Use of Archival Aerial Photos and Images Acquired Using UAV to Reconstruct the Changes of Annual Load of the Suburban Landfill: Case Study of Promnik, Poland

Abstract

The aim of this study is to evaluate the rate of the changes of the annual volume of the municipal waste deposited in the landfill in Promnik, Poland in the period 2003–2020, and to assess the applicability of archival aerial photos for this kind of survey. The landfill analyzed in this article is the main recipient of garbage from the city of Kielce, with a population of 200,000. That assessment is crucial for municipal waste management planning, bio-gases management, expected use time and reclamation of landfill area. Important issues to consider are how the new legal regulations, in effect from 2012, affected the amount of the municipal waste and the rate of landfill growth. Measurement of the volume changes was conducted using a photogrammetric method on the archival aerial photos from 2003 to 2015 and aerial photos acquired using an unmanned air vehicle in 2020. Reference material was digital elevation model (DEM) of this area, derived from aerial laser scanning performed in 2014. Chamber works on the images were conducted using PCI Orthoengine and Agisoft Metashape software. DEMs of the landfill area were generated, and the volume changes of the dump canopy have been determined between 2003, 2014, 2015 and 2020. Height changes were measured along cross-sections and in probe of 150 random points. A significant decrease in the annual load of the municipal waste after 2014 has been found, from over 70,000 m³ to 50,000 m³, which proves the effectiveness of the regulations introduced in 2012. Bio-gas productivity potential of the bio-active municipal waste layer was also assessed.

1. Introduction

An important issue of environmental engineering is the estimation of the amount of waste going to landfills. Bio-degradable organic waste, if not subjected to a utilization process, is the source of greenhouse gases, mainly marsh gas and carbon dioxide [1]. An assessment of the waste volume is the crucial issue for the planning of management of these gases. In Poland, waste management, including the measurement of the waste volume in landfills, is provided through relevant legal regulation [2,3]. Municipal waste cannot be deposited in landfills until after it has been processed in an MBT installation (mechanical and biological treatment). The main aim of these legal regulations is to increase the amount of recyclable waste and to decrease municipal waste dumped on landfills. Moreover, the waste after MBT process should be biologically inactive [4]. It seems that after 2013–2014 not only the amount but the quality of the waste going to the landfills would change, reducing their capacity to produce biogas.

One of the most convenient methods of measuring the increase in the volume of landfills is aerial photogrammetry, especially the use of a UAV equipped with high-resolution digital cameras [5,6,7,8]. The problem is that waste management also needs a volume survey data from a previous period, which is not always available. In this situation the use of archival aerial photos enables measurement of previous state of the dump. In the studies on biogas management, only the mass of waste stored in the previous period is analyzed, according to data provided by the landfill manager [9]; therefore, no data about actual thickness of the waste layer and its volume are available, except the maximum elevation of the dump canopy. The above-mentioned legal regulations [3] require the survey of the dump canopy, but this kind of survey was not conducted before 2013. In this situation, the only available method of measuring landfill changes is photogrammetry using archival aerial photos. Very few studies report using archival aerial photos for the purpose of landfills monitoring [10,11], with most documenting the use of only planar measurements. There are more numerous studies that present the use of archival aerial photos for the purpose of landform measurement and analyses [12,13,14,15], including stereophotogrammetry. The methods used in these studies can also be implemented in the analyses of the landfill volume change, and the degree of their usefulness and possible limitations seems to be a research gap to be filled.

The studied landfill is situated in Promnik, near Kielce, Poland (Figure 1), and it is the main recipient of the waste produced in this city, populated by over 200,000 people, including the suburban area. This makes it a representative example for research on the impact of the implementation of the new law on the functioning of the waste management system for urban areas. The objectives of the study were divided into two significant issues. The first important objective was to assess the magnitude/rate of the reduction in the amount of municipal waste going to the landfill each year (annual load) after introduction of the Polish Waste Act, 2012 [3]. The second issue is related to the photogrammetric measurement methods, and is the identification of the limitations of these methods. The research required the use of a combination of materials: contemporary high-resolution images, and archival aerial photos and DEM of aerial laser scanning (ALS). The use of such data generated a number of problems to be solved: different resolution and quality of the images, recognition and measurement of the ground control points (GCP) for aero-triangulation of the older images, and assessment of the measurement uncertainty on the digital elevation models (DEM), generated using archival aerial photos, referring to the contemporary images acquired using UAV. This approach is quite new to waste management analysis, and uses the method previously applied in the landform analyses [16,17]. It should allow not only the assessment changes in the volume of the landfill, but also the determination of the spatial variability of the waste layers. It is also very useful in the case of lack of applicable surveys of the landfill terrain, caused by a lack of applicable law regulations in the earlier period of its use. Summarizing the objectives described above, two questions can be formulated:

–

how did the annual amount of waste change as a result of the change in the law in 2012?

–

how accurately can changes in the shape of the dump canopy be measured, taking into account the quality of archival photogrammetric material?

The images acquired using a UAV in 2020, and the reference data (aerial photos and ALS—derived DEM) of 2003–2015 were used to assess the rate of dump volume increase. It was assumed, that the crucial date is 2014, because the mentioned legal regulations, introduced in the end of 2012, needed about a year to be fully implemented. The reference photogrammetric data of the 2003–2015 period are available in the polish governmental documentation center.

2. Materials and Methods

The dump volume changes in 2003–2020 period was measured using DEMs derived from the photogrammetric materials: aerial photos of 2003, 2015, 2020 and the point cloud of ALS survey of 2014. The archival aerial images of 2003 and 2015 were acquired as a part of large photogrammetric raids for the purposes of governmental land information system, performed using the cameras described in

Table 1. The cameras used and actual ground resolution of the obtained photogrammetric images.

Year	Camera (Sensor) Type/Image Resolution (Pixels)	Terrain Resolution (GSD in m)	Focal Lenght/Frame Dimensions (mm)	FOV Angle (o)	Flight Altitude above Ground [m]
2003	Analog frame camera Leica RC 20 UAGA/16,840 × 16,565	0.19	153.36/ 230 × 230	90	2070
2015	Digital frame camera UltraCam Eagle/13,080 × 20,010	0.25	79.8/ 68,016 × 104,052	77	3750
2020	Digital frame camera DJI Zenmuse X5S/5280 × 2970	0.027	15/ 18.9 × 10.6	72	120
2020	Multilens digital frame camera MicaSense Rededge MX/1280 × 960	0.084	5.5 4.8 × 3.6	47.2	120

. Selected images of these raids have been shared by the Head Office of Geodesy and Cartography of Poland for the study purposes. In the case of the aerial images from 2003 and 2015, only single stereopairs were needed for elaborate DEM. The reach of these stereomodels is presented in Figure 1. As presented in the figure, they adequately cover the surface of the landfill. In the case of the images of 2003, only two images were used for aero-triangulation. In the case of 2015, three images were needed to get more GCPs for bundle adjustment process, but only one stereomodel was generated. For both 2003 and 2015 images, GCPs have been measured during chamber works, using detailed digital maps for civil engineering purposes (horizontal coordinates of the objects being GCP, such as basis of the poles, corners of the fences), and ALS-derived DEM (height). GCPs were examined and selected according to the method of bundle adjustment in many iterations [17]. In the case of the elaboration of photos of 2003, 8 of 13 examined points were rejected. In the case of the photos of 2015, the previously selected GCPs were used, and nine additional were points examined, three of which were rejected. Finally, the best bundle adjustments (defined by RMS error as presented in

Table 2. Accuracy of DEM used for analysis.

Year	RMS Error of Bundle Adjustment [m]	DEM Resolution [m]	RMS Error of Height [m]
2003	0.38	1	0.48
2014	-	1	0.15 3
2015	0.26	1	0.45
2020 1	0.11	0.1	0.23
2020 2	0.08	0.2	0.17

¹ DJI Zenmuse X5S camera; ² MicaSense Rededge MX camera; ³ According to Kurczyński [21].

) were achieved with the GCPs pattern presented on Figure 1. For the DEM derived from the 2014 ALS, the coverage of a single datasheet obtained from a governmental documentation center is marked in Figure 1. Additionally, Figure 1 depicts the area of the photogrammetric flight using the UAV in 2020. The main issue before the volumetric analysis was the assessment and comparison of the accuracy of survey of these materials due to their different resolution. The properties of the cameras and aerial photos used for the analysis are presented in the Table 1. The actual resolution is varied, due to camera parameters as well as flight parameters. The main factors affecting potential accuracy of the photogrammetric survey are flight altitude and optical parameters. In the case of the cameras used in 2003 and 2015, the lenses were fixed and calibrated in metric system [18,19]. In the case of 2020 UAV-carried cameras, the calibration process was performed in the GCP and tie points on the acquired images. Both sets of the images of 2020 have been acquired during a single flight, using a DJI Matrice 210 RTK v2 quadrocopter. The altitude of the flight was set at 120 m a.g. l., because it provided resolution-meeting conditions of the surveys for the purposes of civil engineering in Poland [20]. The cameras mounted on a two-set gimbal acquired the sets of photos according to separate patterns. This solution was necessary because these cameras have different optical parameters (Table 1). The number of the images acquired by Zenmuse X5S camera was 451, and in the same time Multilens MicaSense Rededge MX camera acquired 1735 images, in 5 separated spectral channels.

The archival aerial photos from 2003 and 2015 have been developed and used for stereo-photogrammetric measurement, using PCI Orthoengine software. The images of 2020 have been elaborated using Agisoft Metashape software, being better adapted to work with UAV-acquired images, with camera locations measured using GNSS/INS system. Automatically extracted photogrammetric DEMs were edited and corrected using reference ALS-derived DEM according to the same method applied in similar research [17]. The waste volumes of each date were calculated using spatial analysis tools of Quantum GIS software, for following assessment of annual load volume separately between 2003 and 2014, 2014 and 2015, 2015 and 2020. The latest volumetric survey was made using set of images acquired on 22 October 2020, using UAV DJI Matrice 210 RTK v2, as described above. ALS-derived DEM of 2014, was developed according to the 1st standard of DEM of ALS data for the ISOK project [8]. The RMS height error for flat surfaces without vegetation was assessed as no more than 0,15 m in this standard and their cell dimension was 1 m. The date of the ALS survey (April 2014) coincides with the crucial date of implementation of a new way of waste management due to a UE directive and a Polish Waste Act [2,3]. High accuracy and appropriate acquisition time made it suitable as a reference surface model. Except for this, survey data of 2016 exist, covering the area of the dump canopy and its close surroundings (drainage ditches). The surveys were carried out to create a network of marks for the subsequent monitoring of landfill subsidence. Marks are not identifiable on aerial photos, and thus their usefulness in the analysis of volume changes is limited.

The landfill volume changes were estimated using raster statistics and raster calculator tools in Quantum GIS software. The DEMs used have been converted to regular grid format, with cell dimensions of 1 m in the case of the models of 2003, 2014 and 2015, and of 0.1 m in the case of the model of 2020. Ground reference level (ground without waste cover) was assumed as 267 m a.s.l. The volume of each model was calculated as the sum of the products of the grid cell area and grid cell value. In addition, the profiles were created to analyze the deformation of the landfill surface. Cross-section lines used for creating the profiles are presented in Figure 2. First one is delimited from N to S, through all quarters of the landfill, from the oldest (N) to the active one (S), along the highest part of the dump canopy. The second one, running from W to E, crosses the active (in 2020) quarter of the dump canopy. Location of the cross-lines allows the assessment of the growth of the dump, as well as the subsidence process associated with the production of gases.

There is also a set of control points marked in Figure 2 that were used for georeferencing of the UAV-acquired images. Three types of points were used: temporary marked GCP on the top of the dump (P1–P4) and points on field details, signed only by numbers, both surveyed using GNSS RTK method; the third one are points signed as K1–K5, measured during chamber works, using ALS-derived DEM, and used as check points for the bundle adjustment of an aerial images. In the case of the points measured on the field details, their distribution is uneven, due to the lack of appropriate objects for measurement. Except for the point presented in Figure 2 and described above, the probe of 30 random points on flat, unvegetated area were measured during chamber works in the area covered by the models

The comparison of the DEMs generated and tested with the mentioned probe points is presented in the Table 2. Residuals of DEMs of 2003, 2015 and 2020 were calculated relative to the DEM of 2014. In each case, no significant pattern of the residuals spatial distribution was noticed, and it was determined random. The accuracy of the reference DEM has been independently tested using a GNSS RTK terrain survey [16,21].

For the 2020 aerial photos, a multispectral camera was used to better measure the control points, but the DEM of the X5S camera was eventually used for further analysis due to its higher resolution and lower probability of major errors. The multispectral camera exhibits better image quality, but lower resolution and a narrower field of view (Table 1). The DEM quality of the X5S camera images was better due to the higher overlap rate of the images. The RMS DEM error was the basis for assessing the uncertainty of the volume measurement. This uncertainty was calculated as the quotient of the average layer growth and the RMS value and expressed as a percentage.

3. Results

The volume changes of the dump canopy are presented in

Table 3. Estimated increase in the volume of the dump canopy of landfill in Promnik in 2003–2020 period.

Year	Volume [m3]	Uncertainty [%]
2003	1,225,190	±6.81
2014	2,008,349	±1.04
2015 1	2,075,359	±5.23
2015 2	2,048,554	±6.11
2020	2,305,191	±1.41

¹ With and ² without heap of temporary stored of waste of MBT process.

. Annual load before 2014 was estimated based on this data as 71,196 m³. Analysis of the DEM and orthophoto map of 2015 demonstrated that 40% of load (volume of 26,805 m³) deposited after 2014 is stored as a separate heap of treated waste after MBT processes, which was removed before 2020. The volume of common municipal waste in 2014–2015 was estimated as 40,205 m³, which means a significant (40%) decrease in the annual load. In the years following 2015, an annual load increased to 51,327 m³, referring to the volume of 2015 without the MBT waste heap, and together with the heap volume of 45,966 m³. This data allow one to assess the amount of the municipal waste being recycling by product. The mean annual load between 2014 and 2020 is assessed as 49,473 m³.

These data do not include the amount of subsidence, which could be estimated by analyzing cross-sections (Figure 3). Comparison of the profiles indicates a significant (approx. 2 m) amount of subsidence between 2003 and 2014 in the oldest, northern part of dump canopy (on the left part of the cross-section, from 30 to 180 m of the cross-section), and measurable (about 1 m) amount of subsidence between 2014 and 2020 in the middle part (from 200 to 280 m of the cross-section). The profile from 2015 is generally lower situated than the ones from 2014 and 2020, which can be partly explained by subsidence, but the greater uncertainty of the 2015 study, compared to 2014 and 2020, should be taken into account.

The thickness of the municipal waste layer of successive time horizons can be estimated as 7–13 m between 2003 and 2014, and 8–10 m between 2014 and 2020. Cross-section of the active cell of the dump canopy from W to E (Figure 4) shows the thickness of the waste from 2003 to 2014 at approx. 6–7 m, and in 2014–2020 at approx. 8–12 m.

The analysis of the thickness of the new dump canopy layers outside the cross-sections was carried out on a sample of 150 random points, generated with analysis tools of QGIS, and used to compare the values of the height on DEMs of 2003, 2014, 2015 and 2020 (Figure 5). The obtained results are similar to those obtained from the cross-sections, which confirms the correct selection of the cross-section lines. Generally, the most intense increase in the thickness of the waste layer can be observed in the southern part of active quarter. In some places, negative values are visible, especially for the 2014–2015 period. It can be explained by both the uncertainty of the 2015 survey, and rearrangement of the part of the dump in this time, which is visible on Figure 4, as the difference between 2014 and 2015 profiles.

4. Discussion

The results obtained prove the undoubted influence of the new law [2,3] on the annual load of municipal waste deposited in the landfill. The decrease in annual load volume is significant, but the amount of the waste is still about 50,000 m³. Analysis of the profiles showed layers of waste stored before 2014 on the entire active surface of the dump canopy. These layers are potentially the source of biogases because it can be assumed that only the waste stored on the dump canopy after 2014 has the characteristics of “stabilat”, i.e., it is not biologically active [15]. According to Rosik–Dulewska [1], unit production of the biogases (determined as m³ of the gas per 1 m³ of the waste per year) decreases from 3 m³ of the waste in the first 12 years of storage to 1 m³ after 16 years of storage. The overall potential productivity of the dump canopy can then be assessed as 1,500,000 m³ for layer of the period 2003–2014. An assessment of real biogas productivity of landfill in Promnik was performed for the middle part of the dump canopy (from 210 to 365 m of the cross-section on Figure 3), which was closed in 2011 [9]. Comparison of two math models and reference data of emission measurement showed that the real emission is even higher than those estimated according to the older method [1], reaching a maximum value of 900,000 m³/year. The active cell of the dump canopy (from 365 to 510 m of the cross-section on Figure 3) consists of two layers: older (before 2014), potentially producing biogas, and newer (after 2014), filled by waste after MBT, which should not be biologically active. The thickness of the first one is about 6 m and the second about 9 m. The point probes showed on Figure 5 are an example of the analysis, useful for the planning of gas discharge system devices. It shows how deep below the current level of the waste surface the bioactive waste layer is. The assessment of real biogas emission is also possible on the basis of images obtained using a thermal camera carried by a UAV [22].

A separate issue is the compatibility of the results of research carried out using photogrammetric methods with field measurements. Field surveys are usually used for estimation of the annual volume of the waste [23]. In the study [4], some differences (RMS 0.07 m horizontal and 0.37 m vertically) between field survey and UAV-based photogrammetric survey were noticed. The differences varied from 4% to 9%, depending on the location of GCP in the active zone of landfill. Similar results were obtained in the case of research [23], but in that case only DSMs were examined, which after editing to DEM should achieve greater accuracy, which was proven in the case of the landform analysis [17]. However, very dense vegetation limits the accuracy of the measurement.

The analysis of the accuracy of the photogrammetric materials used in this study proves that the optimal survey method for municipal waste volume estimation is the UAV flight with the acquisition of photos with a ground resolution of 0.02–0.1 m. The accuracy of the survey is comparable to the ALS measurements, and meets the standards of engineering terrain surveys in Poland [20]. The use of archival aerial photos allows one to perform the surveys and analysis of the older waste layers, but their accuracy is lower (RMS in control points in rate 0.4–0.5 m). That accuracy does not meet the requirements of engineering terrain surveys in Poland, but in the case of rapidly changing objects, such as dump canopy, it is usually the only source of information about the previous state of the object. Another case study requiring the use of archival aerial photos to determine the previous state and pattern of the object is the survey of the river channel changes [17]. This type of research indicated a similar rate of surveys’ accuracy using similar material. The accuracy of the photogrammetric survey in this case was determined by many factors affecting two kinds of error: random and systematic. The random errors are usually much higher than systematic, but their impact on the analysis can be minimized by using appropriate number of probing points and a statistical test. In the analysis mentioned above, a bundle adjustment error of about 0.4 m resulted in random errors exceeding 1 m and systematic errors of less than 0.1 m. As a result, the uncertainty of the assessment of changes in the total height was about 0.17 m. This method based on the use of archival aerial photos can be implemented for the assessment of the subsidence of the dump canopy; however, confirmation of its effectiveness would require a separate study. In the case of the data of ALS survey or high resolution aerial photos acquired using a UAV that meet the standards of survey for land engineering [20], it is enough to compare profiles or calculate a raster map for the analysis of subsidence.

The method of combined DEM analysis of landfills based on archival aerial photographs and cameras carried by UAVs can be widely used in all countries where photogrammetric raids have been regularly repeated over the last 20 years. The 20-year limit was adopted on the basis of waste bio-productivity studies [1,9,15]. A large number of landform studies using this method prove the high availability of archival aerial photos [12,13,14,15,17] and confirm the convenience and benefits of using UAVs to measure the contemporary volume of landfills [5,6,7,8].

5. Conclusions

The study of the landfill in Promnik presented a new approach to the issue of assessing the annual load of the municipal waste. Applying the different images, both the archival (2003, 2015) and contemporary, acquired using UAV or derived from ALS survey (2014), allows for the reconstruction of the shape and volume changes of the landfill in the past. The problem is that archival aerial photos enable measurement with limited accuracy (about 0.5 m), which increases the uncertainty of assessing the volume of the older waste layers (±6.81%). On the other hand, the bio-gas productivity of these layers decreases over time [9], which reduces the scale of the problem. In any case, this method can be considered effective enough to assess the biogas productivity of these layers. The research also presents an analysis of changes in waste management, taking into account the adaptation to new legal regulations. In the case of the Promnik landfill, the crucial data were the DEM derived from the 2014 ALS, which shows the state of the landfill at the time of the commencement of landfilling with the use of new disposal methods applied after the introduction of new regulations. Thus, the surface of active cell on this DEM, can be considered as the spatial boundary between the bioactive and inactive (after the MBT process) waste layer. The survey using the images acquired using a UAV in 2020 can be considered a normal monitoring procedure, meeting the requirements of the measurements for civil engineering purposes. The accuracy rate of the survey (RMS of the DEM about 0.23 m) has a similar value in other studies [5,6,7,8], which confirms the usefulness of this surveying method. The use of this method to estimate the dump volume in the study site allowed us to document a significant decrease in the annual load of municipal waste after 2014, from over 70,000 to 50,000 m³, which proves the effectiveness of the implemented regulations..

To conclude, the combined use of the archival aerial photos and aerial photos obtained using UAV allows one to assess:

–

the actual volume of waste;

–

the thickness of the waste, including storage time and bio-activity;

–

spatial diversification of the shape and thickness of the waste layer.

These parameters are of key importance in the planning of bio-gases management as well as in the process of their reclamation. The data obtained by means of the mentioned methods supplement the documentation kept by the landfill managing authority. Comparison with other studies [17,22] also allows for the development of photogrammetric methods using UAVs in order to assess the subsidence process, also needed in planning the reclamation of landfills.