pRecipe was conceived back in 2020 as part of MRVG’s doctoral dissertation at the Faculty of Environmental Sciences, Czech University of Life Sciences Prague, Czechia. Designed with reproducible science in mind, pRecipe facilitates the download, exploration, visualization, and analysis of multiple precipitation data products across various spatiotemporal scales (Vargas Godoy and Markonis 2023).

“Like civilization and technology, our understanding of the global water cycle has been continuously evolving, and we have adapted our quantification methods to better exploit new technological resources. The accurate quantification of global water fluxes and storage is crucial in studying the global water cycle.”

Before We Start

Like many other R packages, pRecipe has some system requirements:

PROJ
Geospatial Data Abstraction Library (GDAL)
Network Common Data Form (NetCDF)

Data

pRecipe database hosts 27 different precipitation data sets; seven gauge-based, eight satellite-based, seven reanalysis, and five hydrological model precipitation products. Their native specifications, as well as links to their providers, and their respective references are detailed in the following subsections. We have already homogenized, compacted to a single file, and stored them in a Zenodo repository under the following naming convention:

<data set>_<variable>_<units>_<coverage>_<start date>_<end date>_<resolution>_<time step>.nc

The pRecipe data collection was homogenized to these specifications:

<variable> = total precipitation (tp)
<units> = millimeters (mm)
<resolution> = 0.25°

E.g., GPCP v2.3 (Adler et al. 2018) would be:

gpcp_tp_mm_global_197901_202205_025_monthly.nc

Gauge-Based Products

		Spatial Coverage
Data Set	Spatial Resolution	Global	Land	Ocean	Temporal Resolution	Record Length	Get Data	Reference
CPC-Global	0.5°		x		Daily	1979/01-2022/08	Download	P. Xie, Chen, and Shi (2010)
CRU TS v4.06	0.5°		x		Monthly	1901/01-2021/12	Download	Harris et al. (2020)
EM-EARTH	0.1°		x		Daily	1950/01-2019/12	Download	Tang, Clark, and Papalexiou (2022)
GHCN v2	5°		x		Monthly	1900/01-2015/05	Download	Peterson and Vose (1997)
GPCC v2020	0.25°		x		Monthly	1891/01-2022/08	Download	Schneider et al. (2011)
PREC/L	0.5°		x		Monthly	1948/01-2022/08	Download	Chen et al. (2002)
UDel v5.01	0.5°		x		Monthly	1901/01-2017/12	Download	Willmott and Matsuura (2001)

Satellite-Based Products

		Spatial Coverage
Data Set	Spatial Resolution	Global	Land	Ocean	Temporal Resolution	Record Length	Get Data	Reference
CHIRPS v2.0	0.05°		50°SN		Monthly	1981/01-2022/07	Download	Funk et al. (2015)
CMAP	2.5°	x	x	x	Monthly	1979/01-2022/07	Download	Pingping Xie and Arkin (1997)
CMORPH	0.25°	60°SN	60°SN	60°SN	Daily	1998/01-2021/12	Download	Joyce et al. (2004)
GPCP v2.3	0.5°	x	x	x	Monthly	1979/01-2022/05	Download	Adler et al. (2018)
GPM IMERGM v06	0.1°	x	x	x	Monthly	2000/06-2020/12	Download	G. J. Huffman et al. (2019)
MSWEP v2.8	0.1°	x	x	x	Monthly	1979/02-2022/06	Download	Beck et al. (2019)
PERSIANN-CDR	0.25°	60°SN	60°SN	60°SN	Monthly	1983/01-2022/06	Download	Ashouri et al. (2015)
TRMM 3B43 v7	0.25°	50°SN	50°SN	50°SN	Monthly	1998/01-2019/12	Download	George J. Huffman et al. (2010)

Reanalysis Products

		Spatial Coverage
Data Set	Spatial Resolution	Global	Land	Ocean	Temporal Resolution	Record Length	Get Data	Reference
20CR v3	1°	x	x	x	Monthly	1836/01-2015/12	Download	Slivinski et al. (2019)
ERA-20C	1.125°	x	x	x	Monthly	1900/01-2010/12	Download	Poli et al. (2016)
ERA5	0.25°	x	x	x	Monthly	1959/01-2021/12	Download	Hersbach et al. (2020)
JRA-55	1.25°	x	x	x	Monthly	1958/01-2021/12	Download	Kobayashi et al. (2015)
MERRA-2	0.5° x 0.625°	x	x	x	Monthly	1980/01-2023/01	Download	Gelaro et al. (2017)
NCEP/NCAR R1	1.875°	x	x	x	Monthly	1948/01-2022/08	Download	Kalnay et al. (1996)
NCEP/DOE R2	1.875°	x	x	x	Monthly	1979/01-2022/08	Download	Kanamitsu et al. (2002)

Hydrological Model Forcing

		Spatial Coverage
Data Set	Spatial Resolution	Global	Land	Ocean	Temporal Resolution	Record Length	Get Data	Reference
FLDAS	0.1°		x		Monthly	1982/01-2021/12	Download	McNally et al. (2017)
GLDAS CLSM v2.0	0.25°		x		Daily	1948/01-2014/12	Download	Rodell et al. (2004)
GLDAS NOAH v2.0	0.25°		x		Monthly	1948/01-2014/12	Download	Rodell et al. (2004)
GLDAS VIC v2.0	1°		x		Monthly	1948/01-2014/12	Download	Rodell et al. (2004)
TerraClimate	4\(km\)		x		Monthly	1958/01-2021/12	Download	Abatzoglou et al. (2018)

Introduction to pRecipe

In this introductory demo we will first download the GPM-IMERGM data set. We will then subset the downloaded data over South America for the 2001-2015 period, and crop it to the national scale for Bolivia. In the next step, we will generate time series for our data sets and conclude with the visualization of our data.

NOTE: While the functions in pRecipe are intended to work directly with its data inventory. pRecipe can handle most other data sets in “.nc” format, as well as any other “.nc” file generated by its functions.

Installation

install.packages('pRecipe')
library(pRecipe)

Data Download

Downloading the entire data collection or only a few data sets is quite straightforward. You just call the download_data function, which has four arguments dataset, path, domain, and timestep.

dataset is set to “all” by default, but you can specify the names of your data sets of interest only.
path is set to “.” by default. I.e., the current working directory. By replacing it for [your_project_folder], the downloaded files will be stored in [your_project_folder] instead.
domain is set to “raw” by default, but you can specify the domain of your interest only. E.g., “ocean” for ocean only data sets (For availability please check the Data section).
timestep is set to “monthly” by default, but if you prefer you can also download annual data with “yearly”.

Let’s download the GPM-IMERGM data set and inspect its content with infoNC:

download_data(dataset = 'gpm-imerg')
gpm_global <- raster::brick('gpm-imerg_tp_mm_global_200006_202012_025_monthly.nc')
infoNC(gpm_global)

[1] "class      : RasterBrick "
[2] "dimensions : 720, 1440, 1036800, 256  (nrow, ncol, ncell, nlayers)"
[3] "resolution : 0.25, 0.25  (x, y)"
[4] "extent     : -180, 180, -90, 90  (xmin, xmax, ymin, ymax)"
[5] "crs        : +proj=longlat +datum=WGS84 +no_defs "
[6] "source     : gpm-imerg_tp_mm_global_200006_202012_025_monthly.nc "
[7] "names      : X2000.06.01, X2000.07.01, X2000.08.01, X2000.09.01, X2000.10.01, X2000.11.01, X2000.12.01, X2001.01.01, X2001.02.01, X2001.03.01, X2001.04.01, X2001.05.01, X2001.06.01, X2001.07.01, X2001.08.01, ... "
[8] "Date/time  : 2000-06-01, 2021-09-01 (min, max)"
[9] "varname    : tp "

Processing

Once we have downloaded our database, we can start processing the data with:

crop_data to crop the data using a shapefile.
fldmean to generate a time series by taking the area weighted average over each timestep.
remap to go from the native resolution (0.25°) to coarser ones (e.g., 0.5°, 1°, 1.5°, …).
subset_data to subset the data in time and/or space.
yearstat to aggregate the data from monthly into annual.

Subset

To subset our data to a desired region and period of interest, we use the subset_data function, which has three arguments x, box, and yrs.

x Raster* object or a data.table or a filename (character).
box is the bounding box of the region of interest with the coordinates in degrees in the form (xmin, xmax, ymin, ymax).
yrs is the period of interest with years in the form (start_year, end_year).

Let’s subset the GPM-IMERGM data set over South America (-96, -30, -56, 24) for the 2001-2015 period, and inspect its content with infoNC:

gpm_subset <- subset_data(gpm_global, box = c(-96, -30, -56, 24), yrs = c(2001, 2015))
infoNC(gpm_subset)

[1] "class      : RasterBrick "
[2] "dimensions : 320, 264, 84480, 180  (nrow, ncol, ncell, nlayers)"
[3] "resolution : 0.25, 0.25  (x, y)"
[4] "extent     : -96, -30, -56, 24  (xmin, xmax, ymin, ymax)"
[5] "crs        : memory"
[6] "source     : r_tmp_2023-07-07_185756.586009_13478_99533.grd "
[7] "names      :  X2001.01.01,  X2001.02.01,  X2001.03.01,  X2001.04.01,  X2001.05.01,  X2001.06.01,  X2001.07.01,  X2001.08.01,  X2001.09.01,  X2001.10.01,  X2001.11.01,  X2001.12.01,  X2002.01.01,  X2002.02.01,  X2002.03.01, ... "
[8] "min values : 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 2.393141e-04, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, 0.000000e+00, ... "
[9] "max values :    1069.6940,     917.3006,    1102.5068,     915.1335,    1549.5077,    1169.6469,    1483.0360,    1453.4453,    1624.6550,    1233.0613,    1715.0280,    1285.7706,     881.6859,     874.7393,     883.4313, ... "
[10] "time       : 2001-01-01, 2015-12-01 (min, max)"

Crop

To further crop our data to a desired polygon other than a rectangle, we use the crop_data function, which has two arguments x, and y.

x Raster* object or a data.table or a *.nc filename (character).
y is a “.shp” filename (character).

Let’s crop our GPM-IMERG subset to cover only Bolivia with the respective shape file, and inspect its content with infoNC:

gpm_bol <- crop_data(gpm_subset, "gadm41_BOL_0.shp")
infoNC(gpm_bol)

[1] "class      : RasterBrick "
[2] "dimensions : 54, 50, 2700, 180  (nrow, ncol, ncell, nlayers)"
[3] "resolution : 0.25, 0.25  (x, y)"
[4] "extent     : -69.75, -57.25, -23, -9.5  (xmin, xmax, ymin, ymax)"
[5] "crs        : +proj=longlat +datum=WGS84 +no_defs "
[6] "source     : memory"
[7] "names      :  X2001.01.01,  X2001.02.01,  X2001.03.01,  X2001.04.01,  X2001.05.01,  X2001.06.01,  X2001.07.01,  X2001.08.01,  X2001.09.01,  X2001.10.01,  X2001.11.01,  X2001.12.01,  X2002.01.01,  X2002.02.01,  X2002.03.01, ... "
[8] "min values : 4.359420e+01, 5.965404e+01, 9.195424e+00, 2.650523e+00, 4.473422e-01, 4.649702e-03, 3.581941e-04, 1.511060e-02, 3.136731e-01, 5.168897e-01, 3.443884e-01, 1.019173e+01, 3.299495e+00, 2.491986e+01, 1.160967e+01, ... "
[9] "max values :    613.21777,    530.11438,    497.65503,    371.26581,    216.29959,    136.70122,    209.37540,    124.56583,    166.48785,    313.69836,    472.29553,    487.41364,    613.77014,    673.70099,    549.12671, ... "
[10] "time       : 2001-01-01, 2015-12-01 (min, max)"

Generate Time series

To make a time series out of our data, we use the fldmean function, which has one argument x.

x Raster* object or a data.table or a *.nc filename (character).

Let’s generate the time series for our three different GPM-IMERGM data sets (Global, South America, and Bolivia), and inspect its first 12 rows:

gpm_global_ts <- fldmean(gpm_global)
head(gpm_global_ts, 12)

          date     value
        <Date>     <num>
 1: 2000-06-01  93.64844
 2: 2000-07-01  96.05852
 3: 2000-08-01  94.18216
 4: 2000-09-01  90.43190
 5: 2000-10-01  93.91238
 6: 2000-11-01  93.61439
 7: 2000-12-01  96.70333
 8: 2001-01-01  94.67989
 9: 2001-02-01  86.00950
10: 2001-03-01  96.15177
11: 2001-04-01  97.05069
12: 2001-05-01 100.53676

gpm_subset_ts <- fldmean(gpm_subset)
head(gpm_subset_ts, 12)

          date     value
        <Date>     <num>
 1: 2001-01-01 106.52438
 2: 2001-02-01  89.98158
 3: 2001-03-01 113.35350
 4: 2001-04-01 107.26019
 5: 2001-05-01 123.50707
 6: 2001-06-01  94.20347
 7: 2001-07-01 102.07352
 8: 2001-08-01  94.62878
 9: 2001-09-01  96.31932
10: 2001-10-01 112.90529
11: 2001-11-01 102.68565
12: 2001-12-01 113.49551

gpm_bol_ts <- fldmean(gpm_bol)
head(gpm_bol_ts, 12)

          date     value
        <Date>     <num>
 1: 2001-01-01 233.87604
 2: 2001-02-01 183.34294
 3: 2001-03-01 165.97789
 4: 2001-04-01  85.02165
 5: 2001-05-01  63.68961
 6: 2001-06-01  24.68989
 7: 2001-07-01  31.89638
 8: 2001-08-01  17.94735
 9: 2001-09-01  55.87102
10: 2001-10-01 103.33750
11: 2001-11-01 163.95180
12: 2001-12-01 156.72036

Visualize

Either after we have processed our data as required or right after downloaded, we have different options to visualize our data:

plot_box to see a seasonal boxplot.
plot_density to see the empirical density of monthly precipitation.
plot_heatmap to see a heatmap of all monthly values.
plot_line to see the average time series.
plot_map to see the Cartesian lon-lat map of the first raster layer.
plot_summary to see line, heatmap, box, and density plot together in a single plot.
plot_taylor to see a Taylor Diagram (requires a referential data set).

Let’s plot our three different GPM-IMERGM data sets (Global, South America, and Bolivia)

Maps

To see a map of any data set raw or processed, we use plot_map.

plot_map(gpm_global)

plot_map(gpm_subset)

plot_map(gpm_bol)

Time Series Visuals

Line

plot_line(gpm_global_ts)

Introduction to pRecipe

Mijael Rodrigo Vargas Godoy, Yannis Markonis

2024-01-31

Before We Start

Data

Gauge-Based Products

Satellite-Based Products

Reanalysis Products

Hydrological Model Forcing

Introduction to pRecipe

Installation

Data Download

Processing

Subset

Crop

Generate Time series

Visualize

Maps

Time Series Visuals

Line