RiverSonde - Sensing
Humans have relied on fresh and saltwater rivers for transportation, clean drinking water, irrigation, and food since the earliest recorded history. Compared to oceans and other large bodies of water, however, rivers are significantly understudied, especially with respect to the effects of climate change. In an effort to close this research gap, we have set out to develop the RiverSonde, a low profile underwater sensor suite that enables inexpensive remote collection of water quality data from rivers and streams. We’ve designed RiverSonde to be hydroelectrically powered by a river's natural current, making it deployable in low sunlight areas. We've also made it mesh networking ready, meaning that a multiple sonde network can be deployed to take remote measurements even in areas without cellular coverage. By implementing sensors and electronics that are typically only found in 3000 dollar plus laboratory probes while keeping our unit cost under 500 dollars, we hope to give budget constrained research and regulatory agencies a powerful new tool.
One of the fundamental challenges is the detection and measurement of Harmful Algal Blooms (HABs). Phytoplankton, cyanobacteria and algae have hundreds or thousands of different taxonomical categories and sub-categories and produce a diverse array of environmental and health hazards. Many potential species and their effects are either poorly understood or remain undiscovered. IOC-UNESCO has an ongoing effort to document and track the taxonomic information of identified species and provides a publicly accessible list indicating the potential scope of a detection challenge. Additionally, many bacteria or algae are functional parts of an ecosystem and are not necessarily harmful to ecosystem stability or human health and industry. Robustly making these distinctions in various geographical locations comprises the overarching challenge in HAB detection.
Modularity, affordability, accessibility, and coverage are key restrictions in approaching this problem. In addition to those design limitations, our clients the Department of Conservation and Recreation (DCR) and the Charles River Watershed Association have shown interest in collecting data on the following
| Metric | Rationale |
|---|---|
| Chlorophyll-a | Indicative of dispersed biomass |
| Phycocyanin | Indicative of cyanobacterial biomass |
| Phosphorous | Indicative of increased soil mass and erosion |
| Suspended Solids | Indicative of particulate pollution in river |
| Temperature | Generally useful parameter |
| Depth | Useful for sea level rise over time |
| E. Coli | Found with industrial or sewage runoff |
| Dissolved Oxygen | Indicative of ecological health |
| Microcystin-LR(cyanotoxin) or Microcysits (cyanobacteria type) | Most common and problematic form of algal bloom in the Boston area |
| BMAA | Cyanotoxin commonly found in Boston rivers |
| Other Cyanotoxins | Whatever is possible based on time and resources |
Chlorophyll Sensor
Our approach to designing the core sensor suite in RiverSonde is split into three parts. COTS sensors like thermocouples, humidity gauges (for leak detection inside the shell), pressure transducers and more are easily integrated and comprise the first part. Second, taking chlorophyll-a measurements can yield broad and potentially indicative information about algal blooms and biomass concentrations at given points in the river. Leeuw et. al. demonstrated a cheap DIY chlorophyll-a sensor which excites chlorophyll with a blue LED and measures the intensity of the emitted light to calculate a chlorophyll concentration near the surface of the sensor. We integrated this approach into the hull of our platform. Finally, for the third portion, we considered the remaining metrics the DCR and CRWA asked us to attempt to measure.
DIY Cytometer
With one measurement out of chlorophyll and phycocyanin successful, the next consideration was if it was possible to directly evaluate either HAB toxins or bacterial species. Since toxins were smaller and more chemically challenging to evaluate given our constraints, we focused on considering approaches to bacterial or algal species. A key capability of our approach needed to be a capacity to take in situ measurements of our target molecules. Flow cytometry is one approach that has been implemented on a DIY scale that could fit our needs and take in situ measurements.
Benchtop approaches to flow cytometry typically cost several thousand dollars and are not useful for taking remote in situ measurements and process samples at a slow rate. A team at UCLA demonstrated a 2500 dollar DIY cytometer that had a relatively high throughput (100 mL/hr). While that’s an order of magnitude cheaper than traditional cytometers, it was still above our budget. Also, the cytometer relied on computer vision on captured images to identify target organisms. Traditional computer vision is computationally expensive and consumes large amounts of power, well beyond our platform's rated capacity.
In order to address the computation and power consumption issues, we turned to a subsampling technique called single pixel imaging. Subsampling techniques work by taking a fraction of a potentially available dataset and either reconstructing the remaining dataset from the small sample or acting on the fractional sample to perform a task. Single pixel imaging is a technique where an image is captured by a single pixel over a (relatively) long period of exposure. With a single pixel camera, a small fraction of the relevant pixels can be captured until enough information is available to identify the target organisms in the field of view. This drastically cuts the amount of information needed to be processed, bringing the processing and power requirements in line with what could be supported by RiverSonde. We currently are building and testing a benchtop version of this hybrid flow cytometer/single pixel camera to measure the microcystin and potentially E. Coli requested by our clients. Additionally, such a setup is also modular enough to be adapted to other organisms based on the ecosystem specifics.
Channel Design
The first driving component in the cytometer design is the microfluidic channel. The channel dimensions directly relate to the volumetric flow rate of the channel and the velocity of the fluid inside the channel. The channel dimensions are co-dependent on the additive manufacturing constraints, the sampling rate of the sensor and a mechanically feasible flow rate determined by the limitations of the pump. The channel design should take these variables and factors into account and seek to minimize the distance travelled between each sample.
Pump Design
A custom benchtop syringe pump was used to generate the pressure differential through the microfluidic channel. The above image is an example of a pump assembled using printed and COTS components. This approach could concievably be integrated into the 3D printed structure of the Sonde which is why it was used for the benchtop demonstration assembly. Source for original design
Optics Design
A Digital Micromirror Device (DMD) was used to construct a single pixel camera for the imaging sensor. A single pixel camera has several advantages in this application over a traditional CMOS or CCD chip:
- More effective for sub-sampling approaches. Sub-sampling, or undersampling, is a signal processing technique where a fraction of the available signal is processed to either reconstruct or act upon the original signal. This saves significant computational resources in computer vision.
- Can easily select for and focus on regions of interest by turning on mirrors corresponding to field of view.
- Better suited for Super Resolution (SR) imaging where an image is developed via longer exposure that resolves smaller than the size of the pixels (mirrors in this case)
Under Construction
Algorithmic Layout
Project still in progress, check back for further updates.
Selected References
- Zhao, Y., Chen, Q., Sui, X., & Gao, H. (2017). Super Resolution Imaging Based on a Dynamic Single Pixel Camera. IEEE Photonics Journal, 9(2), 1–11. https://doi.org/10.1109/JPHOT.2017.2688351
- Gӧrӧcs, Z., Tamamitsu, M., Bianco, V., Wolf, P., Roy, S., Shindo, K., … Ozcan, A. (2018). A deep learning-enabled portable imaging flow cytometer for cost-effective, high-throughput, and label-free analysis of natural water samples. Light: Science & Applications, 7(1), 66. https://doi.org/10.1038/s41377-018-0067-0
- Phillips, D. B., Sun, M.-J., Taylor, J. M., Edgar, M. P., Barnett, S. M., Gibson, G. G., & Padgett, M. J. (2016). Adaptive foveated single-pixel imaging with dynamic super-sampling. ArXiv:1607.08236 [Physics]. Retrieved from http://arxiv.org/abs/1607.08236