Spatialtemporal Smoothing of Data

There are options in NASAPrecipitation.jl to spatialtemporally smooth data.

smoothing(
    <NASAPrecipitation Dataset>,
    <GeoRegion>,
    <smoothing options> ...,
)

It is important to see the documentation and see which options are available for each type of dataset, but we summarize it in the table below:

Type	Spatial	Temporal
IMERGHalfHourly	Yes	Yes (specify hours)
IMERGDaily	Yes	Yes (specify days)
IMERGMonthly	Yes	No
TRMM3Hourly	Yes	Yes (specify hours)
TRMMDaily	Yes	Yes (specify days)
TRMMMonthly	Yes	No

Setup

using NASAPrecipitation
using CairoMakie
using DelimitedFiles

download("https://raw.githubusercontent.com/natgeo-wong/GeoPlottingData/main/coastline_resl.txt","coast.cst")
coast = readdlm("coast.cst",comments=true)
clon  = coast[:,1]
clat  = coast[:,2]
nothing

Spatial Smoothing Example

Let's define a rectangular domain in Southeast Asia and download some NASAPrecipitation data for this region:

npd = IMERGMonthly(start=Date(2015),stop=Date(2015,3))
geo = RectRegion("TMP","GLB","Southeast Asia",[15,-15,125,95],savegeo=false)
download(npd,geo)
lsd = getLandSea(npd,geo)

We then proceed to perform a spatial smoothing (1.0º longitude, 0.2º latitude) of the data.

npd_smth = IMERGMonthly(start=Date(2015,2),stop=Date(2015,2))
smoothing(npd_smth,geo,smoothlon=1,smoothlat=0.2)

Now let us compare the difference between the raw and smoothed data.

ds_raw  = read(npd_smth,geo,Date(2015))
ds_smth = read(npd_smth,geo,Date(2015),smooth=true,smoothlon=1,smoothlat=0.2)

prcp_raw  = ds_raw["precipitation"][:,:,11]  * 86400
prcp_smth = ds_smth["precipitation"][:,:,11] * 86400

close(ds_raw)
close(ds_smth)

fig = Figure()
aspect = (maximum(lsd.lon)-minimum(lsd.lon)+2)/(maximum(lsd.lat)-minimum(lsd.lat)+2)

ax1 = Axis(
    fig[1,1],width=350,height=350/aspect,
    title="November 2015 (Raw)",xlabel="Longitude / º",ylabel="Latitude / º",
    limits=(minimum(lsd.lon)-1,maximum(lsd.lon)+1,minimum(lsd.lat)-1,maximum(lsd.lat)+1)
)
ax2 = Axis(
    fig[1,2],width=350,height=350/aspect,
    title="November 2015 (Smoothed 1.0ºx0.2º)",xlabel="Longitude / º",
    limits=(minimum(lsd.lon)-1,maximum(lsd.lon)+1,minimum(lsd.lat)-1,maximum(lsd.lat)+1)
)
contourf!(
    ax1,lsd.lon,lsd.lat,prcp_raw,colormap=:Blues,
    levels=range(0,20,length=11),extendlow=:auto,extendhigh=:auto
)
contourf!(
    ax2,lsd.lon,lsd.lat,prcp_smth,colormap=:Blues,
    levels=range(0,20,length=11),extendlow=:auto,extendhigh=:auto
)
lines!(ax1,clon,clat,color=:black)
lines!(ax2,clon,clat,color=:black)

resize_to_layout!(fig)
fig

And we see that the data is smoothed in the longitude-direction. Because they are not smoothed equally in both longitude and latitude, the fields look stretched in the longitude direction.

API

Missing docstring.

Missing docstring for NASAPrecipitation.smoothing. Check Documenter's build log for details.