Landfills are full of contaminated material, which could be harmful for the environment. To reduce the risks of pollutants entering the environment, a watertight base layer and an impermeable cover layer isolate the waste. These layers have to be maintained eternally, which is very costly. To move towards a more sustainable landfill aftercare, Dutch landfill operators have started three demonstration projects, to reduce leachate concentration in 10 years. At one of the projects, this is attempted by recirculation of leachate, which in the long term, reduces the concentrations of pollutants in the leachate. During the project the reduction of emissions via leachate is measured, but in order to make this project successful, it also has to be proofed that the emissions are reduced permanently. To proof that the emission via leachate is permanently reduced a model is required which enables to predict how remaining mass in the landfill will be emitted via leachate, based on time series of leachate quantity and quality. 
This research focuses on adapting the prediction models, created for two demonstration projects, to describe the water balance for the third demonstration field at De Kragge II in Bergen op Zoom. Compared with the other two projects, the water balance for this field is a bit more complicated, since it has a different layout of the drainage system and lateral flow to and from adjacent fields is possible. The aim of this research is to model the water dynamics in the landfill with minimal uncertainty.

Available input and output data are: rainfall, evaporation potential, leachate levels and leachate outflow, available for the pilot field and the adjacent compartment. The leachate outflow is controlled by valves, level meters and pumps in the flow system, also the operator has influence on which compartment is drained. Another complication with the outflow data is that the data from weighed trucks transporting the leachate indicate that sometimes leachate was directly pumped from the landfill, herewith bypassing the flowmeter.

The model consists of three layers, a recultivation-, waste- and drainage layer. In the recultivation layer infiltration into the waste layer is calculated by balancing rainfall, evapotranspiration and storage. The water volume infiltrating the waste layer is distributed stochastically, according to a travel time distribution, that discretizes the infiltrated water to faster and slower moving regimes. In the drainage layer model, the balance of water inflow, leachate outflow, sideflow and storage is calculated. Resulting
in a volume of water ex-filtrating the landfill. To evaluate model uncertainty, visual and quantative criteria are used. The fits of modelled on measured data is used as a visual check. The quantitative analysis consists of evaluating the Kullback-Leibler divergence and the marginalized likelihood. The Kullback-leibler divergence estimates how much information is gained from the parameters, while the marginalized likelihood determines the balance between information content and complexity. 
In order to find a model that describes the water dynamics with minimal uncertainty, three different model implementations were evaluated. In the first approach both leachate level and outflow were used for calibration with measured data. Evaluation of the model performance showed that the outflow could not be determined with acceptable error. This was indicated by large standard deviations of the model and measurement error with respect to outflow measurements. Likely the reason for this is the gap in the water balance and the erratic patterns in the outflow data. A second approach was therefore modelled in which only the leachate levels were fitted with measured data and the outflow data was given as input. This increased the leachate level fits slightly, also the quantitative criteria showed that approach 2 is better than approach 1. Some of the parameters of approach 2 had large uncertainty and the model is quite complex given the available measured data. Therefore a third, simpler and faster model was implemented. In this model the waste layer calculations were simplified using the circular convolution function of MATLAB. This function calculates the travel time distribution continuously instead of discretizing the function over given retention times, which was done in the first two approaches. This model approach gave the best leachate level fits. The Kullback-leibler divergence indicated higher information content compared to approach 2.

In addition, each approach was evaluated with different model scenarios in which the waste compartments where either coupled or uncoupled. For each approach the uncoupled models performed visually and quantitatively better than the coupled models. 
Approach 2 gave the best insight in the water dynamics given its complexity, shown by the higher values for the marginalized likelihood. Approach 3 gave the best fits, of the leachate levels, the highest 퐷_KL values and is the fastest model. From these results it would be advised to use approach 3 to analyse the water dynamics of the landfill. Given the results of the different scenarios, it would be advised to use the uncoupled models for the analysis of water dynamics inside the landfill, since these models showed the best fits, highest 퐷_KL and marginalized likelihood values. 
Based on the obtained results, the following insights could be drawn about the landfill dynamics. The sideflow between the two compartments is about 5 to 25 m3/day. The model showed that water in the landfill, flows fast from the cover layer to the drainage layer. The infiltration flux of the recultivation layer model seemed to be dominated by rainfall and evaporation. Therefore this model could be simplified by omitting water storage and flow through the layer.

The growing world population and climate change have resulted in increasing stress on the Earth’s ecosystems. Especially the overexploitation of agricultural land, overgrazing and extreme droughts have resulted in land degradation, often called desertification. This leads to the disappearing of natural vegetation, the loss of soil quality, lower production capacity in agriculture and increased water scarcity, which is most severe on drylands like the African continent. Globally, organizations such as the United Nations Convention to Combat Desertification (UNCCD) are fighting land degradation by implementing sustainable farming practices and supporting re-vegetation projects such as the Green Wall Initiative in Africa, often using ancient water harvesting techniques. Justdiggit, a Dutch NGO, is currently working on land restoration projects and dug over 72,000 water-retaining semi-circular bumps, or ’bunds’, in Kenya. By retaining rainwater in these bunds, water is given the time to infiltrate into the soil, erosion rates by overland flow are reduced, while vegetation recovers and on the long termcan take over the function of the bunds. Although the first results of the bunds at small scale are promising, the impact of the projects of Justdiggit has never been quantified on a large scale in terms of the amount of water that’s retained, increase of vegetation and decrease of surface temperature.

As dense in-situ networks are expensive and have difficulties in capturing the large spatial variability of parameters such as soil moisture and temperature, often remote sensing (satellite observation) is used, which does well in measuring spatial variability on large scales. However, as optical remote sensing is often affected by cloud cover, data availability is limited. Furthermore, it is difficult to translate optical vegetation parameters such as the Normalized Difference Vegetation Index (NDVI) to vegetation biomass, as these only observe the surface of the canopy. Passive microwave remote sensing, which is measured in the microwave part of the electromagnetic spectrum, has proven to be very accurate in determining parameters such as soil moisture content, surface temperature and vegetation optical depth (VOD, related to vegetation thickness and water content). The advantage of using microwaves is that it can ’see’ through clouds and can therefore be used under almost all atmospheric circumstances. However, due to the low intensity of passive microwaves, resolutions are low (¼ 50x30km). VanderSat, a Dutch remote sensing company, downscales the low-resolution passive microwave observations to 100x100m field-scale resolution. The goal of this research is therefore to test if the project areas of Justdiggit can be monitored using these downscaled passive microwave remote sensing soil moisture, surface temperature (night-time) and vegetation optical depth products of VanderSat.

Using in-situ data from sensors that were installed inside and outside the bunds during a fieldwork in Kenya in March 2018, the hydrological behaviour of the bunds was investigated. Furthermore, a cross-validation was carried out to compare the downscaled passive microwave soil moisture, surface temperature (nighttime) and VOD products to the in-situ sensors, thermal derived surface temperatures (day- and night-time) from MODIS and NDVI fromSentinel-2 and MODIS. After these cross-comparisons, the differences between the project areas of Justdiggit and the reference area were analysed using the downscaled passive microwave products. Finally, using the Tau-Omega model, the passive microwave signal from the bunds was simulated.

All in all, the analysis of the results have shown that the bunds have a significant effect on the soil moisture content, especially at the deeper soil layers, based on the in-situ sensors, while results of NDVI have shown a clear contrast in vegetation ’greenness’ between the project areas and the reference area. Due to the short measuring period, the correlations between the in-situ sensors and the downscaled soilmoisture and surface temperature products were still low. Furthermore, highest correlations were found between the downscaled surface temperature product and the night-time surface temperatures of MODIS. As it turned out, no significant differences were found between the project areas and the reference area based on the downscaled passive microwave products. This, in spite of the fact that the soil moisture, surface temperature and VOD products have shown to be consistent. Finally, modelling of the microwave signal of the project areas, has shown that the effect of the bunds is still too small in terms of emissivity and is therefore not being measured yet. Increasing the current percentage of bunds from10 to roughly 20%, or waiting five years after implementation of the bunds in 2016, should give significant differences in terms of soil moisture, surface temperature and vegetation optical depth, as a result of vegetation development and land restoration in the project areas.

Evaluation of Distributed Temperature Sensing to monitor temperatures within a landfill on a field scale.