A hybrid algorithm for estimating the chlorophyll-a
concentration across different trophic states in Asian inland waters
Optical water type discrimination and tuning remote sensing
band-ratio algorithms: application to retrieval of chlorophyll and kd
(490) in the Irish and Celtic seas
Influence of a red band-based water classification approach on
chlorophyll algorithms for optically complex estuaries
A soft-classification-based chlorophyll-a estimation method using
MERIS data in the highly turbid and eutrophic Taihu Lake
A system to measure the data quality of spectral remote-sensing
reflectance of aquatic environments
An improved optical classification scheme for the Ocean Colour
Essential Climate Variable and its applications
A
system to measure the data quality of spectral remote-sensing
reflectance of aquatic environments
Water Classification
The reference Rrs spectra were first normalized by their respective
root of sum of squares (RSS), \[
nR_{rs}(\lambda)=\frac{R_{rs}}{[\sum^N_{i=1}R_{rs}(\lambda_i)^2]^{1/2}}
\] where the index N represents the total number of wavelengths,
varying from 1 to 9 and \(\lambda_i\)
corresponds to the wavelengths of 412, 443, 488, 510, 531, 547, 555,
667, and 678 nm. The nRrs spectra vary over the range between 0 and 1,
while it retains the ‘‘shapes’’ pertaining to the original Rrs spectra,
i.e., the band ratios of nRrs(\(\lambda_i\)) remain the same as Rrs(\(\lambda_i\)).
The number of data clusters k was evaluated using the gap method
[Tibshirani et al., 2001]. The gap value is defined as: \[
GAP_n(k)=E_n^*[log(W_k)]-log(W_k)
\] where n is the sample size, k is the number of clusters being
evaluated, and Wk is the pooled within-cluster dispersion measurement,
with \[
W_k=\sum_{r=1}^k\frac{1}{2n_r}D_r
\] where nr is the number of data points in cluster r, and \(D_r\) is the sum of the pair-wise distances
for all points in cluster r.
The expected value \(E_n^*[log(W_k)]\) is determined by Monte
Carlo sampling from a reference distri- bution, and $log(Wk) is computed
from the sample data. According to the gap method, the optimum cluster
number of the nRrs data is determined as 23. Interestingly, this number
is nearly the same as that of Forel- Ule water type classes developed
100 years ago [Arnone et al., 2004].
The unsupervised method, K-means clustering technique, was further
used to group the nRrs spectra.
But what i want to do is just"Case1, Case2, and Case2 into slightly
turbid, moderet turbid and highly turbid"
Anyway, I'm gonna to read it over as I also need it to do the
verification/validation.
Quantitatively measurement
match up \(R_{rs}^*(\lambda^{'})\) with \(nR_{rs}(\lambda)\) with regard to the
wavelengths. If \(R^*_{rs}\) has more
spectral bands than that of \(nR_{rs}(\lambda)\), we will only choose the
same wavelengths with \(nR_{rs}(\lambda)\) for further analysis. If
\(R^*_{rs}\) has fewer wavelengths than
the \(nR_{rs}(\lambda)\) spectra (i.e.,
\(N(\lambda^{'})<9\)), a subset
of \(nR_{rs}(\lambda^{'})\) and
associated upper boundary spectra \(nR_{rs}^U(\lambda^{'})\) and lower
boundary spectra \(nR_{rs}^L(\lambda^{'})\) will be
extrated first for \(\lambda^{'}\).
the normalization of \(R_{rs}^*\) spectra following equation (1).
For the case of \(N(\lambda^{'})<9\), the new \(nR_{rs}(\lambda^{'})\) specta will be
rescaled through the normalization procedure of eq(1) so that the RSS of
\(nR_{rs}(\lambda^{'})\) is equal
to 1 (?). Further, the new upper and lower boundary
spectra \(nR_{rs}^{U}(\lambda^{'})\) and \(nR_{rs}^{L}(\lambda^{'})\) will also be
rescaled by the newly rescaled \(nR_{rs}(\lambda^{'})\) spectra as below
\[
nR_{rs}^{U}(\lambda)=\frac{nR_{rs}^{U}(\lambda)}{[\sum_{i=1}^NnR_{rs}(\lambda_i)^2]^{1/2}}
\]
assign a water type to the target spectrum by comparing it with
the reference nRrs spectra. The spectra similarity between the target
spectrum \(nR_{rs}^*\) and refference
spectra nRrs are estimated using a spectral angle mapper(SAM)[Kruse et
al., 1993], \[
cos\
\alpha=\frac{\sum_{i=1}^{N}[nR_{rs}^**nR_{rs}]}{\sqrt{\sum_{i=1}^{N}[nR_{rs}^{*}(\lambda_i)]^2\sum_{i=1}^{N}[nR_{rs}(\lambda_i)]^2}}
\] Where \(\alpha\) is the angle
formed between the refference spectrum nRrs and the normalized target
spectrum \(nR_{rs}^*\). As a spectral
classifier, SAM is able to determine the spectral similarity by treating
them as vertors in a space with dimensionality equal to the number of
bands, N. The water type of the target spectrum \(nR_{rs}^*\) is identified as one with the
largest cosine values (equivalent to the smllest angles).
the computation of QA scores by comparing the target spectrum
\(nRrs\) with the upper and lower
boundaries (\(nRrs^U\) and \(nRrs^L\)) of the corresponding water type.
The number of wavelengths where \(nR_{rs}^*\) falling within the boundaries
is counted, and used to derive the total score (\(C_{tot}\)) for the \(nR_{rs}\) spectrum, \[
C_{tot}=\frac{C(\lambda_1)+C(\lambda_2)+\ldots+C(\lambda_N)}{N}
\]
Where \(C_{i}\) is the
wavelength-specific score with N the total number of wavelengths for
both \(R_{rs}^*\) and \(R_{rs}^{ref}\). At wavelength \(\lambda_i\) , for example, if \(R_{rs}^*(\lambda_i)\) is found beyond
either the upper or lower boundary of nRrs, a score of 0 will be
assigned to this wavelength; otherwise score=1. As suggested by equation
(7), the total score \(C_{tot}\) will
vary within the range of [0, 1]. A higher score indicates higher data
quality.
To account for the measurement uncertainty and possible
data-processing errors and likely insufficient data coverage, the
original upper boundary and lower boundary are slightly modified by
\(\pm0.5%\),\(nRrs^U=nRrs^U\times(1+0.005)\ and\
nRrs^L=nRrs^L\times(1-0.005)\), respectively. Note that this
added range of 0.5% is one order of magnitude smaller than the projected
accuracy for radiance measurement [Hooker et al., 1992].
image-00030115182324282
Result
Because all field measurements are discrete, it is likely that the
database used here does not cover every water types and/or there are
situations where the range of Rrs variability goes beyond the domains
defined here. Such a limitation can be updated or revised when more
high-quality in situ measurements are avail- able. A MATLABVR script is
made available (http://oceanoptics.umb.edu/score_metric/) to facilitate
the evaluation and refinement of the score metrics. Nevertheless, this
QA scheme provides an easily applicable system to quantitatively
evaluate the quality of individual Rrs spectra.
function[maxCos, cos, clusterID, totScore] = QAscores_matrix(test_Rrs, test_lambda) % Quality assurance system for Rrs spectra (version 1) % % Author: Jianwei Wei, University of Massachusetts Boston % Email: Jianwei.Wei@umb.edu % Nov-01-2016 % % ------------------------------------------------------------------------------ % KNOWN VARIABLES : ref_nRrs -- Normalized Rrs spectra per-determined from water clustering (23x9 matrix) % ref_lambda -- Wavelengths for ref_nRrs (1x9 matrix) % upB -- Upper boundary (23x9 matrix) % lowB -- Lower boundary (23x9 matrix) % % INPUTS: test_Rrs -- Rrs spectra for testing (units: sr^-1); % a row vector % test_lambda-- Wavelengths for test_Rrs % % OUTPUTS: maxCos -- maximum cosine values % cos -- cosine values for every ref_nRrs spectra % clusterID -- idenfification of water types (from 1-23) % totScore -- total score assigned to test_Rrs % ------------------------------------------------------------------------------ % % NOTE: % 1) Nine wavelengths (412, 443, 488, 510, 531, 547, 555, 667, 678nm) are assumed in the model % 2) If your Rrs data were measured at other wavelength, e.g. 440nm, you may want to change 440 to 443 before the model run; % or modify the code below to find a cloest wavelength from the nine bands. % 3) The latest version may be found online at HTTP://oceanoptics.umb.edu/score_metric % % Reference: % Wei, Jianwei; Lee, Zhongping; Shang, Shaoling (2016). A system % to measure the data quality of spectral remote sensing % reflectance of aquatic environments. Journal of Geophysical Research, % 121, doi:10.1002/2016JC012126 % % ------------------------------------------------------------------------------ % Apr. 5 2017, Keping Du % 1) Vectorize code % 2) totScore takes account of NaN bands % 3) add input data check % Apr. 11, 2017 % 4) compatible with previous matlab version (tested on v2014a) % % INPUTS: % test_Rrs -- matrix (inRow*inCol), each row represents one Rrs spectrum % OUTPUTS: % maxCos,clusterID,totScore -- row vector (1*inRow) % cos -- matrix (refRow[23]*inRow) % % Note: % 1) nanmean, nansum need statistics toolbox % 2) on less memory and multi-core system, it may further speedup using % parfor % % ------------------------------------------------------------------------------ %% check input data [row_lam, len] = size(test_lambda); if( row_lam ~= 1 ) test_lambda = test_lambda'; [row_lam, len] = size(test_lambda); end
[row, col] = size(test_Rrs); if( len~=col && len~=row) error('Rrs and lambda size mismatch, please check the input data!'); elseif( len == row ) test_Rrs = test_Rrs'; end
I believe this is actually not the true evaluation that do not need
any 'in-situ' data.
It should met a lot of problem when applied it to other regions and
sensors.
I need to modify the former code if I want to use it in the
evaluation of fusion result.
A
hybrid algorithm for estimating the chlorophyll-a concentration
across different trophic states in Asian inland waters
image-00030115201801830
\(MCI\le0.001\) Slightly turbid
\(0.001<MCI\le0.0016\) Moderate
turbid
\(MCI>0.0016\) Highly turbid
Optical
water type discrimination and tuning remote sensing band-ratio
algorithms: Application to retrieval of chlorophyll and Kd(490)
in the Irish and Celtic Seas
image-00030115202628690
This is a very very emprical classification.
Influence
of a red band-based water classification approach on chlorophyll
algorithms for optically complex estuarie
image-00030115202825460
I can try this. Along with the Chia/TSM ratio
A
soft-classification-based chlorophyll-a estimation method using
MERIS data in the highly turbid and eutrophic Taihu Lake
SGLI didn't have red edge wavelength, which introduced a lot of
problem in this work.
Phytoplankton
pigments cause high reflectance in red edge wavelengths (e.g.
709 nm), and low reflectance in red wavelengths (e.g. 681 nm)
because of the phytoplankton pigment absorption peak around 681 nm.
An
improved optical classification scheme for the Ocean Colour Essential
Climate Variable and its applications
An absorption-related optical classification approach was developed
by considering the contributions of different particle sources,
aph(443)/ ad(443) (Table 3), to divide the study sites into three water
classes, i.e., detritus-dominated waters (aph(443)/ad(443) < 0.2,
Wd), pigment- dominated waters (aph(443)/ad(443) ≥ 1.0, Wp) and
intermediate waters (0.2 ≤ aph(443)/ad(443) < 1.0, Wm),
自己可以尝试一下。
Remotely
estimating total suspended solids concentration in clear to extremely
turbid waters using a novel semi-analytical method