overview
DESCRIPTION
Overview. Overview of Dataflow data processing method Chlorophyll- a correction method Dataflow processing output Thoughts on supplementary data analysis methods. Processing the 2005 – 2013 DATAFLOW Data. Chlorophyll- a Correction Method. - PowerPoint PPT PresentationTRANSCRIPT
1
•Overview of Dataflow data processing method• Chlorophyll-a correction method•Dataflow processing output• Thoughts on supplementary data analysis methods
Overview
2
Processing the 2005 – 2013 DATAFLOW Data
Pair YSI chlorophyll-a
with extracted chlorophyll-a
data
Correct/calibrate YSI data using extracted
chlorophyll-a
InterpolationPost-process interpolation
results
3
• Purpose: To correct/calibrate YSI chl-a based on extracted chl-a• Different methods used in past by DEQ, VIMS,
HRSD• Variants:• How to lump data (by season, segment, year, etc.)• Whether or not to ln-transform• Whether to include additional explanatory variables
(turbidity, temperature)
• Magnitude of correction is small compared to normal variation (over space and time)
Chlorophyll-a Correction Method
4
• Recently recommended method was a compromise between monthly and multi-year lumping• Lump by season-segment-year• Use ln transform• Exclude turbidity and temperature
• However, it results in some unreasonable corrections; e.g.:• 192 ug/L 808 ug/L• 333 ug/L 954 ug/L
• Cause: Outliers and fewer number of high chl-a values (limited dynamic range) in some combinations
Recent developments
5
Appears that seasons can be lumped
0 50 100 150 200 250 300 350 4000
100
200
300
400
500
600
700
R² = 0.815755152261404
R² = 0
SpringLinear (Spring)Summer
YSI Chla (ug/L)
Extr
acte
d C
hla
(u
g/L
)
6
TF different, but little evidence that other segments are fundamentally different
0 50 100 150 200 2500
20
40
60
80
100
120
140
160
180
R² = 0.486964803348307
R² = 0.747940161656906R² = 0R² = 0
ELIMH
Linear (ELIMH)
JMSMH
YSI Chla (ug/L)
Extr
acte
d C
hla
(u
g/L
)
7
Some years are different
0 50 100 150 200 250 3000
50
100
150
200
250
300
350
2005Linear (2005)2006
YSI Chla (ug/L)
Extr
acte
d C
hla
(u
g/L
)
8
Log transform improves some regression properties but can contribute to unreasonable corrections
-2 -1 0 1 2 3 4 5 60
1
2
3
4
5
6
7
2005
Lin-ear (2005)
2006
Ln YSI Chla
Ln
Extr
acte
d C
hla
9
• Use consistent correction method• Remove outliers• Lump non-TF segments by year• Simple linear regression• These regressions have already been
developed by HRSD.
Recommendation: Annual Correction Method
10
Dataflow Processing Output:Maps
Mesohaline
Polyhaline
…can be made “prettier” using GIS
11
• By segment, individual cruise• Spatial exceedance, no temporal averaging
• By segment-season• Averaged spatially and temporally
• Can perform CFD calculations
Dataflow Processing Output:Tables of Threshold Exceedance Rates
Cruise Date Segment Season YearArithmetic Mean Chla
Geometric Mean Chla
% Area Exceed 10
ug/L
% Area Exceed 20
ug/L
% Area Exceed 30
ug/L
% Area Exceed 40
ug/L
% Area Exceed 50
ug/L
% Area Exceed 60
ug/L
% Area Exceed 70
ug/L
% Area Exceed 80
ug/L
% Area Exceed 90
ug/L
% Area Exceed
100 ug/L03/02/2010 JMSMH Spring 2010 6.99 6.30 9.7 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.003/05/2013 JMSMH Spring 2013 18.57 6.99 38.3 25.4 20.3 15.1 12.0 10.4 9.1 8.8 7.6 5.903/06/2006 JMSMH Spring 2006 74.76 42.60 96.5 67.5 57.4 50.8 47.4 43.3 39.2 35.2 32.2 29.203/06/2008 JMSMH Spring 2008 5.42 4.40 14.2 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.003/07/2012 JMSMH Spring 2012 27.11 14.86 63.3 39.2 25.6 19.3 17.3 16.5 11.6 11.6 8.5 6.303/08/2006 JMSMH Spring 2006 36.40 8.12 36.9 22.1 19.0 16.6 14.8 14.0 13.2 13.2 13.2 12.803/08/2007 JMSMH Spring 2007 3.79 3.60 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.003/08/2010 JMSMH Spring 2010 5.09 4.74 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.003/08/2011 JMSMH Spring 2011 3.57 2.92 1.5 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.003/10/2009 JMSMH Spring 2009 18.49 9.60 36.7 19.2 16.8 14.9 10.0 8.2 7.6 6.4 6.4 4.1
12
Dataflow Processing Output:Charts of %Area by Chl-a Bin
2005 2006 2007 2008 2009 2010 2011 2012 20130%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
JMSMHSummer
Mean % Area > 100 ug/LMean % Area 90 - 100 ug/LMean % Area 80 - 90 ug/LMean % Area 70 - 80 ug/LMean % Area 60 - 70 ug/LMean % Area 50 - 60 ug/LMean % Area 40 - 50 ug/LMean % Area 30 - 40 ug/LMean % Area 20 - 30 ug/LMean % Area 10 - 20 ug/LMean % Area 0 - 10 ug/L
14.5 7.1 8.2 19.9 9.17 17.3 6.1 10.3 9.3 Seasonal mean
13
Dataflow Processing Output:Graphics of Mean vs. Threshold Exceedance Rates
0 5 10 15 20 25 300.0
5.0
10.0
15.0
20.0
25.0
% Area Exceeding 80 ug/LJMSMH Summer - 2005-2013 - Individual Cruises
Geometric Mean Chl-a (ug/L)
% A
rea
Exce
ed
ing
80
ug
/L
14
• Conditional probability• Chl-a bin charts are a version of this• Some bins are data limited• Inter-bin thresholds?
• Potential supplementary analyses• Simple graphical (e.g., scatterplots)• Locally-weighted averaging (LOWESS)• Continuous conditional probability methods• Changepoint analysis
Other Potential Graphical/Statistical Analysis Methods
15
Scatterplot/LOWESS Example: Cocholodinium density vs. chlorophyll-a
16
Scatterplot/LOWESS Example: C. polykrikoides density (transformed) vs. chlorophyll-a
10,000 cells/mL
17
Locally-weighted conditional probability example: Probability that C. polykrikoides exceeds 10k cells/mL vs. chlorophyll-a
18
Changepoint analysis example: Probability of C. polykrikoides > 10k cells/mL split by chlorophyll-a
Significantly higher probability when chla > 100 ug/L
Brown and Caldwell 19
Extra Slides
Brown and Caldwell 20
2005 2006 2007 2008 2009 2010 2011 2012 20130%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
JMSMHSpring
Mean % Area > 100 ug/LMean % Area 90 - 100 ug/LMean % Area 80 - 90 ug/LMean % Area 70 - 80 ug/LMean % Area 60 - 70 ug/LMean % Area 50 - 60 ug/LMean % Area 40 - 50 ug/LMean % Area 30 - 40 ug/LMean % Area 20 - 30 ug/LMean % Area 10 - 20 ug/LMean % Area 0 - 10 ug/L
Brown and Caldwell 21
2005 2006 2007 2008 2009 2010 2011 2012 20130%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
JMSPHSpring
Mean % Area > 100 ug/LMean % Area 90 - 100 ug/LMean % Area 80 - 90 ug/LMean % Area 70 - 80 ug/LMean % Area 60 - 70 ug/LMean % Area 50 - 60 ug/LMean % Area 40 - 50 ug/LMean % Area 30 - 40 ug/LMean % Area 20 - 30 ug/LMean % Area 10 - 20 ug/LMean % Area 0 - 10 ug/L
Brown and Caldwell 22
2005 2006 2007 2008 2009 2010 2011 2012 20130%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
JMSPHSummer
Mean % Area > 100 ug/LMean % Area 90 - 100 ug/LMean % Area 80 - 90 ug/LMean % Area 70 - 80 ug/LMean % Area 60 - 70 ug/LMean % Area 50 - 60 ug/LMean % Area 40 - 50 ug/LMean % Area 30 - 40 ug/LMean % Area 20 - 30 ug/LMean % Area 10 - 20 ug/LMean % Area 0 - 10 ug/L