Paper: Eyes in the Sky: Estimating Inland Water Quality Using Landsat Data - #1111
Conversation
|
Curvenote Preview
|
dkedar7
changed the title
rowanc1
added
paper
|
@cdlindsey will serve as editor for this paper. |
fr1ll
left a comment
fr1ll
left a comment
There was a problem hiding this comment.
Well-written, clear work. The one place I got lost is how the WQA data is spatially indexed - by the name of the body of water, by lat/long, or other. If you clear this up, it will help to explain how the Landsat data ties. I also became curious how the data moves over time for any bodies with many measurements. The fact that these two questions came up indicates it's an interesting and accessible topic, clearly presented.
|
|
||
| Harmful algal blooms (HABs) can severely impact both the environment and human health. Environmentally, HABs deplete oxygen in water bodies, leading to fish kills and loss of aquatic biodiversity. They can block sunlight, disrupting aquatic plant growth and altering food webs. Many HABs produce toxins that accumulate in the ecosystem, affecting wildlife and domestic animals. For humans, exposure to these toxins—through drinking water, recreation, or consumption of contaminated fish and shellfish—can cause a range of health issues, including skin irritation, respiratory problems, gastrointestinal illness, and, in severe cases, liver or neurological damage. HABs also threaten water supplies and recreational activities, resulting in economic losses for affected communities. Therefore, detecting and mitigating HABs is an area of continued attention. | ||
|
|
||
| A key indicator of HABs is the concentration of Chlorophyll a (Chl-a) in water. Chl-a is a pigment found in all photosynthetic algae, and its presence is directly correlated with the abundance of algal biomass. Elevated Chl-a levels typically signal increased algal growth, which can indicate the onset or presence of a bloom. Therefore, monitoring Chl-a concentrations provides a reliable proxy for detecting and quantifying HABs in aquatic environments. |
There was a problem hiding this comment.
Consider adding a reference tying Chl-a to algal blooms. This could help an interested reader to understand more of the scientific underpinnings of your work.
| Values of Chl-a predicted by the best performance model plotted against actual values. | ||
| ::: | ||
|
|
||
| Furthermore, using Landsat-based satellite remote sensing insights is scalable and yields more than 30 times as many Chl-a readings as in-situ readings obtained from WQP (@fig:nycomparison). |
There was a problem hiding this comment.
By my understanding, the WQA dataset gives Chl-a "readings", while the LANDSAT data lets you predict Chl-a. I would suggest to re-phrase this sentence to indicate that these are predictions or inferences.
There was a problem hiding this comment.
Rephrased in the text as well as figure caption.
| or if you prefer LaTeX you can try `tabular` or `longtable` environments: | ||
| Querying the WQP for Chl-a measurements within the bounds of New York State between January 2015 and October 2023 yields a dataset of 68,075 measurements. WQP is not consistent in its nomenclature and data quality checks, so it was essential to specify different ways of referring to Chl-a. Some of those ways are _Chlorophyll_, _Chlorophyll A_, _Chlorophyll a_, _Chlorophyll a (probe relative fluorescence)_, _Chlorophyll a (probe)_, _Chlorophyll a - Periphyton (attached)_, _Chlorophyll a - Phytoplankton (suspended)_, _Chlorophyll a, corrected for pheophytin_, and so on. HyRiver, a Python library published by [](https://doi.org/10.21105/joss.03175), was used to connect to the WQP and export data. | ||
|
|
||
| To reduce the effect of repeated measurements from the same water body and around the same time, we deduplicate data by removing consecutive values with the same date and time if the relative standard deviation (RSD) of those values was less than 10. Passive remote sensing measurements like Landsat are not sensitive to measuring water quality beyond 100 meters depth. So, we removed all measurements that were reported to be measured beyond 100m in depth. After applying these filters and removing unclean values (where measurement values were not numerical), we are left with 13,717 measurements. |
There was a problem hiding this comment.
Since you will be joining SQA data and LANDSAT data, it would be helpful to specify what location information is present in each SQA measurement. Do the measurements feature a lat/long value, a named body of water, or something else?
There was a problem hiding this comment.
Added a few lines explaining how the spatial joining happens with stress on indexing fields: location (latitude, longitude) and dates.
|
|
||
| :::{figure} nycomparison.png | ||
| :label: fig:nycomparison | ||
| The number of remotely sensed Chl-a estimations (B) is more than 30x the number of in-situ measurements obtained from the Water Quality Portal (A) across NY’s 200 biggest lakes and in the same 8-year period (2015-2023). |
There was a problem hiding this comment.
This image would benefit from a key for the bubble size. Does each bubble represent the center of a body of water, and is the size indicating different measurements e.g. from different times? For large bodies of water, can there sometimes be more than one measurement from the same day, but from different locations?
There was a problem hiding this comment.
Added a bubble size legend and explanation about what bubble centers and sizes mean.
|
By the way, @dkedar7 - this is my first time doing one of these reviews, so please let me know if my review has any issues I can fix! |
|
Interesting work, and well-written paper @dkedar7 🙂 ! I had a few points for discussion and potential improvement.
|
|
Thank you for the comments. Based on the feedback, I'm listing the following action items to help us track.
|
|
@madhavmk My thoughts about your discussion points:
|
|
For the third point from the checklist, I attempted to find cases where there is sufficient data across WQP and Landsat-inferred values, but unfortunately, there aren't any. This is due to strict data cleaning steps. There are three cases (distinct lakes) where there are some overlapping points, but not enough to show a comparison trend plot. These are Greenwood Lake, Skaneateles Lake, and Owasco Lake. The overlaps can be seen using the accompanying web app: https://watersat-hpn4y2dvda-ue.a.run.app/ |
|
The updates look good to me ! |
|
@cdlindsey What should be our next steps here? |
fwkoch
added
approved
8 participants
Algal blooms threaten human health and aquatic ecosystems, making monitoring essential. While Chlorophyll-A (Chl-a) effectively indicates algal presence, laboratory analysis is complex. This study utilizes satellite imagery as an alternative, addressing previous research limitations caused by scarce lab data. Moreover, it also demonstrates how openly available Chl-a measurements obtained from the Water Quality Portal (WQP) can enable communities and organizations of all sizes to measure Chl-a in their waters without access to specialized in-situ water sampling skills or laboratory analysis equipment.
By combining the extensive WQP dataset with Landsat satellite imagery, these models estimate Chl-a levels in New York's inland waters. Training with eight years of data demonstrated a strong correlation between satellite-derived and actual measurements (MAPE: 0.96%; RMSE: 3.2 μg/L), enabling improved spatial and temporal monitoring capabilities.