{
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# Ocean Color Classifications\n",
"\n",
"Here we apply simple & standard classifiers to `ocean color data` over a wide region. \n",
"\n",
"_Credit: sample data file was provided by T. Jackson from [Plymouth Marine Lab](https://www.pml.ac.uk)_\n",
"\n",
"![\"Drawing\"](\"../figs/cbiomes-01.png\")
"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "subslide"
}
},
"source": [
"### Activate packages for later use\n",
"\n",
"It is assumed that listed packages have aleary been installed using `julia`'s package manager (documentation available [here](https://docs.julialang.org/en/)). "
]
},
{
"cell_type": "code",
"execution_count": 1,
"metadata": {},
"outputs": [],
"source": [
"using OceanColorData, Plots, NCDatasets"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "subslide"
}
},
"source": [
"### Observed wavebands\n",
"\n",
"Currently, the `OC-CCI` [satellite data set](https://esa-oceancolour-cci.org) provides remotely sensed reflectance at 6 wavelengths (`wv_cci` in `nm`)"
]
},
{
"cell_type": "code",
"execution_count": 2,
"metadata": {},
"outputs": [],
"source": [
"wv_cci=[412, 443, 490, 510, 555, 670];"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "subslide"
}
},
"source": [
"### Optical classification using reflectances\n",
"\n",
"`Fuzzy logic` classifiers defined in [Moore et al 2009](https://doi.org/10.1016/j.rse.2009.07.016) and [Jackson et al 2017](http://dx.doi.org/10.1016/j.rse.2017.03.036) can be used to assign optical class memberships from an `Rrs` vector. While Moore et al define `n=8` classes using an in-situ database, Jackson et al instead define `n=14` classes using a satellite database. The latter benefits from better data coverage across all of the ecological provinces of the global ocean and is used in `OC-CCI`. \n",
"\n",
"In both cases the classifier is encoded in a mean reflectance spectra (`M[i][1:6]`) and a covariance matrix (`S[i][1:6,1:6]`) provided for each optical class (`i` in `1:n`). Class memberships are then derived by computing the squared Mahalanobis distance to each `M[i]` and passing the result to cumulative chi-squared distribution function (Equations 11 and 12 in [Moore et al 2011](https://doi.org/10.1109/36.942555))."
]
},
{
"cell_type": "code",
"execution_count": 3,
"metadata": {
"slideshow": {
"slide_type": "subslide"
}
},
"outputs": [],
"source": [
"(M,Sinv)=Jackson2017();"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "slide"
}
},
"source": [
"### Read In Data Sample\n",
"\n",
"_Credit: sample data file was provided by T. Jackson from [Plymouth Marine Lab](https://www.pml.ac.uk)_"
]
},
{
"cell_type": "code",
"execution_count": 4,
"metadata": {},
"outputs": [],
"source": [
"dir0=\"../samples/\"; \n",
"fil=dir0*\"ESACCI-OC-RRS-sample-fv4.0.nc\"\n",
"ds = Dataset(fil)\n",
"\n",
"Rrs_412=ds[\"Rrs_412\"]; Rrs_443=ds[\"Rrs_443\"]; Rrs_490=ds[\"Rrs_490\"]; \n",
"Rrs_510=ds[\"Rrs_510\"]; Rrs_555=ds[\"Rrs_555\"]; Rrs_670=ds[\"Rrs_670\"];\n",
"lon=ds[\"lon\"]; lat=ds[\"lat\"];"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "subslide"
}
},
"source": [
"Plot one of the wave bands as an example."
]
},
{
"cell_type": "code",
"execution_count": 5,
"metadata": {
"slideshow": {
"slide_type": "-"
}
},
"outputs": [
{
"data": {
"image/svg+xml": [
"\n",
"\n"
]
},
"execution_count": 5,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"c=transpose(reverse(ds[\"Rrs_490\"][:,:,1],dims=2))\n",
"heatmap(lon,reverse(lat),c,title=\"Rrs at 490nm\")"
]
},
{
"cell_type": "markdown",
"metadata": {
"slideshow": {
"slide_type": "slide"
}
},
"source": [
"### Compute memberships\n",
"\n",
"Apply Classification to our 2D data sample and plot out a map."
]
},
{
"cell_type": "code",
"execution_count": 6,
"metadata": {
"lines_to_next_cell": 0
},
"outputs": [],
"source": [
"#Find points that have a full set of input\n",
"tmp=fill(false,size(Rrs_412))\n",
"for ii in eachindex(Rrs_412)\n",
" !ismissing(Rrs_412[ii]) ? tmp[ii]=!ismissing(Rrs_443[ii].*Rrs_490[ii].*Rrs_510[ii].*Rrs_555[ii].*Rrs_670[ii]) : nothing\n",
"end\n",
"ii=findall(tmp);\n",
"\n",
"#Apply classifier\n",
"mbrshp=fill(NaN,(1440,960,14))\n",
"for jj=1:length(ii); \n",
" kk=ii[jj]\n",
" Rrs_tmp=[Rrs_412[kk] Rrs_443[kk] Rrs_490[kk] Rrs_510[kk] Rrs_555[kk] Rrs_670[kk]]\n",
" mbrshp[kk[1],kk[2],:]=FuzzyClassification(M,Sinv,vec(Rrs_tmp))\n",
"end"
]
},
{
"cell_type": "code",
"execution_count": 7,
"metadata": {
"slideshow": {
"slide_type": "subslide"
}
},
"outputs": [
{
"data": {
"image/svg+xml": [
"\n",
"\n"
]
},
"execution_count": 7,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"c=transpose(reverse(mbrshp[:,:,10],dims=2))\n",
"heatmap(lon,reverse(lat),c,title=\"Membership for class 10\")"
]
}
],
"metadata": {
"@webio": {
"lastCommId": null,
"lastKernelId": null
},
"celltoolbar": "Slideshow",
"jupytext": {
"formats": "ipynb,jl:light"
},
"kernelspec": {
"display_name": "Julia 1.3.1",
"language": "julia",
"name": "julia-1.3"
},
"language_info": {
"file_extension": ".jl",
"mimetype": "application/julia",
"name": "julia",
"version": "1.3.1"
},
"rise": {
"enable_chalkboard": true
}
},
"nbformat": 4,
"nbformat_minor": 4
}