Runoff estimation using digital image processing for residential areas

Water supply–demand has increased multiple times in the last decade and is expected to increase to 55% by 2050 (Wang et al. 2013; Haque et al. 2016). This demand is mainly due to the exponential growth of the population and the rapid expansion of and development of urban areas and industries. Moreover, the variability of rainfall in the spatio-temporal direction due to climate change has complicated the situation. Severe water shortages have been predicted by 2050 if no measures are taken to find new water supply sources (IFPRI 2012). Thus, there is an urgent need for water conservation to cater to the need in these changing times.

Studies of rainfall trend analysis have shown a decreasing trend in various cities worldwide (Amanatidis et al. 1993; Jain & Kumar 2012; Zhou et al. 2017). Thus, a continuation of this decreasing trend scenario will affect the required availability of freshwater, especially during the non-rainy seasons. Due to urbanization, a significant part of the land cover is transformed into imperviousness (Lindh 1983). Thus, during the rainy season in urban areas, infiltration is low due to a lack of pervious surface, and most of the rainfall flows as surface runoff. The higher surface runoff may cause flood situations where the intensity is high (Hafizi Md Lani et al. 2018). Also, the frequency of groundwater recharge becomes low due to urbanization (Foster 1999) which may cause scarcity of the available groundwater. Overexploitation of the groundwater reserves and studies have shown that nations' social and economic growth will be affected due to decreased groundwater availability (Kemper 2004; Garg & Hassan 2007).

Several studies have shown an inverse relationship between rainfall and temperature (Kothyari & Singh 1996; Sharma & Babel 2014; Issahaku et al. 2016). Also, an increase in the mean temperature on a global scale is reported by Parker et al. (2000). This increase in temperature is attributed to global warming; thus, a decrease in rainfall may be expected. Hence, proper planning and management of the available water resources should be carefully studied to negate any detrimental effects in the future. Also, rainwater harvesting (RWH) methods in such changing environment is a sustainable method to tackle water shortcomings in the future.

The rainwater interacts with the physical characteristics of the basin such that a part is infiltrated depending on the permeability of the surface, and the other part flows as the overland flow. The overland flow can be stored using a RWH system, which otherwise would flow out of the basin (Abdulla & Al-Shareef 2009). The RWH system has been successfully used in various countries to cater to the population's demands (Handia et al. 2003). Studies have also shown that the RWH system has been effectively used for agriculture (Falkenmark & Stockholm International Water Institute 2001). In an RWH system, the overland flow runoff can be trapped and stored. However, RWH is not limited to runoff storage alone; in residential areas, the rain that falls on the roof can also be collected. The RWH system helps decrease the dependency on the groundwater and reduce flood risks (Eckart et al. 2018; Qi et al. 2019; Huang et al. 2021).

Studies have shown the RWH system to be cost-efficient and more profitable if implemented on a large scale (Tam et al. 2010). Studies regarding RWH conducted in the water-scarce regions in Malaysia pointed to economic, social, and technical challenges. However, it was concluded to be a potential water resources alternative for the region (Lee et al. 2016). Other studies also suggest the RWH system as an efficient alternative for reducing the dependencies on fresh water in water-scarce regions (Sample & Liu 2014; Thomas et al. 2014; Morales-Pinzón et al. 2015). Haque et al. (2016) conducted a study on the impacts of climate change on RWH by selecting five locations in Australia. They concluded that the dry season affects the RWH system more than the wet season. Ward et al. (2012) studied the performance of a large-building RWH system that showed an average water-saving efficiency of 87% in 8 months. The findings show the significant potential in the savings of water and cost. In Jordan, rooftop RWH system studies by Abdulla & Al-Shareef (2009) showed considerable water saving; however, bacteriological parameters exceeded the limits required for potable water. The water from the RWH system can be used for various non-potable purposes. However, some studies showed that it could be treated so that it can be supplied for drinking and domestic purposes as well (Hartigan 2009; Helmreich & Horn 2009; Zhou et al. 2010; Al Qudah et al. 2012; Qi et al. 2019; Huang et al. 2021).

The RWH system studies show great potential in newly developed urban areas (Zabidi et al. 2020), thus drawing the interest of many in recent years (Farahbakhsh et al. 2009; Belmeziti et al. 2014; Vieira et al. 2014; Fonseca et al. 2017). Research studies have used the SCS-CN (Soil Conservation Service-Curve Number) method to estimate the runoff (Al-Ghobari & Dewidar 2021). In this method, the coefficient, CN, depends on factors like the soil type, land use, land cover, and antecedent soil moisture content. The rational formula is also widely used to compute the runoff for a catchment with known runoff coefficients based on the land use and land cover (LULC) (Biswas & Mandal 2014). Remote sensing and GIS (Geographic Information System) techniques are widely applied in urban environmental analysis for roof surface runoff estimation (Radzali et al. 2018; Norman et al. 2019) and to identify impervious surfaces and RWH sites (Forkuo 2013; Gaikwad 2015; Mahmoud & Alazba 2015; Ammar et al. 2016).

There have been very few studies that have used Google Satellite images for estimation of LULC and runoff coefficients. Medina et al. (2012) used high-resolution satellite imagery data from Google Maps in an automated fashion based on the fuzzy set classification to extract the runoff coefficients from satellite images and found promising results. Aher et al. (2014) applied the K-means clustering algorithm and textural parameters based on the gray-level co-occurrence matrix (GLCM) to classify the Google Earth images into land cover and land use sectors. They found that the K-means algorithm works well for classifying satellite images due to its excellent accuracy. Aung & Thant (2019) also applied the K-means clustering algorithm to classify Google Satellite images into three general classes: (1) building, (2) vegetation, and (3) road. Ekbote et al. (2017) used a correlation of the template image with the main image employing image normalized cross-correlation to estimate the green spots (tree) area.

The traditional method of computing the runoff coefficient is time-intensive as the LULC over the catchment is different and thus have different runoff coefficients. This process, if done manually, can make the computation even more complex. Thus, in this study, a framework is proposed to estimate the parameters using digital image processing techniques with the help of Google Satellite images.

The present study proposes a simple and novel method to estimate the runoff generation capacity of a limited area using Google Satellite images. It can be used to quickly and efficiently estimate the runoff from the residential areas and then use it to design RWH systems.

The MNNIT campus is cantered at 25.4920 °N, 81.8639 °E and houses approximately 6,000 residents. The campus residents are mostly dependent on the groundwater which is recharged during the monsoon season.

For the purpose of designing a RWH system for a small residential area, the watershed cannot be defined by the natural drainage, but by the available drainage network. The drainage network of the campus was obtained from the civil maintenance office, MNNIT, Allahabad. Based on the available drainage network of the institute, the four catchments are defined as shown in Figure 1.

The results of the image processing were compared with the actual runoff coefficients computed using the Arc-GIS.

Table 2 shows the details of the LULC in the second catchment. The total area of the catchment is 24,170.27 m² out of which barren land comprises 56.09% of land cover. The least percentage cover is 9.81% that includes the commercial areas. The forest cover and road coverage are 17.79 and 16.30%, respectively. The equivalent runoff coefficient for the second catchment is obtained as 0.3553.

Finally, the yield is computed for the third catchment. Table 3 shows the percentage of LULC coverage, where the total area is 51,000.63 m². In the third catchment, the percentage coverage of the commercial area is 39.99%, which is the maximum, followed by the forest area, which is 31.77% of the catchment area. The percentage coverage of barren land and roads is 14.22 and 14.00%, respectively. The equivalent runoff coefficient using the LULC is obtained as 0.4111 for the third catchment.

The comparison of the volume of water generated from the two methods is shown in the bar plots below in Table 4. It can be seen that there is less than 10% error in the image processing method by using the default calibration parameter (k=0.8).

Since the error is systematic, we estimated the runoff generated from image processing for k=0.9. The results for the case are shown in Table 5. It can be seen that average error is reduced to near 2% only.

Furthermore, the image processing method is easy and consumes less time in runoff estimation. On average, the time taken for capturing the screenshot, saving it, and running the code takes only less than 1 min for a catchment. However, in the conventional method, proper georeferencing is a crucial step that takes about 5–10 min. Then, marking different types of areas by polygons is also a time-consuming method, taking more than 1 h for the study area. Therefore, the time taken and efforts are significantly reduced in the proposed method without compromising the accuracy.

In this study, a quantitative comparison has been conducted to investigate the potential of the RWH system in the MNNIT, Allahabad campus in Prayagraj. The campus houses infrastructures like the administrative building, halls, etc., having a large rooftop area that has the potential to collect plenty of rainwater. Thus, the potential of setting an RWH system in the MNNIT campus was studied. Runoff estimation studies in the past are time-intensive as parameter estimation for an area is complex by the conventional method (Arc-GIS). A novel framework is developed to obtain the parameters by simple digital image processing (Google Satellite Image) which were then used to obtain the runoff. This framework is easy and consumes less time in runoff estimation; as the framework can be easily applied to any new area.

The conclusions of the study are as follows:

• The suggested methodology is an easy and fast method as compared to the conventional method for estimating the runoff coefficient for a residential catchment. The residential catchments are defined by the man-made drainage network, hence capturing the residential catchment is easier and can be done without any use of GIS software. However, for the natural catchments, the drainage network is defined based on a digital elevation model, which requires the use of GIS software for the catchment delineation.

• The sum of RGB pixels of the image obtained from Google image was found to be directly correlated with the runoff coefficient for easy calculation.

• Considering the maximum runoff coefficient as nearly 0.9, the default parameter was kept equal to 0.8.

• The entire study area was divided into three catchments. Using the conventional method (Arc-GIS), the equivalent runoff coefficients for these catchments were found to be 0.2780, 0.3553, and 0.4111, respectively. The range of error (compared to conventional method) in runoff obtained from the proposed method with default k value (0.8) were found to be 8.16–13.55% with an average value of 9.91%. However, with slightly modified value of k (0.9), the errors were significantly reduced to 1.94–3.32% with an average of 2.15%.

Thus, it can be concluded that the proposed method is an easy and effective method for quickly computing the runoff for any given residential area for designing RWH system. However, there are certain limitations of this method. The method requires the delineated catchment area and the area must be calculated prior. Furthermore, this method is suitable only for the residential area, not for the natural regions.

Data cannot be made publicly available; readers should contact the corresponding author for details.

The authors declare there is no conflict.