisoPPts[,1]

plot(x.ts, main=main, ylab= ylab, col="blue")
grid()

Plotting the accumulated precipitation estimates for the first eight months of 1983 from CHIRPS, and overlying the boundaries of the study area (only its first attribute):

chirps.total   <- sum(CHIRPS5km, na.rm= FALSE)

plot(chirps.total, main = "CHIRPSv2 [Jan - Aug] ", xlab = "Longitude", ylab = "Latitude")
plot(ValparaisoSHP[1], add=TRUE, col="transparent")

Running VARMER

Setup

Finally, you need to define the output directory (drty.out) before running VARMER. Then, you can run the varmer.ts function as follows:

varmer.ts(x=ValparaisoPPts, x.metadata=ValparaisoPPgis,
          v=CHIRPS5km, lat='lat', lon='lon',
          drty.out="~/Documentos/dataset_valparaiso/")

\newpage

Expected outputs

If VARMER runs without problems, the final merged product (individual GeoTiff files) will be stored in your user-defined drty.out directory.

Software details

This tutorial was built under:

sessionInfo()$platform
sessionInfo()$R.version$version.string 
paste("VARMER", sessionInfo()$otherPkgs$VARMER$Version)

References

1. Ulloa, J., Samaniego, E., Campozano, L., & Ballari, D. (2018). A variational merging approach to the spatial description of environmental variables. Journal of Geophysical Research: Atmospheres, 123. https://doi.org/10.1002/2017JD027982.

2. Zambrano-Bigiarini, M.; Baez-Villanueva O.M.; Giraldo-Osorio, J. (2020). RFmerge: Merging of Satellite Datasets with Ground Observations using Random Forests. R package version 0.1-10. URL https://cran.r-project.org/package=RFmerge. DOI:10.5281/zenodo.3581515.

3. Baez-Villanueva, O. M.; Zambrano-Bigiarini, M.; Beck, H.; McNamara, I.; Ribbe, L.; Nauditt, A.; Birkel, C.; Verbist, K.; Giraldo-Osorio, J.D.; Thinh, N.X. (2020). RF-MEP: a novel Random Forest method for merging gridded precipitation products and ground-based measurements, Remote Sensing of Environment, 239, 111610. doi10.1016/j.rse.2019.111606.

4. Funk, C., Peterson, P., Landsfeld, M., Pedreros, D., Verdin, J., Shukla, S., Husak, G., Rowland, J., Harrison, L., Hoell, A., and Michaelsen, J. (2015) The climate hazards infrared precipitation with stations-a new environmental record for monitoring extremes, Sci Data, 2, 150 066, doi:10.1038/sdata.2015.66.

VARMER documentation built on Feb. 8, 2021, 1:07 a.m.