HomeMy WebLinkAboutNSTEPS NC CyAN Analysis_v3.0_02922023Analysis of CyAN Remote Sensing Data for
North Carolina Coastal Waters under Nutrient
Scientific Technical Exchange Partnership
Support (N-STEPS)
February 2023
v. 3.0
Prepared for:
U.S. Environmental Protection Agency
Office of Science and Technology,
Health Ecological Criteria Division
1200 Pennsylvania Avenue, NW
Washington, DC 20460
Prepared by:
Tetra Tech, Inc.
1 Park Drive, Suite 200
Research Triangle Park, NC 27709
NC CyAN Report
January 2023 Page | 2
Contents
1 Introduction...........................................................................................................................................................4
2 Methods................................................................................................................................................................4
2.1 Estuary Data ...............................................................................................................................................4
2.2 Imagery Data ..............................................................................................................................................5
2.3 CyAN Processing .......................................................................................................................................6
2.4 CyAN Metrics and Thresholds....................................................................................................................7
2.5 Analysis ......................................................................................................................................................9
3 Results .............................................................................................................................................................. 10
3.1 Frequency – Status and Trends .............................................................................................................. 10
3.1.1 Boxplots ......................................................................................................................................... 10
3.1.2 Heatmaps ...................................................................................................................................... 15
3.1.3 Time Series.................................................................................................................................... 19
3.2 Magnitude – Status and Trends .............................................................................................................. 21
3.2.1 Boxplots ......................................................................................................................................... 21
3.2.2 Heatmaps ...................................................................................................................................... 22
3.2.3 Time Series.................................................................................................................................... 23
3.3 Extent – Status and Trends ..................................................................................................................... 25
3.3.1 Boxplots ......................................................................................................................................... 25
3.3.2 Heatmaps ...................................................................................................................................... 27
3.3.3 Time Series.................................................................................................................................... 30
3.4 Trend Analysis ......................................................................................................................................... 32
4 Conclusions ....................................................................................................................................................... 33
5 References ........................................................................................................................................................ 35
NC CyAN Report
January 2023 Page | 3
Figures
Figure 1. Map of the estuaries used for this study .....................................................................................................5
Figure 2. Illustration of a given waterbody (A), resolvable pixels for that waterbody (B), and classification of
the DN values into threshold values (C). Schaeffer et. al. 2022.................................................................7
Figure 3. Illustration of the temporal (A) and spatial (B) dimensions of CyAN data. Coffer et al. 2021. ....................8
Figure 4. Boxplots of No Bloom Frequency data. Estuaries are ordered by their median value. ........................... 10
Figure 5. Boxplots of High Bloom Frequency data. Estuaries are ordered by their median value. ........................ 11
Figure 6. Boxplots of Total Bloom Frequency data. Estuaries are ordered by their median value. ........................ 12
Figure 7. Boxplots of High Bloom Frequency data for the growing season. Estuaries are ordered by their
median value. .......................................................................................................................................... 13
Figure 8. Boxplots of frequency threshold data for Chowan River. ......................................................................... 14
Figure 9. Boxplots of frequency threshold data for Alligator River. ......................................................................... 14
Figure 10. Boxplots of frequency threshold data for Little River. ............................................................................ 15
Figure 11. Heatmap of the No Bloom Frequency data. Color denotes annual averaged value from 0 (green)
to 1 (red). ................................................................................................................................................. 16
Figure 12. Heatmap of the High Bloom Frequency data. Color denotes annual averaged value from 0 (green)
to 1 (red). ................................................................................................................................................. 17
Figure 13. Heatmap of the Total Bloom Frequency data. Color denotes annual averaged value from 0
(green) to 1 (red). ..................................................................................................................................... 18
Figure 14. Heatmap of the High Bloom Frequency data for the growing season. Color denotes annual
averaged value from 0 (green) to 1 (red). ................................................................................................ 19
Figure 15. Time series plots of Total Bloom Frequency data for each estuary. ...................................................... 20
Figure 16. Boxplots of magnitude data. Estuaries are ordered by their median value. .......................................... 21
Figure 17. Boxplots of magnitude data for the growing season. Estuaries are ordered by their median value. ..... 22
Figure 18. Heatmap of the magnitude-normalized data. Color denotes annual averaged value. ........................... 23
Figure 19. Time series plots of magnitude-normalized data for each estuary. ....................................................... 24
Figure 20. Boxplots of No Bloom Extent data. Estuaries are ordered by their median value. ................................ 25
Figure 21. Boxplots of High Bloom Extent data. Estuaries are ordered by their median value. ............................. 26
Figure 22. Boxplots of Total Bloom Extent data. Estuaries are ordered by their median value. ............................. 27
Figure 23. Heatmap of the No Bloom Extent data. Color denotes annual averaged value from 0 (green) to 1
(red). ........................................................................................................................................................ 28
Figure 24. Heatmap of the High Bloom Extent data. Color denotes annual averaged value from 0 (green) to 1
(red). ........................................................................................................................................................ 29
Figure 25. Heatmap of the Total Bloom Extent data. Color denotes annual averaged value from 0 (green) to
1 (red). ..................................................................................................................................................... 30
Figure 26. Time series plots of Total Bloom Extent data for each estuary.............................................................. 31
Tables
Table 1. Surface area of the estuaries .......................................................................................................................5
Table 2. CyAN threshold values .................................................................................................................................7
Table 3. List of CyAN metric–threshold variables ......................................................................................................9
Table 4. Seasonal Mann-Kendall significant results for No Bloom thresholds ........................................................ 32
Table 5. Seasonal Mann-Kendall significant results for High Bloom thresholds ..................................................... 32
Table 6. Seasonal Mann-Kendall significant results for Total Bloom thresholds .................................................... 32
Table 7. Seasonal Mann-Kendall significant results for Magnitude-Normalized ..................................................... 33
NC CyAN Report
January 2023 Page | 4
1 INTRODUCTION
The Nutrient Scientific Technical Exchange Partnership & Support (N-STEPS) program was created by the U.S.
Environmental Protection Agency (USEPA) in 2005 to serve as a technical and scientific resource for numeric
nutrient criteria development efforts for states, territories, and authorized tribes. The program is intended to
provide technical assistance to water quality scientists who are working to develop numeric nutrient criteria to
protect the designated uses of their state, territorial, or tribal surface waters. N-STEPS has developed materials,
tools, and offered technical assistance for all stages of numeric nutrient criteria development – i.e., planning, data
preparation and management, data exploration, analysis and model development, scientific literature review, and
peer review.
The North Carolina Department of Water Resources (NCDWR) is working to develop numeric nutrient criteria for
the Chowan River/Albemarle Sound (CR/AS) in accordance with the state’s Nutrient Criteria Development Plan
(NCDP). Waters for which criteria are being developed include the open waters of the CR/AS designated primarily
as Class SB (primary recreation, 15A NCAC 02B .0222) and inland waters of the Chowan River watershed.
These waters currently have numeric criteria for chlorophyll a and dissolved oxygen (see 15A NCAC 02B .0220).
This NSTEPS effort will be used to help the NCDWR and its associated Scientific Advisory Council (SAC)
establish appropriate tools and indicators of nutrient effects to derive numeric nutrient criteria concentrations that
protect beneficial uses in the CR/AS.
NSTEPS previously worked with NCDWR to support criteria development efforts in Albemarle Sound in 2015.
That effort developed statistical classifications of the open sound waters and the adjacent tributaries based on
differences in salinity, depth, residence time, and turbidity. The results of that effort are captured in the Albemarle
Sound Classification and Analysis conducted under the Nutrient Scientific Technical Exchange Partnership
Support (March 14, 2015).
NCDWR has also worked previously with NASA to compare satellite imagery of chlorophyll a concentrations with
in-situ measurements collected by NCDWR. This comparison generated representative images for monthly
means of chlorophyll a concentration for the study area. This work showed low correlation between the imagery
and in-situ measurements. However, advances in satellite algorithms have produced improved tools for remotely
sensing chlorophyll and cyanobacteria in this area.
This NSTEPS effort will leverage information gained through the above-mentioned studies, as well as additional
data available since the last NSTEPS work, namely remotely sensed cyanobacterial biomass imagery developed
through the Cyanobacteria Assessment Network (CyAN; https://www.epa.gov/water-research/cyanobacteria-
assessment-network-cyan), to estimate the status and trends in cyanobacteria biomass over the period of
available image record for the CR/AS and adjacent estuaries in North Carolina (NC).
2 METHODS
2.1 Estuary Data
Seven estuaries located in northeast NC were used for this study. Albemarle Sound and adjacent estuaries were
analyzed. Estuary delineations were based on geographic information system (GIS) polygons provided by NC.
Delineations for the Chowan arm were extended eastward to the NC-32 bridge (Figure 1). Due to its small size,
Edenton Bay was merged into the Chowan River Extended spatial subunit (herein referred to as Chowan River in
this document). The range of estuary sizes spans two orders of magnitude, from Little River (12 sq km) to
Albemarle Sound (1,024 sq km) (Table 1).
NC CyAN Report
January 2023 Page | 5
Figure 1. Map of the estuaries used for this study
Table 1. Surface area of the estuaries
Estuary Area (sq km)
Albemarle Sound 1,024
Chowan River Extended 208
Alligator River 178
Pasquotank River 59
North River 36
Perquimans River 23
Little River 12
2.2 Imagery Data
The Cyanobacteria Assessment Network (CyAN; https://www.epa.gov/water-research/cyanobacteria-assessment-
network-cyan)) is a multi-agency project among the USEPA, the National Aeronautics and Space Administration
(NASA), the National Oceanic and Atmospheric Administration (NOAA), and the United States Geological Survey
NC CyAN Report
January 2023 Page | 6
(USGS). The goal of CyAN is to support management and use of US lakes and estuaries by supporting the
detection and quantification of cyanobacteria algal blooms.
CyAN processes satellite imagery data across the US to detect cyanobacteria. Imagery data for CyAN has been
collected from various satellites across the years. Images from the from the European Space Agency (ESA)
Medium Resolution Imaging Spectrometer (MERIS) satellite sensor are from 2002 to 2012. Images from ESA’s
Ocean and Land Color Imager (OLCI) satellite sensor are from 2016 to present.
Images used by CyAN were downloaded as tag image file format (.tiffs) files from the NASA Ocean Color Web
website (NASA 2018; https://oceancolor.gsfc.nasa.gov/projects/cyan/). The estuaries of interest for this project
are located in CyAN tile 8-3. CyAN data are published in daily, 7-day, or 14-day versions, where 7-day and 14-
day files contain the 7- and 14-day maximum values for each cell, respectively. For this project, 7-day files from
January 2008 to February 2022 were downloaded. There was a four-year gap in CyAN imagery data from April
2012 to April 2016, due to the gap between the end of the MERIS and the start of the OLCI satellite sensors. In
total, 527 image files were downloaded.
The image data consist of 300m x 300m cells (pixels). Data from the raw images were processed using an
algorithm calibrated to detect cyanobacteria (Lunetta et. al. 2015). The result was a “digital number” (DN) value
for each pixel. The DN range for detectable cyanobacteria values is from 1 to 253, with values representing color
intensity. A DN value of 0 denotes no cyanobacteria detected (a nondetect). DN values of 254 and 255 denote
either no data or land/cloud masking (i.e., interference from land or clouds).
DN values were converted to cyano index (CI) values, which translates to an approximate cell density estimate
(cells/mL) of cyanobacteria:
𝐶𝐼=10 3
250∗𝐶𝐿−4.2
Research from Wynn and others (2021) identified and quantified a bias between the older MERIS and newer
OLCI data. Data from the OLCI images were found to be smaller in magnitude than MERIS data. Therefore, an
adjustment factor of 1.06 was applied to the OLCI CI values. This is equivalent to applying a 2.11 bias adjustment
to OLCI DN data:
𝐶𝐼𝐿𝐶𝑅𝐼𝑅=1.06 ∗𝐶𝐼𝐿𝐿𝐶𝐼
𝐶𝑀𝐿𝐶𝑅𝐼𝑅=𝐶𝑀𝐿𝐿𝐶𝐼+250
3 ∗𝑙𝑙𝑎10(1.06)
𝐶𝑀𝐿𝐶𝑅𝐼𝑅=𝐶𝑀𝐿𝐿𝐶𝐼+2.11
OLCI DN values of zero were retained as zero for this analysis.
2.3 CyAN Processing
After downloading from the CyAN website, pixels were clipped using the estuary polygons. To remove any
potential land masking issues due to a pixel overlapping both the shoreline and the estuary, a negative buffer
equal to the diagonal of the raster cell size (425 m) was applied to the estuary polygons (refer to Figure 2 (B)
NC CyAN Report
January 2023 Page | 7
below). This shrinks the estuary polygons, but ensures only pixels fully overlaying water were used in the
analysis.
2.4 CyAN Metrics and Thresholds
The amount of CyAN data is vast. For example, the Albemarle Sound contains over 11,000 pixels. Each pixel has
its own value updated weekly with each new image. To help summarize these data, USEPA developed CyAN
metrics that can be calculated for a given waterbody across a set timeframe. These metrics are referred to as
temporal frequency, magnitude, and spatial extent.
Before defining CyAN metrics, it is important to discuss the CyAN thresholds. Simply put, CyAN thresholds are a
binned version of the DN data. CyAN has outlined five threshold bins representing a gradient in cyanobacterial
densities (Table 2). The Low/Medium/High classification system is similar in framework to the World Health
Organization (WHO) 1999 cyanobacteria guidelines for freshwater (https://www.epa.gov/cyanohabs/world-health-
organization-who-1999-guideline-values-cyanobacteria-freshwater).
Table 2. CyAN threshold values
DN Values Threshold Event
0 No Bloom
[1, 29] Low Bloom
[30, 99] Medium Bloom
[100, 253] High Bloom
>1 Total Bloom
Square brackets denote “exclusive” range
Note that “Total Bloom” is a summation of the number of Low, Medium and High Bloom events in a given spatial
unit over a given time period. The thresholds and metrics work together. For example, we can calculate the “High
Bloom Frequency” value for a given waterbody. To calculate this value, we use the binary “Yes/No” data from
each pixel, as determined by the High Bloom cutoff threshold (illustrated in Figure 2).
Figure 2. Illustration of a given waterbody (A), resolvable pixels for that waterbody (B), and classification of the
DN values into threshold values (C). Schaeffer et. al. 2022.
NC CyAN Report
January 2023 Page | 8
CyAN data are spatiotemporal data. As such, CyAN metrics are aggregations of data across three dimensions.
Each image raster spans two spatial dimensions – latitude and longitude. The third dimension – time – is
comprised of the layers of image files, where each file represents another point in time (Figure 3). Data from
these three dimensions are aggregated for a given waterbody and set timeframe (e.g., monthly or annual
timesteps).
Figure 3. Illustration of the temporal (A) and spatial (B) dimensions of CyAN data. Coffer et al. 2021.
CyAN frequency measures the average number of cyanobacteria threshold events across a waterbody:
𝐶𝑞𝑎𝑞𝑞𝑎𝑙𝑎𝑥𝑖=𝑙𝑎𝑎𝑙(∑𝑀𝑞𝑙𝑀𝑎𝐶𝑞𝑎𝑙𝑞𝑞𝑖𝑖
∑𝑀𝑞𝑙𝑀𝑎𝑀𝑎𝑞𝑎𝑞𝑞𝑎𝑎𝑀𝑖𝑥𝑎𝑙𝑞𝑖𝑖
)
𝑖
where i is the pixel (spatial) index, j is the time (image) index, and k is the waterbody index. NumOfEvents is the
number of threshold events (e.g., No Bloom, High Bloom, etc.) for each pixel location i, summed across all the
images j. Likewise, NumOfObservedPixels is the number of observable pixels (pixels unobscured by land or
clouds) for each pixel location i, summed across each image j. The resulting quotient is a grid of values, with each
value representing the frequency of the threshold event for that pixel. Finally, the average of those pixels is
calculated for each waterbody. CyAN frequency values range from 0 to 1, exclusive. See Clark et. al. (2017) and
Coffer et. al. (2021) for additional information on the CyAN frequency metric.
CyAN magnitude measures the average intensity of cyanobacteria threshold events across a waterbody:
𝑀𝑎𝑎𝑙𝑖𝑞𝑞𝑎𝑎𝑖=𝑙𝑎𝑎𝑙(∑𝐶𝐼𝑖𝑖𝑖)𝑖
The subscripts are the same as CyAN frequency. The CI of each pixel is summed across each waterbody–image
combination. Then, the average across the images is calculated for each waterbody. CyAN magnitude values are
greater than or equal to zero. See Mishra et. al. (2019) for additional information on the CyAN magnitude metric.
NC CyAN Report
January 2023 Page | 9
CyAN extent measures the spatial extent of a cyanobacteria threshold event across a waterbody:
𝐶𝑥𝑞𝑎𝑙𝑞𝑖=𝑙𝑎𝑎𝑙(∑𝑀𝑞𝑙𝑀𝑎𝐶𝑞𝑎𝑙𝑞𝑞𝑖𝑖𝑖
∑𝑀𝑞𝑙𝑀𝑎𝑀𝑖𝑥𝑎𝑙𝑞𝑖𝑖𝑖
)
𝑖
The subscripts are the same as CyAN frequency. For extent, the number of threshold events for each pixel i are
summed across each Waterbody-Date combination. Likewise, the total number of pixels are summed across each
waterbody-image combination. The resulting quotient is the proportion of pixels with threshold events across the
waterbody-image combination. Finally, the average of those values is calculated for each waterbody. CyAN extent
is normalized by the waterbody size (via the denominator above) and ranges from 0 to 1, exclusive. See Urquhart
et. al. (2017)(and update in Schaeffer et al. 2022) for additional information on the CyAN extent metric.
CyAN metrics are calculated for a given time period or timestep. For this study, a monthly timestep with roughly 4
images (weekly) per timestep was used. Monthly resolution allows for the exploration and analysis of intra-annual
seasonality and growing season differences. Recall that CyAN metrics essentially convert the image data into
tabular data at the waterbody–timestep level. There were 847 observations, one for each estuary–year–month
combination (7 estuaries * ~10 years * 12 months = 847). Each estuary had 121 monthly observations. For each
observation, 11 metric–threshold combinations were calculated (Table 3).
Table 3. List of CyAN metric–threshold variables
Frequency Magnitude Extent
No Bloom Magnitude-Normalized No Bloom
Low Bloom - Low Bloom
Medium Bloom - Medium Bloom
High Bloom - High Bloom
Total Bloom - Total Bloom
Note. Magnitude may be calculated as-is or normalized by estuary area. This study reports magnitude-normalized
values.
2.5 Analysis
Processed data were visually explored using boxplots, heatmaps, and time series plots. Spatial and temporal
comparisons were made across estuaries and years. Analyses using both year-round and growing season
(defined as May to September) data were performed.
A seasonal Mann-Kendall (MK) trend test was performed. MK is a rank-based, non-parametric analysis to detect
monotonic trends across time (Mann 1945). MK is robust to outliers and has a long history of use in water
resources. MK can easily be extended to account for seasonality (Hirsch and Slack 1984). The seasonal MK
(SMK) was applied to data observed on/after 2016, due to the gap in satellite coverage before 2016.
GIS processing of estuarine polygons was performed using ArcGIS 10.7 (ESRI 2022). Data processing and
analyses were performed using R (R Core Team 2022). The “raster” package (Hijmans 2021) and the “rgdal”
NC CyAN Report
January 2023 Page | 10
package (Bivand et. al. 2021) were used for raster and polygon processing in R. The SMK test was calculated
using the “rkt” package (Marchetto 2021).
3 RESULTS
Results are provided for the three CyAN metrics. For brevity, this section focuses on No Bloom, High Bloom, and
Total Bloom. Additional output from all thresholds can be made available as supplemental material as desired.
3.1 Frequency – Status and Trends
3.1.1 Boxplots
In the boxplots below, each point represents a month of image data. The No Bloom frequency results (Figure 4)
show that Chowan River had the largest proportion of no bloom events (median = 93%). Conversely, Little River,
North River, and Pasquotank River had median values below 0.5. Recall that “No Bloom” means that the DN was
<1 (DN ranges from 0–253).
Figure 4. Boxplots of No Bloom Frequency data. Estuaries are ordered by their median value.
NC CyAN Report
January 2023 Page | 11
When viewing the High Bloom Frequency results (Figure 5), Little River had the largest median value (25%). The
medians of the other estuaries were less than 10%. The data are right-skewed, meaning there are a handful of
months that reach much larger frequencies, relative to most of the data.
Figure 5. Boxplots of High Bloom Frequency data. Estuaries are ordered by their median value.
The Total Bloom Frequency results show lower values for Chowan River, on average, compared to the other
estuaries (Figure 6). Chowan River extended had a median value of 7%, compared to the larger median values of
56%, 66%, and 71% for Pasquotank River, North River, and Little River (respectively). For all metrics, note that
the Total Bloom and No Bloom thresholds are complimentary, since “Total Bloom = 1 – No Bloom” and “No Bloom
= 1 – Total Bloom”.
NC CyAN Report
January 2023 Page | 12
Figure 6. Boxplots of Total Bloom Frequency data. Estuaries are ordered by their median value.
When focusing on the growing season, the inter-quartile range (75th – 25th percentile; the width of the boxes in the
boxplot) tended to be higher than the year-round results. However, the median value was not consistently higher
or lower but varied per estuary and threshold. For example, when looking at High Bloom Frequency (Figure 7), we
see a wider range of likely values across all estuaries. The median value for Little River increased from 25% to
41%, and from 7% to 15% for North River. However, median values for Albemarle Sound, Alligator River, Chowan
River, and Perquimans River remained virtually the same.
NC CyAN Report
January 2023 Page | 13
Figure 7. Boxplots of High Bloom Frequency data for the growing season. Estuaries are ordered by their median
value.
We can also compare the threshold values across a single estuary. Below we see that Chowan River had a large
proportion (93% median) of No Bloom values and a small proportion (7% median) of Total Bloom (Figure 8).
NC CyAN Report
January 2023 Page | 14
Figure 8. Boxplots of frequency threshold data for Chowan River.
In comparison, Alligator River had more spread across its thresholds, with lower No Bloom frequencies (68%
median), and larger Total Bloom frequencies (32% median) (Figure 9). Looking at Little River, we see lower No
Bloom frequencies (29% median) and larger Total Bloom frequencies (71% median) (Figure 10).
Figure 9. Boxplots of frequency threshold data for Alligator River.
NC CyAN Report
January 2023 Page | 15
Figure 10. Boxplots of frequency threshold data for Little River.
3.1.2 Heatmaps
Another way to visualize the data is with heatmaps, with the values averaged for each year. Recall that the gap
from 2013 to 2015 is due to a lack of CyAN satellite images from April 2012 to April 2016. Therefore, heatmap
data for 2012 and 2016 do not reflect a full 12-month period.
As in the boxplots, Chowan River had larger “No Bloom” values compared to other estuaries (Figure 11). With the
heatmap, we observe a general decrease in No Blooms (i.e., an increase in cyanobacteria) for most of the
estuaries. Values for 2021 appear to have improved slightly for Little River, North River, and Pasquotank River,
compared to 2020.
NC CyAN Report
January 2023 Page | 16
Figure 11. Heatmap of the No Bloom Frequency data. Color denotes annual averaged value from 0 (green) to 1
(red).
High Bloom Frequencies were generally low, with some medium-range values for Little River, North River, and
Pasquotank River from 2018 to 2020 (Figure 12). Total Bloom frequencies showed increased values for those
same three estuaries from 2018 to 2020 (Figure 13).
NC CyAN Report
January 2023 Page | 17
Figure 12. Heatmap of the High Bloom Frequency data. Color denotes annual averaged value from 0 (green) to 1
(red).
NC CyAN Report
January 2023 Page | 18
Figure 13. Heatmap of the Total Bloom Frequency data. Color denotes annual averaged value from 0 (green) to 1
(red).
Growing season frequencies appeared similar to the year-round frequencies. As with the boxplots, some but not
all estuaries showed higher growing season values compared to annual values (Figure 14).
NC CyAN Report
January 2023 Page | 19
Figure 14. Heatmap of the High Bloom Frequency data for the growing season. Color denotes annual averaged
value from 0 (green) to 1 (red).
3.1.3 Time Series
Time series plots of Total Bloom Frequency for all the estuaries is presented in Figure 15. The oscillating pattern
suggests seasonality, which is to be expected for these data. Patterns present in the boxplots and heatmaps were
also observed here, with Chowan River presenting lower values relative to the other estuaries. Note that the gap
from 2013-2015 is connected by a straight line.
NC CyAN Report
January 2023 Page | 20
Figure 15. Time series plots of Total Bloom Frequency data for each estuary.
NC CyAN Report
January 2023 Page | 21
3.2 Magnitude – Status and Trends
3.2.1 Boxplots
Little River magnitude values were the largest of the estuaries (0.0074 median), and Chowan River were the
smallest (0.0010 median) (Figure 16). The data are right-skewed, with each estuary having a few notably larger
months. When focusing on the growing season, the values are more spread out (Figure 17). Overall, the median
values did not shift much, with the largest shift being Little River (from 0.0074 to 0.0100).
Figure 16. Boxplots of magnitude data. Estuaries are ordered by their median value.
NC CyAN Report
January 2023 Page | 22
Figure 17. Boxplots of magnitude data for the growing season. Estuaries are ordered by their median value.
3.2.2 Heatmaps
Heatmaps of magnitude showed perhaps a slight increase over time for most estuaries (Figure 18). As with
frequency, 2021 appears to be improved compared to 2020. Much larger magnitude values were observed in
2019 for Little River and Pasquotank River. The growing season results (not shown) were very similar to the year-
round results
NC CyAN Report
January 2023 Page | 23
Figure 18. Heatmap of the magnitude-normalized data. Color denotes annual averaged value.
3.2.3 Time Series
Time series plots of magnitude for all the estuaries are presented in Figure 19. Patterns observed in the boxplots
and heatmaps were also present here, with large spikes in 2019 for Little River and Pasquotank River.
NC CyAN Report
January 2023 Page | 24
Figure 19. Time series plots of magnitude-normalized data for each estuary.
NC CyAN Report
January 2023 Page | 25
3.3 Extent – Status and Trends
3.3.1 Boxplots
The No Bloom extent boxplots shows that Little River, North River, and Pasquotank River had lower No Bloom
annual median extents compared to the other estuaries (Figure 20). Chowan River had the largest No Bloom
extent at 94%.
Figure 20. Boxplots of No Bloom Extent data. Estuaries are ordered by their median value.
Little River had the largest High Bloom event extent (24%), while the other estuary medians were under 10%
(Figure 21). Total Bloom values (Figure 22) show Pasquotank River, North River, and Little River with median
extent values larger than 50% (i.e., greater than 50% of pixels with blooms). In comparison, Chowan River’s
median Total Bloom extent was 6%.
NC CyAN Report
January 2023 Page | 26
Figure 21. Boxplots of High Bloom Extent data. Estuaries are ordered by their median value.
NC CyAN Report
January 2023 Page | 27
Figure 22. Boxplots of Total Bloom Extent data. Estuaries are ordered by their median value.
3.3.2 Heatmaps
No Bloom extents appeared to decrease slightly from for most estuaries (Figure 23). Chowan River and
Perquimans River had the largest values of No Bloom. Little River, North River, and Pasquotank River showed
relatively lower No bloom values from 2018 to 2020.
NC CyAN Report
January 2023 Page | 28
Figure 23. Heatmap of the No Bloom Extent data. Color denotes annual averaged value from 0 (green) to 1 (red).
Little River, North River, and Pasquotank River showed moderate levels of High Bloom Extent (Figure 24). The
other estuaries showed lower values across all the years. Similarly, Little River, North River, and Pasquotank
River showed increased Total Bloom Extent values (Figure 25). Chowan River showed consistently lower Total
Bloom values across the observed years.
NC CyAN Report
January 2023 Page | 29
Figure 24. Heatmap of the High Bloom Extent data. Color denotes annual averaged value from 0 (green) to 1
(red).
NC CyAN Report
January 2023 Page | 30
Figure 25. Heatmap of the Total Bloom Extent data. Color denotes annual averaged value from 0 (green) to 1
(red).
3.3.3 Time Series
Time series plots of Total Bloom Extent for all the estuaries is presented in Figure 26. Alligator River and
Perquimans River appear to show a small yet steady increase starting in 2016.
NC CyAN Report
January 2023 Page | 31
Figure 26. Time series plots of Total Bloom Extent data for each estuary
NC CyAN Report
January 2023 Page | 32
3.4 Trend Analysis
A SMK trend analysis was performed on data collected since 2016. Results are presented for No Bloom (Table
4), High Bloom (Table 5), Total Bloom (Table 6), and magnitude (Table 7). Non-significant results (p-value >0.05)
are labeled as “Flat”. Overall, we see increases in Total Bloom frequency and extent for Alligator River, Chowan
River, Little River, and Perquimans River. Of those, Alligator River and Perquimans River also showed an
increase in High Blooms and magnitude.
Table 4. Seasonal Mann-Kendall significant results for No Bloom thresholds
Estuary Frequency Extent
Albemarle Sound Flat Flat
Alligator River Decreasing Decreasing
Chowan River Decreasing Decreasing
Little River Decreasing Decreasing
North River Flat Flat
Pasquotank River Flat Flat
Perquimans River Decreasing Decreasing
Table 5. Seasonal Mann-Kendall significant results for High Bloom thresholds
Estuary Frequency Extent
Albemarle Sound Flat Flat
Alligator River Increasing Increasing
Chowan River Flat Flat
Little River Flat Flat
North River Flat Flat
Pasquotank River Flat Flat
Perquimans River Increasing Increasing
Table 6. Seasonal Mann-Kendall significant results for Total Bloom thresholds
Estuary Frequency Extent
Albemarle Sound Flat Flat
Alligator River Increasing Increasing
Chowan River Increasing Increasing
NC CyAN Report
January 2023 Page | 33
Estuary Frequency Extent
Little River Increasing Increasing
North River Flat Flat
Pasquotank River Flat Flat
Perquimans River Increasing Increasing
Table 7. Seasonal Mann-Kendall significant results for Magnitude-Normalized
Estuary Magnitude
Albemarle Sound Flat
Alligator River Increasing
Chowan River Flat
Little River Flat
North River Flat
Pasquotank River Flat
Perquimans River Increasing
4 CONCLUSIONS
The following conclusions were drawn from the analyses described above:
• The CyAN images provided a relative catalog of cyanobacterial assemblage status and trends in the
entire Chowan and adjacent Albemarle Sound system since 2008
• The CyAN images generated measures of cyanobacterial bloom frequency, magnitude, and extent for
each of the sub-estuarine areas
• The results indicate that the Chowan River sub-estuary has generally low total and high bloom frequency,
magnitude, and extent compared to other sub-estuaries, but that total cyanobacterial bloom frequency
has increased since 2016
• In contrast, the Little, North, and Pasquotank sub-estuaries show the highest frequency, magnitude, and
extent of high bloom and total bloom events across the region, followed by the central Albemarle Sound
• In terms of trends,
o the Alligator and Perquimans sub-estuaries increased in all three metrics – cyanobacteria bloom
frequency, magnitude, and extent since 2016;
NC CyAN Report
January 2023 Page | 34
o the Chowan and Little increased in cyanobacteria bloom frequency and extent since 2016;
o the North and Pasquotank did not show significant trends, but both sub-estuaries have had
among the highest cyanobacteria bloom frequency, magnitude, and extents in the region and this
did not change (decrease or increase)
NC CyAN Report
January 2023 Page | 35
5 REFERENCES
Bivand, R, T Keitt, and B Rowlingson. 2021. rgdal: Bindings for the 'Geospatial' Data Abstraction Library. R
package version 1.5-23. https://CRAN.R-project.org/package=rgdal.
Clark, JM, BA Schaeffer, JA Darling, EA Urquhart, JM Johnston, AR Ignatius, MH Myer, KA Loftin, PJ Werdell,
and RP Stump. 2017. Satellite monitoring of cyanobacterial harmful algal bloom frequency in recreational
waters and drinking water sources. Ecological Indicators. 80:84-95.
Coffer, MM, BA Schaeffer, WB Salls, E Urquhart, KA Loftin, RP Stumpf, PJ Werdell, and JA Darling. 2021.
Satellite remote sensing to assess cyanobacterial bloom frequency across the United States at multiple
spatial scales. Ecological Indicators. 128:107822.
ESRI 2022. ArcGIS Desktop: Release 11.7. Redlands, CA: Environmental Systems Research Institute.
Hijmans, RJ. 2021. raster: Geographic Data Analysis and Modeling. R package version 3.4-13. https://CRAN.R-
project.org/package=raster.
Lunetta, R., BA Schaeffer, RP Stumpf, D Keith, S Jacobs, and M Murphy. 2015. Evaluation of cyanobacteria cell
count detection derived from MERIS imagery across the eastern USA. Remote Sensing of the Environment.
Elsevier Science Ltd, New York, NY, 157(0):24–34.
Marchetto A. 2021. rkt: Mann-Kendall Test, Seasonal and Regional Kendall Tests. R package version 1.6,
https://CRAN.R-project.org/package=rkt.
Mishra, S, RP Stumpf, BA Schaeffer, PJ Werdell, KA Loftin, and A. Meredith. 2019. Measurement of
cyanobacterial bloom magnitude using satellite remote sensing. Scientific reports, 9(1), pp.1-17.
NASA Goddard Space Flight Center, Ocean Ecology Laboratory, Ocean Biology Processing Group. 2018. Sea-
viewing Wide Field-of-view Sensor (SeaWiFS) Ocean Color Data, NASA OB.DAAC. doi:
10.5067/ORBVIEW-2/SEAWIFS/L2/OC/2018.
R Core Team. 2022. R: A language and environment for statistical computing. R Foundation for Statistical
Computing,Vienna,Austria. https://www.R-project.org/.
Schaeffer, BA, E Urquhart, M Coffer, W Salls, RP Stumpf, KA Loftin, and PJ Werdell. 2022. Satellites quantify the
spatial extent of cyanobacterial blooms across the United States at multiple scales. Ecological Indicators, 140,
p.108990.
Urquhart, EA, BA Schaeffer, RP Stumpf, KA Loftin, and PJ Werdell. 2017. A method for examining temporal
changes in cyanobacterial harmful algal bloom spatial extent using satellite remote sensing. Harmful Algae.
67:144–152.