According to University of Michigan’s center for sustainable systems, Municipal Solid Waste (MSW), commonly called “trash” or “garbage”, includes wastes such as durable goods (e.g., tires, furniture), nondurable goods (e.g., newspapers, plastic plates/cups), containers and packaging (e.g., milk cartons, plastic wrap), and other wastes (e.g., yard waste, food). This category of waste refers to common household waste, as well as office and retail waste, but excludes industrial, hazardous, and construction waste. Municipal solid waste management (MSWM) encompasses the functions of collection, transfer, resource recovery, recycling, and treatment ( Henry, Yongsheng, & Jun, 2006 ). ( Kimwatu & Ndiritu, 2016 ) in their study n applications of GIS and remote sensing in solid waste management defined municipal solid waste as a heterogeneous collection of wastes produced in urban areas, the nature of which varies from region to region.

In developing countries, although open dumping is common, there is also a realization that this is inadequate ( Shekdar , 2009 ). Despite dumps being the most common method according to waste management hierarchy, it is the final and least recommended solution ( Faitli et al., 2015 ). Globally, sustainable waste management has been a major challenge. Waste degradation in the dumps lead to generation of emissions. They are the most significant contributors of GHG emissions ( Kumar et al., 2019 ). Currently, more than half, 55%, of the world’s population resides in cities. The urban share of the world’s population is projected to increase to 68% by 2050 (United Nations Department of Economics and Social Affairs, UN DESA). This has led to the increase in generation of municipal solid waste due to the increased urbanization levels and increase in income levels. This makes waste management one of the most important priorities for governments. The East African Community (EAC), Kenya, Uganda and Tanzania have attempted to address proper waste management through the East African Community (2016) . Illegal waste management is a critical issue for contemporary governments. Locally, poor waste management hinders the delivery of the constitutional right to a clean and healthy environment. Most communities across Kenya do not receive proper waste management and disposal services. This leads to the rise of informal and illegal dumpsites. This is a challenge in proper waste management ( Ministry of Environment and Forestry, 2021 ).

Kenya’s national waste policy envisions transition from low collection, illegal dumping and uncontrolled dumpsites to affordable waste collection, recycling, and composting, and waste minimization. The continued growth of Kenya’s economy and cities due to devolution has compounded its waste management challenges. The policy emphasizes the necessity to develop new and innovative solutions that include research and development in emerging technologies to help map, collect, segregate, quantify and utilize waste. It also encourages engagement with all the stakeholders at all levels of waste management. It is on the realization of these objectives that this project is formulated. NEMA’s national solid waste management strategy is being implemented through five key objectives that include formulation of policies, legislations, and economic instruments to reduce waste quantities, inculcating responsible public behaviour on waste management, promoting waste segregation at source and resource recovery for materials and energy generation, establishing environmentally sound infrastructure and systems for waste management. To meet the five key objectives, it is imperative that solid waste dumps are identified and mapped.

Reliable data on waste generation are recorded and are available. Data is collected daily and provides a rational basis for planning and executing waste management operations. By contrast, in developing economies the data on solid waste generation is collected through surveys that are only deployed for a brief time and may be limited to certain cities. Moreover, the data related to solid waste that is transported by the system, which may not be equivalent to the quantity of waste generated ( Shekdar , 2009 ).

Currently no comprehensive and integrated monitoring or mapping of illegal and legal waste is undertaken in Kenya and the national sustainable waste management policy categorizes the inculcation of timely inventories on quantities and types of waste generated as a gap that needs to be addressed.

The illegal dumping of waste materials presents a significant challenge for environmental conservation efforts due to the potential risks it poses to human health and natural habitats ( Helvoort , 2023 ). Poor disposal of municipal solid waste has major adverse effects on health and the environment. The bi products resulting from the waste mass could infiltrate the local environment posing serious threats that degrade the quality of the environment as well as the human health ( Slonecker et al., 2010 ). Organic waste products degrade and produce toxic substances, such as biogas, leachate, and heat. The other biproducts produced due to waste decomposition, biogas and leachate, have an injurious effect on the surrounding vegetation. Dumping of MSW not only damages aesthetic beauty but also causes air pollution and groundwater contamination ( Kumar et al., 2019 ).

Previous studies have assessed the impact of heat generation of dumpsites and the leachate leakage on the surrounding vegetation and while that has been a noble contribution to the scientific body of work, an improved understanding of the distribution of illegal waste disposal can support improvements to the cost-effectiveness and efficiency of waste management efforts ( Glanville & Chang, 2015 ).

Waste disposal into landfill and dumps is the most adopted technique of waste management in many countries, despite being the least recommended waste management solution ( Gill et al., 2019 ). Dumps are significant air and soil pollution sources especially when they are not sanctioned and not controlled ( Notarnicola, 2004 ). The lack of proper waste treatment facilities has also majorly contributed to the problem of many informal garbage dumps. Existence of a proper waste monitoring and characterization system is one of the most important requirements in waste management.

The first analysis on the application of remote sensing for waste management was published by Garofalo in 1974. This study discusses the utilization of aerial photographs to support estimation techniques of solid waste distribution and production. The methodology is based on the visual interpretation of land use and the incorporation of these data into solid waste production models ( Ottavianelli et al., 2005 ). Yonezwa investigated the usefulness of monitoring waste on land areas using data from the currently operated earth observation satellites ALOS (Advanced Land Observing Satellite) and Quickbird ( Yonezawa, 2009 ).

BJ-1 micro-satellite remote sensing images can provide support to open-air solid waste dumps management, and it has potential to reduce the operational costs, so it can meet the challenges of the solid waste management agenda ( Yalana, et al., 2008 ), while ( Richter et al., 2019 ) gathered and combined vector data with RS indices (vegetation, built-up area, and moisture) in a GIS, ranked and assessed Nova Scotia’s seven waste management regions based on a normalized ranking classification. In another study, the researchers used from the sensor Thematic Mapper on Landsat 5 in collaboration with digital ortho-photos (1:10000) and land cover map Corine 1990 data to identify degraded lands with a high environmental hazard in the Apulia Region in Southern Italy. They monitored and identified dump presence by the spectral signature specificity and individuated areas characterized by the same spectral properties ( Notarnicola, 2004 ). In his study, results revealed that the efficiency of waste management systems can be maximized by the proper use of remote sensing and GIS techniques. The study also revealed that these techniques were most used for sitting the landfill and waste bin for waste disposal and evaluation of environmental impact of buried waste ( Singh, 2019 ).

In their study, ( Mahmood et al., 2019 ) compare the suitability of different satellite-based vegetation indices (VIs) for environmental hazard assessment of municipal solid waste (MSW) open dumps. The compared VIs, as bio-indicators of vegetation health, are normalized by different vegetation index (NDVI), soil adjusted vegetation index (SAVI), and modified soil adjusted vegetation index (MSAVI) that have been subject to spatial-temporal analysis. ( Ottavianelli et al., 2005 ) sought to identify practical ways in which EO data can support landfill management and monitoring, providing quantitative data where possible. ( Jasravia Gill et al., n.d. ) in their paper on detection of waste dumping locations demonstrated how to utilize thermal remote sensing techniques to measure the land surface temperature (LST), which can aid in outlining the waste dumping regions within a landfill.

( Elsadiq Ali et al., 2009 ) used Radarsat to assess the potential of spaceborne SAR systems for mapping waste-disposal sites. The results showed that high resolution SAR imagery from space platforms could be a viable tool for the inventory of waste-disposal sites. An inventory that if monitored regularly can mitigate waste impacts on the environment while ( Glanville & Chang, 2015 ) analyzed existing remote sensing methods and sensors used to monitor and map illegal waste disposal sites to support the evaluation of existing remote sensing methods for mapping illegal domestic waste sites in Queensland, Australia. The informal garbage dumps in Beijing were monitored by SPOT-6 image and identified by the establishment of the interpretation key using multivariate data; and then classified by the established interpretation key ( Cheng & Sun, 2021 ). The geography of illegal waste disposal sites is not random, but a complex pattern influenced by a multitude of economic, environmental, and social factors such as demographics, availability of waste facilities or services, affordability of legal waste facilities or services, effectiveness of law, site access, remoteness of illegal waste disposal location, sociocultural acceptability of illegal waste disposal ( Glanville & Chang, 2015 ) while ( Yalan et al., n.d. ) in their study to detect the location of the open-air solid waste dumps in Beijing based on BJ-1 image found that the output and distribution of solid wastes has correlation with the quantity and density of the permanent population.

In their studies, ( Dabija et al., 2021 ) address the need for a comparative analysis of SVM and RF algorithms across different European regions using the same datasets to determine regional accuracy variations in land cover mapping. Their results informed the choice of support vector machine as a classifier of choice for land use and landcover mapping in this study. ( Cheng & Sun, 2021 ) detected and classified informal garbage dumps, particularly smaller, irregular sites that were not previously well-monitored using remote sensing due to cost and resolution constraints, which builds on this study’s utilization of various image interpretation techniques for identification of MSW dumping sites. It also informed the application of high-resolution satellite imagery for accurate mapping and identification. ( Glanville & Chang, 2015 ) focused on the limited use of RS in Australia for detecting waste disposal in Australia, a situation like Kenya’s application of RS and GIS in solid waste management.

Several studies have successfully demonstrated the application of remote sensing and GIS in waste management. For example, ( Dabija et al., 2021 ) in their study, they employed machine learning algorithms—specifically, Support Vector Machines (SVM) and Random Forests (RF) to classify land cover based on Sentinel-2 and Landsat 8 satellite images. The minimal mapping unit of 500 x 500 m used in the Corine program reduced the image’s informational potential.

In addition, ( Notarnicola, 2004 ) study used a remote sensing-based observational design to detect illegal dumps through spectral signature analysis. Cloud cover interference affected data quality for the February 1995 image and identification of small, heterogeneous areas like dumps was less accurate due to spectral variability.

Mobile GIS refers to an integrated software/hardware framework for the access of geospatial data and services through mobile devices via wireline or wireless networks ( Tsou, 2004 ). Mobile computing systems and hardware are changing the way mobile mapping technology is being used by moving GIS from the desktop into the user’s hands, providing flexibility in data acquisition, data accuracy and integrity, validation in real-time reducing errors and process costs, providing more information with much less time and effort, faster communication protocols, and high productivity, making the mobility an enticing aspect of GIS ( de Abreu Freire & Painho, 2014 ). Mobile GIS has emerged as a significant advancement in waste management, enabling real-time data collection and community engagement ( Kumar et al., 2019 ). By providing citizens with tools to participate in mapping and reporting MSW dumping sites, mobile GIS enables communities to play an active role in the circular economy. This participatory approach is important for cultivating ownership and responsibility among communities, hence resulting in improved waste management.

Waste management is a shared responsibility. Unmanaged waste can be a threat to public health, and community participation is essential in waste management. Community participation in waste management can be implemented through active involvement in the process of disposal, transportation, and waste management, with a sense of awareness and responsibility to create a clean and healthy environment ( Tarigan , Rogaleli, & Waangsir, 2020 ). Successful case studies demonstrate that integrating mobile GIS with community participation can lead to significant improvements in waste management systems, as communities are more likely to adhere to waste management guidelines when actively involved in the process ( Kumar et al., 2019 ).

In the prior years, waste was viewed only as a problem and not as a resource and economic opportunity. This can only be achieved through proper waste management ( Ministry of Environment and Forestry, 2021 ). An improved understanding of the distribution of illegal waste disposal sites is critical to enhance the cost-effectiveness and efficiency of waste management efforts. However, there has been limited use of remote sensing for monitoring and mapping illegal waste disposal despite its potential to address this knowledge gap. Remote sensed images provide effective, efficient monitoring and mapping which support improved management outcomes ( Glanville & Chang, 2015 ). The critical information provided by remote sensed images is important in meeting targets for proper solid waste management.

Juja Ward is an administrative unit in Kiambu County with an area of approximately 326.6 square kilometers. Juja town is located about thirty kilometers north of Nairobi between Thika and Ruiru town as shown in Figure 1 , and it is under the Nairobi metropolitan authority as envisaged in the vision 2030 of Kenya. Estimated population is 272,737 people ( National Government Constituencies Development Fund [NG-CDF] Juja, n.d. ).

Dandora area is in Embakasi North and Ruaraka Constituencies as shown in Figure 2 , respectively. As reported by UNEP in 2018, Dandora dumpsite covered approximately thirty acres and is the destination for approximately 850 tonnes of solid waste daily from Nairobi. This location was chosen due to its elevated levels of illegal dumping and insufficient waste management services, making it an ideal case study for examining the application of remote sensing and GIS technologies in waste management ( Kumar et al., 2019 ). Juja area has smaller, irregularly distributed dumpsites while the Dandora area of interest contains a well-known, large dumpsite and therefore its identification through classification and the extraction of the spectral reflectance curves of the dumpsite and of the other land use/cover classes zones will provide a benchmark for identifying and validating areas characterized by the same spectral properties in Juja area.

Data collection for this study involved various sources as shown in Table 1 , that ensured comprehensive identification of MSW disposal sites. The software used for this practical was ArcGIS to design the forms and collection of dumpsite data for analysis.

Field data collection involved the collection of solid waste dump characteristics listed in Table 2 as generator name, waste type, location, ownership and whether the dumpsite is active or not. A total of 171 random points were collected in ArcGIS online within the administrative boundary of Juja Ward. An ArcGIS online project group was constituted which contained the data collection form and an Esri base map for reference. Data Collection form was created in Survey123 desktop. This was in the form of an editable xlsx with the following attributes:

Six enumerators with smartphones running android operating systems were used for data collection. Survey123 android application was downloaded from google Play Store and installed on each of the phones. Six named users were assigned using the enumerator’s email addresses and unique login details for each generated. Five of the enumerators were assigned fieldworker user type while one was the administrator with publishing privileges. The form was published to the ArcGIS online group android phone from Google play store. The enumerators logged in to the Survey123 mobile app and downloaded the published dumpsite survey form ready for field data collection.

Data collection involved spreading out within Juja Ward starting at Jkuat on day one and spreading outward in all directions on days two and three interviewing household members and recording the answers and taking pictures of the dumped waste. This was repeated until the sampling size of 180 was slightly surpassed. The data collected was automatically syncing into ArcGIS online cloud database. Downloaded data was cleaned for visualization and analysis. This included cleaning typos and re-coding the dumpsite code from 1 to 183. The data will be visualized on a dashboard and a web map created for ease of identification of the locations of dumping areas. A summary of the materials and methods is represented in Figure 3 below.

This imagery was acquired from planet explorer portal. Planet offers two geometry types for PlanetScope imagery: Basic and Ortho. Basic Scene is not orthorectified or corrected for terrain distortions. The geometry used in this study is Ortho Scene, which is orthorectified with additional post processing applied. For ortho products, GeoTIFFs are resampled at 3 meters and are projected into the UTM projection using the WGS84 Datum ( Planet Labs, n.d. ). The 3m resolution planet imagery was preferred because moderate-resolution sensors like Sentinel, LANDSAT and ALOS had limited success in identifying smaller illegal waste sites due to their lower spatial resolution (4 - 50 meters) ( Glanville & Chang, 2015 ). The analytic products are accompanied by a Usable Data Mask (UDM2) file. This is a raster file consisting of eight bands with information about each pixel quality in the scene. Each band provides context on whether a pixel is clear, cloudy, or shadowed, amongst other characteristics. These bands allow users to remove non-useful pixels post download of the image. UMD2 also compensates for effects due to vignetting, low signal-to-noise, or hot or cold pixels.

The PlanetScope Ortho Scene product is orthorectified and is designed for applications that require imagery with accurate geolocation and cartographic projection. These scenes have been processed to remove distortions caused by terrain. Ortho Scenes are delivered as Visual (RGB) and Analytic products. Ortho Scenes are radiometrically, sensor, and geometrically corrected products projected to a cartographic map projection ( Planet Labs, n.d. ).

Analysis-Ready PlanetScope combines data from the three available PlanetScope sensors (Dove Classic, Dove-R, and SuperDove) while enhancing temporal and spatial consistency with Landsat, Sentinel-2, MODIS, VIIRS). It delivers four surface reflectance bands: blue, green, red, and near infrared. The 4-band analytic product used for this study was delivered as Surface Reflectance (SR) ( Planet Labs, n.d. ).

Analysis-Ready PlanetScope band order:

The 4-band multispectral band order is:

Planet’s satellite imagery was acquired from planet explorer for purposes of LULC mapping. The RGB band combination was used for analysis and the raster was clipped to the area of interest in ArcGIS Pro. Training samples were then collected for five classes namely dumpsites, built up areas, water, bare land and vegetation by drawing polygons around known locations of each of the five classes and their definition in a classification schema. We then ran the imagery classification wizard on the support vector machine algorithm. The output was visualized a few times and the process refined, validated and compared. In their study to assess the automatic land classification accuracy according to Corine guidelines based on Sentinel-2 and Landsat 8 multitemporal images ( Dabija et al., 2021 ). The Sentinel-2 satellite images allowed to classify land cover with better overall accuracy (8% - 10%) than the Landsat 8 data, and the Support Vector Machines algorithm with an RBF kernel function achieved the best results and obtained higher overall accuracy results (6% - 7%) than Random Forest.

Supervised classification is performed when an analyst defines the spectral characteristics of the classes by identifying training areas. It is a requirement for the analyst to be familiar with the area of interest, hence the inclusion of Dandora dumpsite, which is large, known and easily identifiable on high resolution satellite images. It will therefore be easier to identify sample areas of the unknown information class of a dumpsite ( Lillesand, Kiefer, & Chipman, 2009 ). The choice of an SVM classifier was influenced by ( Dabija et al., 2021 ), whose study employed machine learning algorithms, specifically, Support Vector Machines (SVM) and Random Forests (RF) to classify land cover based on Sentinel-2 and Landsat 8 satellite images. The landcover was classified into six classes namely water, vegetation/forests, dumpsite, bare land/grassland, built-up areas and roads. The design was aimed at comparing the performance of these algorithms in land cover mapping in various European regions. The research also focused on assessing classification accuracy for different land cover classes, considering regional and spatial variability across diverse geographic areas. For accuracy assessment, they calculated metrics like overall accuracy (OA), producer accuracy (PA), user accuracy (UA), F1-score, and Kappa coefficient. SVM was preferred due to the small training data sets, robustness to overfitting and its strength in modelling non-linear boundaries, which suits Dandora area’s landscape of mixed and transitional landcover. Radial Basis Function (RBF) Kernel using a 2D grid with optimal parameter pairs (C, gamma) in ArcGIS Pro was used due to less susceptibility to noise, correlated bands, and an unbalanced number or size of training sites within each class ( Esri, n.d.-a, n.d.-c ).

Several pixels were used to extract spectral reflectance curves in homogeneous regions to avoid a great variation of values, comparable to the class of interest. The regions of interest are of different shapes. The dump in Dandora has a large and smooth continuous surface located in the North-Western side of the study area. Sensing in several spectral bands simultaneously allows one to relate properties that show up well in specific spectral bands, therefore enabling the discrimination of materials of interest like dumpsites and their various components based on their spectral reflectance curves. We can establish for each material type of interest a reflectance curve. Such a curve shows the portion of incident energy that is reflected as a function of wavelength ( Lillesand, Kiefer, & Chipman, 2009 ).

Image classification results in a raster file in which the individual raster elements are class labeled. As image classification is based on samples of the classes, the actual quality of the result should be checked. This is usually done by a sampling approach in which several raster elements of the output are selected, and both the classification result and the true world results are compared. Comparison is done by creating an error/confusion matrix from which different accuracy measures can be calculated. The true world class is preferably derived from field observations ( Lillesand, Kiefer, & Chipman, 2009 ). A total of 500 samples were collected for the purposes of accuracy assessment using ground truth data from the 2021 raster. Through equal stratified sampling 100 cases each of built-up areas, dumpsites, vegetation, water and bare land respectively were found in the real world for all four epochs. Sample points were created by dividing the study area into five distinct and contiguous raster region strata of built-up areas, dumpsites, water, bare land and vegetation, and performing simple random sampling separately within each stratum.

The measures of mapping accuracy used were the overall, user, producer accuracy and Kappa statistics. Overall accuracy is the number of correctly classified divided by the total number of pixels checked. The user accuracy is the probability that a certain reference class has also been labeled that class. The producer’s accuracy is the probability that a sampled point on the map in that class is that class ( Lillesand, Kiefer, & Chipman, 2009 ). The burn area index (BAI) uses the reflectance in the red and NIR portion of the spectrum to identify the areas of the terrain affected by fire. This index highlights burned land in the red to near-infrared spectrum by emphasizing the charcoal signal in post-fire images ( Wu et al., 2022 ). In calculating the BAI of Juja and Dandora areas, we hypothesize that the field data collection points and the visual observation of the large and known Dandora dumpsite will overlay with high BAI results. Due to unavailability of training datasets for BAI, we automatically selected a threshold that minimizes intra-class variance, which created the new BAI raster that divided the data into two distinct classes, creating a low value class displayed with black pixels, and a high value class displayed with white pixels Esri (n.d.-b) . BAI outputs were subsequently integrated with the SVM map by creating a burn mask.

Informed consent was obtained from all participants involved in the study, ensuring their understanding of the research purpose and their right to withdraw at any time. Data confidentiality and privacy will also be maintained.

This study’s classification results from the SVM classifier resulted in a high degree of accuracy in identifying and mapping MSW disposal sites. Specifically, the SVM model was able to classify dumpsites, water, vegetation, bare land and built-up areas as shown in the LULC maps in Figures 4-7 below. An overall accuracy of 98, 95,92, 95 % for years 2021, 2022, 2023 and 2024 respectively was achieved as measured through the confusion matrices comparing classifications to the training data collected from a high-resolution multispectral imagery. Kappa coefficients were found to be 0.98, 0.95, 0.9 and 0.94 for the four epochs, showing strong agreement between the classification and actual ground truth. In this table five classes brackets C_1, C_2, C_3, C_4, C_5 representing built up areas, dump sites vegetation, water and bare land are listed. A total of five hundred samples were collected. Through equal stratified sampling 100 cases each of built-up areas, dumpsites, vegetation, water and bare land respectively were found in the real world for all the four epochs, while classification results for 2021 yielded 100,99,100,102 and 99 cases of C_1, C_2, C_3, C_4, C_5; in 96, 97, 100,100, and 99 cases they agree, as shown in Table 3 . In 2022, classification results yielded 88,111,101,100 and 100 cases of C_1, C_2, C_3, C_4, C_5; in 84, 97,100,99 and 99 cases they agree as shown in Table 4 . In 2023, classification results yielded 69, 118,100,101 and 112 cases of C_1, C_2, C_3, C_4, C_5; in 65,96,100,100 and 99 cases they agree as shown in Table 5 . Finally, classification results in 2024 yielded 94, 106, 101,99 and 100 cases of C_1, C_2, C_3, C_4, C_5 whereby in 85, 94,100, 100,99 and 97 cases they agreed as shown in Table 6 .

SVM accurately differentiated dumpsites from other land cover types in 2021, with 3 out of the 100 samples interpreted as the other classes, resulting in a 3% error of omission (Type 2 error) while 2 out of the 99 interpreted samples were incorrectly classified, hence leading to a 3% error of commission (Type 1). In 2022, SVM classifiers accurately differentiated dumpsites from other landcover types with 3 out of 100 samples interpreted as other classes, resulting in a 3% error of omission while fourteen out of 111 interpreted samples were wrongly classified, leading to a 13% error of commission. In 2023, 4 out of 100 samples were interpreted as other classes, resulting in a 4% error of omission while twenty-two out of 118 interpreted samples were incorrectly classified, leading to a 19% error of commission. Finally in 2024, 6 out of 100 samples were interpreted as other classes, resulting in a 6% error of omission while twelve out of 106 samples were incorrectly classified, leading to a 12% error of commission.

Dumping sites, especially those characterized by burning, were mapped with higher precision through the Burned Area Index (BAI) as a validation input in the classification process. The 2022 and 2024 BAI maps indicate a correlation between high BAI and location of Dandora dumpsite, as shown in Figure 15 and Figure 16 . BAI provided additional differentiation between burned dumpsites and other LULC types in Dandora, hence improving the performance of RS techniques in mapping and identifying MSW disposal sites.

Spectral signatures of all the four epochs are shown in Figures 8-11 . Vegetation (class 3), water (class 4), and bare land (class 5) show different and distinct spectral signatures. Wavelengths 0.865-0.021 µm indicate the peculiarities of dumpsites (class 2) with the spectral reflectance curves being stable. Although dumpsites (class 2) spectral signatures are well depicted, the maximum and minimum values are not as distant as they are in vegetation and water. This may be because Dandora dumpsite has heterogeneous materials. Built up areas (class1) exhibits spectral response similar to class 2 (dumpsites), with the most similarity being recorded in 2023. This could be the influence of seasonal/weather changes. The vegetation reflectance is low in the

visible spectrum and sharply rises in the NIR band. Water’s reflectance is also low in the visible spectrum, and sharply rises in the NIR, but its maximum reflectance is lower that vegetation’s across all epochs. Out of all the five classes, bare soil’s reflectance is the highest in visible band, with a steady and gradual increase towards the blue band and NIR.

The data from the field survey exercise conducted in January 2022 recorded 183 dumpsites as shown in Figure 12 , and 16 other attributes. 176 of the identified dumpsites are active while seven were inactive as shown in Figure 14 , translating 96% of the dumpsites operational in January 2022, hence indicating a persisting and improperly managed waste disposal. 166 of the dumpsites were illegal while thirteen were designated as shown in Figure 13 , showing that 91% of the dumpsites are unauthorized and pose safety, health, and environmental risks. Waste went uncollected in 119 of the sites, collection by individuals and vehicles/trucks was recorded at 27 and 10 dumpsites respectively while decomposition was used as a management method in nine of the sites. Waste went uncollected in 65% of the sites, hence manual and informal systems dominate. Containers and packaging, nondurable goods and household organic waste were the top waste types, an opportunity for single-use and organic waste recycling and composting, respectively. Finally, dumpsite ownership is formal and fragmented with limited institutional control. Eighty-one are community-based, thirty-three individuals while forty-three had unspecified ownership.

Some case studies involve the application of remote sensing to informal garbage dumps recognition in Beijing City ( Cheng & Sun, 2021 ) and remote sensing analysis techniques and sensor requirements to support the mapping of illegal domestic waste disposal sites in Queensland, Australia ( Glanville & Chang, 2015 ), which successfully identified informal garbage dumps using SPOT-6 satellite images, demonstrating an accuracy of over 90% in the interpretation of these sites and found that very high-resolution satellite data was most effective for detecting illegal waste disposal sites. Lower-resolution data was less useful for identifying smaller sites, respectively. The authors emphasized that the accuracy of remote sensing in detecting informal dumps shows great promise for future urban waste management, highlighting the efficiency and cost-effectiveness of using satellite data compared to traditional methods and suggested that the spatial resolution of satellite data is crucial for accurate waste detection and recommend combining GIS analysis with remote sensing to improve detection accuracy.

The results of this study illuminate on the critical issue of MSW disposal in Kenya, with significant implications for circularity, public health and sustainable development. Application of GIS and remote sensing techniques has demonstrated its ability and efficacy in identifying and mapping waste disposal sites, hence providing a cost-effective, valuable, and effective tool for national environmental authorities, circular economy entrepreneurs and policymakers. This study’s findings contribute to the pressing and broader discourse on sustainable waste management while offering actionable insights for realization of SDG 11 on sustainable cities.