In the present case study, the use of an advanced, efﬁcient and low-cost technique for monitoring an urban stream was reported. Physicochemical parameters (PcPs) of Jungnangcheon stream (Seoul, South Korea) were assessed using an Internet of Things (IoT) platform. Temperature, dissolved oxygen (DO), and pH parameters were monitored for the three summer months and the ﬁrst fall month at a ﬁxed location. Analysis was performed using clustering techniques (CTs), such as K-means clustering, agglomerative hierarchical clustering (AHC), and density-based spatial clustering of applications with noise (DBSCAN).
An IoT-based Arduino sensor module (ASM) network with a 99.99% efﬁcient communication platform was developed to allow collection of stream data with user-friendly software and hardware and facilitated data analysis by interested individuals using their smartphones. Clustering was used to formulate relationships among physicochemical parameters. K-means clustering was used to identify natural clusters using the silhouette coefﬁcient based on cluster compactness and looseness. AHC grouped all data into two clusters as well as temperature, DO and pH into four, eight, and four clusters, respectively.
DBSCAN analysis was also performed to evaluate yearly variations in physicochemical parameters. Noise points (NOISE) of temperature in 2016 were border points(p), where as in 2014 and 2015 they remained core points(q) indicating a trend toward increasing stream temperature. We found the stream parameters were within the permissible limits set by the water quality standards for river water, South Korea.
The sample site platform was located in the Jungnangcheon stream of the Han River in Haengdang-dong, Seongdong-gu, as shown in Figure 1. This urban stream passes through the center of Seoul, a major metropolitan city in South Korea. The ASM was used to measure discreet parameters at a static location every second during three summer months and the ﬁrst fall month for three consecutive years (2014, 2015, and 2016).
The experiment was designed to monitor the 24-h behavior of the stream on week days during summer and fall for three years (2014 to 2016). Instruments were placed downstream and anchored to the ﬁrst pier for safety. A detailed schematic diagram of the ASM frame is provided in Figure 2a. The ASM instrument was placed in a porous galvanized steel strainer (70 cm length, 8.9 cm diameter, 3.35 kg weight) with a galvanized zinc cap (10 cm length). The holes of the strainer were 3 mm in diameter. Figure 2b,c show the strainer (bracket for sensor) ﬁxed vertically to the pile with two bolts for protection against wave action. The strainer was horizontally ﬂanged on the pile cap to keep the instrument safe from external impacts.
RESULTS AND DISCUSSIONS
Figure 5. The DO concentration increased smoothly with an average difference in concentration of 0.22 mg/L between midnight and dawn. The drop-off in DO concentration was high during the day and low at night. The pH of the stream (7.63) was slightly acidic. A decrease of 0.10 decrease in pH was observed at midnight and an increase of 0.17 at midday.
We used the K-means algorithms to cluster the raw data obtained from the sensors. Separate PcP clusters are shown in Figure 6a. This ﬁgure shows the partition of the raw data into three clusters. Since we collected the data over the speciﬁed study period, we clustered each PcP value into a sub-cluster. Figure 6b shows the raw temperature datasets. In this ﬁgure, it is observed that cluster 2 is well classiﬁed as the values on the x-axis are higher, followed by clusters 3 and 1.
In this study, we developed the IoT-based Arduino platform for continuous monitoring of Jungnangcheon stream. Monitoring was conducted for collected data every second every day for ﬁve days a week for four months (June, July, August, September) in 2014, 2015, and 2016.
(1) Three parameters, temperature, DO, and pH, were measured at a ﬁxed location with 99.99% efﬁciency using an IoT Arduino platform. Simpliﬁed information was provided to residents (end users) on their smartphones. Hence, the proposed IoT platform is highly efﬁcient and reliable in data transmission.
(2) AHC analysis segmented all data into two clusters, temperature into four clusters, DO into eight clusters, and pH into four clusters. AHC did not provide signiﬁcant results;however, the optimal time for monitoring individual samples was identiﬁed allowing for a reduction in the number of sampling sites.
(3) We performed monitoring using the IoT prototype based on the Arduino shield, only installed a handful of sensors, and only monitored conditions over a period of four months. However, the results indicated this application can help identify seasonal behavior and efﬁciently monitor PcPs in a low-cost manner.
Source: Hanyang University
Authors: Byungwan Jo | Zafar Baloch