Professional Documents
Culture Documents
ABSTRACT
In situ assessment of waste settlements is a critical consideration for the operation, closure,
and post-closure development of municipal solid waste (MSW) landfills. Settlement evolution is
influenced by the biodegradation processes underway in the waste mass. Presently, the
measurement of such settlements involves instrumentation at specific locations on a landfill, or
aerial surveys that are generally expensive and not conducted often. An alternative approach is
presented herein that involves the use of unmanned aerial vehicles (UAVs) to collect overlapping
imagery and create 3D models of landfills using the structure-from-motion technique. Repeated
surveys allow for differencing between 3D models to calculate settlement and compute
secondary compression indices that can be used in subsequent landfill development. Main
advantages of the technique include the generation of cm-level data resolution, the scalability of
the approach that allows assessments throughout the entire landfill, and the ability to execute
repeated flights for continuous, inexpensive monitoring of spatially resolved settlement. This
approach was tested at an MSW landfill in Midland, Michigan, from May 2018 to April 2019,
where spatial profiles of the modified secondary compression index were successfully derived
from UAV aerial imagery.
INTRODUCTION
Settlement at municipal solid waste (MSW) landfills is caused by a reduction in the volume
of waste and a weakening of its structure due to a combination of creep, raveling,
physicochemical change, and biochemical decay (waste degradation) (El-Fadel and Khoury
2000; Ivanova et al. 2008). Waste settlement is of particular interest due to its airspace
implications. Increased settlement of the waste mass allows landfills to dispose of additional
waste, yielding cost savings (Foye et al. 2007). Furthermore, understanding landfill settlement in
the field, particularly differential settlement, is crucial to ensure the resiliency and longevity of
the facility. Differential settlement refers to non-uniform settlement across the waste mass. This
type of settlement can cause cracks in the liner and cover systems (adversely affecting their
performance), ponding of water on the landfill, damage and potential backflow of leachate and
biogas collection systems, as well as environmental contamination. Differential settlement also
remains an issue after the landfill is closed by compromising the structural integrity of any post-
closure developments (Chen et al. 2009; Durmusoglu et al. 2005; Edelmann et al. 1999; El-Fadel
1999; El-Fadel and Khoury 2000; Sharma 2000).
© ASCE
Geo-Congress 2020
Geo-Congress 2020 GSP 319 250
Landfill settlement can be divided into three main phases: immediate, primary, and
secondary compression (Fei and Zekkos 2013). To assess settlement behavior over the course of
the design lifespan of a landfill, secondary compression is of great interest. The secondary
compression phase, which represents a significant portion of total settlement, begins shortly after
the time of waste placement (following immediate and primary compression) and can extend for
many years after landfill closure (Durmusoglu et al. 2005; Landva et al. 2000). Secondary
compression of landfills is commonly expressed using the modified secondary compression
index, Cαε , first introduced by Sowers (1973):
Downloaded from ascelibrary.org by Imperial College London on 02/23/20. Copyright ASCE. For personal use only; all rights reserved.
Cα ΔH εv
Cαε (1)
1 e H 0 Δlogt log t2 log t1
where Cα secondary compression index, e void ratio, ΔH settlement, H 0 initial
thickness of waste, ε v vertical strain, and t1,2 time (in days) into the secondary compression
phase for each measurement. The secondary compression index, Cα , is modified to eliminate the
need to measure waste void ratio in the field, which is practically impossible to do (Wall and
Zeiss 1995; Zekkos et al. 2013). It is important to note that in contrast to the constant secondary
compression index observed in soils, the Cαε for municipal solid waste has been shown to vary
depending on the biodegradation stage (Fei and Zekkos 2018).
In Situ Settlement Monitoring: In practice, settlement is typically measured at landfills
using settlement plates (Abichou et al. 2013; Andersen et al. 2004; Bareither et al. 2012; Simões
and Catapreta 2013; Yuen and McDougall 2003), surface monuments (El-Fadel 1999; Sharma
2000; Spikula 1997), inclinometers (Foye et al. 2007; Sharma 2000), and vertical extensometers
(Andersen et al. 2004; Sharma 2000). A major drawback of these instrumentation methods is that
the data collected are point measurements and no information is collected regarding the spatial
variation of the settlement between these points. Many studies have only a few measurement
locations to represent the behavior of the entire landfill, but there is significant uncertainty in this
assumption. Furthermore, many of these instruments are embedded into the final cover system,
meaning settlement monitoring only begins once the cell is closed. There are, however, many
reasons why a landfill owner would be interested in monitoring settlement while the cell is
active: (1) early post-closure planning; (2) damaging differential settlements can occur during
waste placement, thus putting the liners and collection systems at risk; and (3) mitigation of
differential settlement before the installation of the final cover system. There are also challenges
in deploying instrumentation at an active landfill, primarily due to the impact on landfill
operations. In studies that involved settlement monitoring during active landfilling, there are
reports of instrumentation being damaged by landfill construction equipment (Abichou et al.
2013; Andersen et al. 2004).
As an alternative to spatially isolated physical instrumentation, optical imagery collected
autonomously by unmanned aerial vehicles (UAVs) can be used to generate high resolution 3D
models. Models from subsequent surveys can be differenced to generate a continuous spatial
profile of settlement. The capability of measuring settlement spatially is particularly appealing
due to the significant variability and heterogeneity of waste composition, properties, and activity
at landfills (Durmusoglu et al. 2005; El-Fadel 1999; El-Fadel and Khoury 2000). Additionally,
collecting the data remotely from an aerial platform eliminates the risk of instrument damage by
landfilling equipment, and does not impede landfill operations.
© ASCE
Geo-Congress 2020
Geo-Congress 2020 GSP 319 251
DATA ACQUISITION
Site Overview: UAV-captured optical imagery was used to compute landfill settlements at
an MSW landfill in Midland, Michigan. The city of Midland experiences about 80 cm of
precipitation annually, with typical temperature highs between -1 and 29°C and typical lows
between -9 and 16°C (Midland, Michigan Period of Record Monthly Climate Summary, 2016).
The landfill has three currently active cells (14, 15, and 16) spanning about 15 hectares that have
been operational since April 1998, December 2003, and March 2008, respectively. The eastern
Downloaded from ascelibrary.org by Imperial College London on 02/23/20. Copyright ASCE. For personal use only; all rights reserved.
side of the landfill (cell 14 and eastern half of cell 15) is operated as a conventional “dry-tomb”
landfill while the western side (western half of cell 15 and cell 16) is operated as a bioreactor
landfill, as illustrated in Figure 1. Prior to November 2018, about 45,000 liters per week of
treated sludge from the local wastewater digester was introduced into the waste on the bioreactor
side to enhance waste biodegradation, accelerate decomposition, and increase biogas production
for energy generation. This section of the landfill was permitted as a bioreactor until April 2019.
© ASCE
Geo-Congress 2020
Geo-Congress 2020 GSP 319 252
lawnmower-type UAV flights. For all surveys executed in this study, the following parameters
were used: 45 m flight altitude, moderate flight speed, 90° camera angle (nadir), 80% front
overlap, and 75% side overlap. The application is designed to remotely connect to the UAV and
command it to execute the pre-planned flight autonomously, while simultaneously capturing
photographs at the target intervals.
In addition, two high precision Septentrio Altus APS3G GNSS receivers (a base and a rover)
were used to collect ground control points (GCPs) on the ground surface within the area to be
mapped by the UAV. Twelve GCP locations were carefully selected throughout the landfill to
Downloaded from ascelibrary.org by Imperial College London on 02/23/20. Copyright ASCE. For personal use only; all rights reserved.
ensure a consistent distribution in both the horizontal and vertical planes and minimize geo-
referencing errors (Incekara et al. 2019). The base receiver was used to set a stationary GNSS
reference position, while the rover receiver was used to store the positions of each GCP with
respect to the base receiver using the real-time kinematic (RTK)-GNSS technique.
ANALYSIS
In this paper, the analysis is based on data collected from three separate field expeditions
conducted over a period of nearly a year: May 29, 2018 (988 images collected), September 28,
2018 (1439 images collected), and April 10, 2019 (1204 images collected). This allowed for two
settlement estimations: one between the first two surveys, and another between the last two
surveys. More frequent surveys could be conducted, if necessary, to measure settlements at
smaller time intervals.
3D Model Generation: The Structure-from-Motion (SfM) technique was used to create 3D
models from the UAV optical imagery using Pix4D Mapper. SfM is a technique that generates
3D models by identifying common features from overlapping imagery and computing 3D data
(Greenwood et al. 2019). The resulting 3D model is first a sparse point cloud (collection of
points in 3D space). The model is then refined using the GCPs and the camera positions from
every image to generate a dense point cloud (Incekara et al. 2019; Zekkos et al. 2018). LiDAR
(light detection and ranging) and TLS (terrain laser scanning) also produce point clouds, but both
methods are more expensive while yielding very similar results to SfM when used with high
resolution UAV imagery (Greenwood et al. 2019). The SfM technique has been used extensively
in a diverse range of applications in geotechnical engineering (Zekkos et al. 2018).
For georeferencing, eight GPS survey measurements were selected to be used as GCPs in the
SfM model, while the remaining were set as Check Points (CPs). The CPs were used after
processing to assess the accuracy of the model. Points were carefully assigned as GCPs or CPs to
ensure adequate distribution throughout the landfill.
Point cloud analyses were conducted using the open-source point cloud processing software
CloudCompare. CloudCompare was used to finely align the point clouds to each other and to
classify vegetated areas. Although the 3D point clouds were already georeferenced using the
camera positions and GCPs in Pix4D, fine tuning is still needed. The CloudCompare Align tool
was used to align the point clouds to one another by selecting four corresponding points on each
cloud and applying a transformation. For this application, it was imperative that the four points
were selected outside the active landfill area so that they were not affected by waste settlement.
Using this technique, the September point cloud was transformed to align with the May point
cloud. Similarly, the April point cloud was transformed to align to the (already transformed)
September point cloud.
The presence of vegetation can be an issue for this type of settlement analysis because
cutting grass or trimming shrubs could be falsely characterized as volume reduction due to waste
© ASCE
Geo-Congress 2020
Geo-Congress 2020 GSP 319 253
settlement. The CANUPO point cloud classification routine in CloudCompare was used to
identify and remove vegetation. First, two training datasets were generated to represent
vegetation and soil/waste. These datasets were generated by extracting subsets of the full point
cloud as examples of the types of surfaces that belong to each class. It is crucial to include
different types, colors, and textures of vegetation in the training dataset to ensure comprehensive
classification.
Different training datasets were needed for different point clouds (i.e. May, September, and
April) because the vegetation had a significantly different appearance after the summer months
Downloaded from ascelibrary.org by Imperial College London on 02/23/20. Copyright ASCE. For personal use only; all rights reserved.
versus after the winter months, as shown in Figure 2. The classifier did not perform well in
classifying all surveys using the same training dataset. Once the training datasets were
developed, the classifiers were trained and implemented for each point cloud.
© ASCE
Geo-Congress 2020
Geo-Congress 2020 GSP 319 254
values greater than 1 m (representing large volume reductions not associated with settlements
such as a stockpile or construction vehicle being moved between surveys) were identified and
excluded from the analysis. The classification rasters were also used to exclude vegetated regions
from subsequent settlement estimation.
Strain, ε v , was computed by dividing the settlement raster by the thickness raster, and Cαε
was computed by dividing the strain raster by the change in the logarithm of time (shown in the
denominator of Eq. 1) for each landfill location. To determine the t0 from which to calculate t1
Downloaded from ascelibrary.org by Imperial College London on 02/23/20. Copyright ASCE. For personal use only; all rights reserved.
and t2 , the average of (1) the date of initial waste placement per cell, and (2) the midpoint of the
two survey dates was used as the date corresponding to the mean age of waste in each cell. This
approach assumes a constant and uniform rate of waste placement per cell. Provided detailed
waste placement data, a more robust approach may be to use a weighted average to derive t0 ,
dependent on when the majority of waste was placed in each cell. Note that t0 does not
correspond to the date of initial waste placement because new waste is continuously being added.
RESULTS
Model Resolution and Accuracy: The georeferencing error for each 3D model was
calculated and is shown in Table 1, by comparing the final computed positions of the check
points against their actual positions from marked images. All three models were able to
accurately position the check points within centimeters (or less) of their markers. Additionally,
the ground sample distance (GSD) of the May 2018, September 2018 and April 2019 models are
1.61 cm/pixel (0.63 in/pixel), 1.28 cm/pixel (0.50 in/pixel), and 1.34 cm/pixel (0.53 in/pixel),
respectively.
Mean
-0.013 -0.024 -0.049 0.015 -0.020 0.028 -0.002 0.003 0.035
(m)
Sigma
0.012 0.014 0.105 0.012 0.024 0.045 0.008 0.012 0.017
(m)
RMS
0.018 0.027 0.116 0.019 0.031 0.053 0.008 0.012 0.038
(m)
When the point clouds were aligned in CloudCompare, the transformation of the September
2018 cloud produced an RMS (root-mean-square) error of 0.076 m, and the transformation of the
April 2019 cloud produced an RMS error of 0.086 m.
Vegetation Classification: The CANUPO classifier in CloudCompare was used to
accurately classify vegetation in the point clouds using a user-defined training dataset. Figure 3
displays the orthomosaic image from the May survey, for example, with its corresponding
classification output to demonstrate the achieved agreement.
© ASCE
Geo-Congress 2020
Geo-Congress 2020 GSP 319 255
Downloaded from ascelibrary.org by Imperial College London on 02/23/20. Copyright ASCE. For personal use only; all rights reserved.
Spatial Spatial
Mean Mean
Variation (σ) Variation (σ)
Figure 4 shows an example of the spatially resolved settlement and Cαε outputs which allow
for the identification of areas that are settling faster than others. Areas without data correspond to
locations where vegetation was present (and thus not analyzed) or waste was added, and
therefore settlement could not be computed. Figure 5 displays the mean Cαε values for the
conventional and bioreactor cells. The Cαε in the bioreactor side is higher than that in the
conventional side, which is expected due to the enhanced biodegradation occuring in the
© ASCE
Geo-Congress 2020
Geo-Congress 2020 GSP 319 256
bioreactor by design. Additionally, a decrease in Cαε was observed in the bioreactor side in the
second survey period, compared to the conventional side, and may be due to the fact that sludge
was no longer pumped into the landfill from November 2018 onward.
Downloaded from ascelibrary.org by Imperial College London on 02/23/20. Copyright ASCE. For personal use only; all rights reserved.
Figure 4. Spatially resolved settlement and Cαε values for September 2018 to April 2019
CONCLUSIONS
UAVs with optical cameras are a powerful data acquisition approach for monitoring long-
term landfill settlement by developing 3D models using the Structure-from-Motion technique.
The results obtained from this new remote sensing approach are spatially resolved and able to
achieve cm-level resolution and high accuracy. Additionally, data collection causes minimal
disruption to landfill operations, and eliminates the risk of damage to the instrumentation. In
agreement with most of the literature, this study has shown that the rate of long-term landfill
© ASCE
Geo-Congress 2020
Geo-Congress 2020 GSP 319 257
settlement is higher for bioreactor landfills than for conventional “dry tomb” landfills.
ACKNOWLEDGEMENTS
This research was supported by the National Science Foundation (NSF) Division of
Computer and Communication Foundations under Grant no. 1442773. Any opinions, findings,
conclusions and recommendations expressed in this paper are those of the authors and do not
necessarily reflect the views of the NSF. The authors would also like to thank the Environmental
Research & Education Foundation and the Dow Sustainability Fellows Program for their support
Downloaded from ascelibrary.org by Imperial College London on 02/23/20. Copyright ASCE. For personal use only; all rights reserved.
to Cassandra Champagne, City of Midland Landfill for allowing access to their facility, John
Manousakis of Elxis Group for his expertise and support in geomatics, and Gabriel Draughon,
Chenghang Liu, Sampurna Datta, Michelle Basham, and William Greenwood for their assistance
with field testing.
REFERENCES
Abichou, T., Barlaz, M. A., Green, R., and Hater, G. (2013). “The Outer Loop bioreactor: A case
study of settlement monitoring and solids decomposition.” Waste Management, Elsevier Ltd,
33(10), 2035–2047.
Andersen, E. O., Balanko, L. A., Lem, J. M., and Davis, D. H. (2004). “Field Monitoring of the
Compressibility of Municipal Solid Waste and Soft Alluvium.” International Conference on
Case Histories in Geotechnical Engineering, 6.
Bareither, C. A., Benson, C. H., Barlaz, M. A., Edil, T. B., and Tolaymat, T. M. (2010).
“Performance of North American Bioreactor Landfills. I: Leachate Hydrology and Waste
Settlement.” Journal of Environmental Engineering, 136(8), 824–838.
Bareither, C. A., Breitmeyer, R. J., Benson, C. H., Barlaz, M. A., and Edil, T. B. (2012). “Deer
Track Bioreactor Experiment : Field-Scale Evaluation of Municipal Solid Waste Bioreactor
Performance.” Journal of Geotechnical and Geoenvironmental Engineering, 138(6), 658–
670.
Chen, Y. M., Zhan, T. L. T., Wei, H. Y., and Ke, H. (2009). “Aging and compressibility of
municipal solid wastes.” Waste Management, Elsevier Ltd, 29(1), 86–95.
Durmusoglu, E., Corapcioglu, M. Y., and Tuncay, K. (2005). “Landfill Settlement with
Decomposition and Gas Generation.” Journal of Environmental Engineering, 131(9), 1311–
1321.
Edelmann, L., Hertweck, M., and Amann, P. (1999). “Mechanical behaviour of landfill barrier
systems.” Proceedings of the Institution of Civil Engineers - Geotechnical Engineering,
137(4), 215–224.
El-Fadel, M. (1999). “Leachate Recirculation Effects on Settlement and Biodegradation Rates in
MSW Landfills.” Environmental Technology (United Kingdom), 20(2), 121–133.
El-Fadel, M., and Khoury, R. (2000). “Modeling Settlement in MSW Landfills: A Critical
Review.” Critical Reviews in Environmental Science and Technology, 30(3), 327–361.
Fei, X., and Zekkos, D. (2013). “Factors Influencing Long-Term Settlement of Municipal Solid
Waste in Laboratory Bioreactor Landfill Simulators.” Journal of Hazardous, Toxic, and
Radioactive Waste, 17(4), 259–271.
Fei, X., and Zekkos, D. (2018). “Coupled experimental assessment of physico-biochemical
characteristics of municipal solid waste undergoing enhanced biodegradation.”
Geotechnique, 68(12), 1031–1043.
Foye, K. C., Zhao, X., Voice, T. C., and Hashsham, S. A. (2007). “Settlement Monitoring for
© ASCE
Geo-Congress 2020
Geo-Congress 2020 GSP 319 258
© ASCE
Geo-Congress 2020