This thesis investigates the efficacy of artificial intelligence (AI) models, particularly convolutional neural networks (CNNs) and U Net architectures, in reconstructing datasets with missing velocity data in river flow analysis. Optical flow and Particle Image Velocimetry (PIV) techniques have emerged as valuable tools for analyzing river flow patterns.
Through a comprehensive literature review, CNNs, and UNet models are identified as promising tools for this task due to their ability to capture intricate patterns in datasets. The study compares the performance of the U-Net model against a statistics-based hydrological benchmark model, revealing the superior performance of the U-Net model.
Furthermore, the analysis explores how the performance of AI models varies with differing quantities of missing data, by masking available data and comparing reconstructed values against the ground truth, highlighting the importance of data availability.
Additionally, the study investigates the influence of spatial patterns in training data on model performance, including patchy versus random missing data in the field of view, simulating more datasets more likely available in reality. This clarifies the challenges encountered in predicting grid points under different training dataset conditions.
Finally, the study identifies areas within the dataset that are particularly challenging to predict, shedding light on factors contributing to prediction errors. These findings underscore the potential of AI models in hydrological applications and provide valuable insights for future research in the field.
Our findings show that U net is capable of reconstructing velocity fields from a river flow better than an average benchmark that uses the average values, with varying accuracy depending on input data.
The average benchmark model had a relative error close to 0.2 in every instance, whereas the U-Net model showed relative errors ranging from 0.085 to 0.006. Errors from a patchy mask are ranging from 0.09 8 to 0.031.

Water is essential for life on earth and is vital for numerous sectors of our society. Pressures arising from climate change, growing populations, and the shift towards clean energy accentuate the importance of effective water management. To make decisions about water resource allocation, hydraulic modeling studies have been undertaken on large African rivers. These studies employ global terrain models, utilize remotely sensed water surface elevation data, and often involve estimating river channel bathymetry. Typically, these models also require an estimation of the river channel bathymetry. However, observed bathymetry data is seldom available and crucial for the hydraulic model performance. It has resulted in a key challenge of modelling large rivers in this data-sparse context.
This research aims to model a medium to large sized river with wide floodplains in three dimensions by integrating discharge data and a highly accurate bathymetry. The primary objective is to quantify the friction coefficient and establish a reliable rating-curve for the river system. By utilizing these key components, the study seeks to provide a comprehensive understanding of the hydraulic behavior of the river, contributing to improved water flow predictions and management strategies. The bathymetry data is acquired through two different methods. The dry bathymetry is obtained using an UAV (DJI Phantom 4) and photogrammetry (WebODM). The wet bathymetry data is collected using both, sonar with the Deeper Chirp+ and spatial referencing with the RTK-GNSS from ArduSimple. These methods are cost-effective and require minimal manpower, making them practical options for acquiring accurate bathymetric information. The discharge data is acquired using the open-source software, OpenRiverCam. OpenRiverCam uses Large Scale Particle Image Velocimetry (LSPIV) to determine the surface velocities and combines the results with the bathymetry data to calculate discharges, providing an efficient solution for assessing river flow characteristics. LSPIV has the advantage that it is a non-intrusive method of measuring the flow velocity and does not require physical probes or instruments in the water. The bathymetry data and discharge data are integrated into the Delft3D FM Suite to assess the accuracy of the measurements and estimate the friction coefficient in both the river and the floodplain. This modeling approach enables a comprehensive analysis of the hydraulic characteristics of a medium to large sized river and supports the evaluation of flow resistance in the study area.
The data acquisition took place at three study sites close to the Bui Dam, in the Black Volta Region, Ghana. The Bui Dam is the second largest hydro-power dam in Ghana managed by the Bui Power Authority (BPA). The Bui Bridge and Bamboi Bridge study sites are positioned downstream of the Bui Dam, allowing for accurate quantification of the discharge and the bathymetry measurements. The third study site, Chache, is positioned upstream of the dam, where daily water level measurements are taken. BPA has observed that the rating curve at this location is outdated. Therefore, efforts are made to update the rating curve and quantify the friction coefficient at this site in both the river and the floodplain.
This research has made significant progress in developing a three-dimensional discharge model and rating curve for medium to large rivers using advanced data collection methods and integration techniques. The study successfully combined photogrammetry and sonar measurements to effectively determine the bathymetry of the river, overcoming challenges related to high water velocities and dense vegetation. The LSPIV technique and OpenRiverCam were utilized to integrate surface velocities and discharge measurements, leading to a more comprehensive understanding of river dynamics. However, limitations were encountered in assessing the accuracy of the model at the Bamboi Bridge site due to the LSPIV results. This highlights the importance of obtaining more comprehensive data and observations to enhance the model’s accuracy. The comparison of rating curves at the Chache site resulted in positive results. Although, further verification during the wet period is required through velocity and discharge measurements to determine the accuracy. Overall, this research contributes to a better understanding of river behavior and provides valuable insights for water flow prediction in an efficient, cost-effective manner with minimal intensive manpower, ensuring a non-intrusive approach.

Small to medium-sized man-made freshwater reservoir are a reliable source for drinking water supply, hydropower generation and irrigation purposes worldwide. However, water volumes in these reservoirs can be significantly affected by prolonged droughts, resulting in severe impacts on society (Kozacek, 2014; Mahr, 2018). To mitigate the impact of such events it is crucial for decision makers to know when the available water resources are lacking. Although many reservoirs are closely monitored, this data is not always readily available. Inadequate information sharing, inaccessibility, and a lack of tools to predict future reservoir storages contribute to this problem.

Remote sensing has the potential to address this problem. The Global Water Watch is a platform that and provides earth-observed surface area dynamics that can be used to monitor small to medium-sized reservoirs worldwide and detect trends in water availability. While this method serves as a valuable indicator of water availability, it falls short in providing decision-makers with the necessary absolute volume time series and volume predictions. Currently, no platform exists beyond in-situ measurements to meet this essential need.

This thesis presents a novel method for retrieving near real-time volume time series in small to medium-sized man-made reservoirs worldwide using remotely sensed open data. The method utilises the MERIT-Hydro digital elevation model, HydroMT and stream flow methods by Eilander et al. (2023), and literature by Messager et al. (2016) to reconstruct reservoir bathymetry. This novel approach in reconstructing reservoir bathymetry enables the conversion of available reservoir area time series into volume time series. These were employed in autoregressive and multi-linear regression models to predict water availability up to six months in advance. The models incorporate ERA5 precipitation data by Hersbach's (2020) and the Standardised Precipitation and Evaporation Index (SPEI) by Beguería et al. (2021) to improve the accuracy of the volume predictions.

When comparing the novel method to the method proposed by Messager et al. (2016), the novel method yielded more accurate reservoir volume estimations. The method successfully obtained bathymetries and accurate volume estimations when validating using 2 reservoirs in Zambia and 48 in India, demonstrating the potential of this novel approach. However, some reservoirs with complex shapes faced initial delineation challenges, resulting in inaccurate volume predictions. These issues could be resolved by manually delineating the area for bathymetry reconstruction. Moreover, regression models were applied to case study reservoirs in Eswatini and Lesotho, demonstrating reasonable predictive capabilities with the Heidke Skill Scores ranging from 0.77 to 1 for up to 2 months ahead. However, precise prediction of extreme decreases in reservoir levels requires a physically based approach that incorporates the volumetric time series provided by this novel method. The study emphasises the necessity of considering the volume time series’ memory to predict water availability and provides a valuable foundation for volume time series analysis using remotely sensed data.

The unavailability of consistent accurate river flow data is a significant impediment to understanding water resources availability, and hydrological extremes. This is particularly true for remote, difficult to access, morphologically active and therefore rapidly changing rivers. The state of global river discharge monitoring with respect to water infrastructure and frequency of data collection has been on the decline over the past few decades. This is despite the significant importance of these data for river flow predictions. Fortunately, rapid advancements in technologies open up possibilities for water resource authorities to increase their ability to accurately, safely and efficiently establish river flow observation through remote and non-intrusive observation methods. Low-cost Unmanned Aerial Vehicles (UAVs) in combination with Global Navigation Satellite Systems (GNSS) can be used to collect geometrical information of the riverbed and floodplain. Such information, in combination with hydraulic modelling tools, can be used to establish physically based relationships between river flows and permanent proxies. This study attempts to monitor flow in volatile, dangerous and difficult to access rivers using only affordable and easy to maintain new technologies. This thesis consists of three main components: i) generating a workable framework for monitoring rivers using low-cost technologies; ii) establishment of river geometry using a combination of airborne photogrammetry and low-cost GNSS equipment iii) and physically based rating curve development through hydraulic modelling of surveyed river sections.

The first three chapters of this thesis provide an introduction in the form of a literature review, justification for the study and a description of the study area. In chapter 4, a framework is developed through an intensive review of traditional river monitoring processes. Uniquely effective and low-cost individual components are selected and placed within a framework. The ideal outcome is an interconnected framework which clearly presents the steps which are necessary for river monitoring in remote locations. The manner in which each critical step is related to the other is explained. Furthermore, the method by which modern technologies are assimilated into the method is described. Within the framework, critical thresholds are set up in order to signal the to the water manager whether the proposed model in its current state continues to perform as required.

Chapter 5 investigates how low-cost technologies such as UAVs in combination with low-cost GNSS devices can be used to generate river geometry for the purposes of application in a hydraulic model. Furthermore, performance of the open-source photogrammetry software substantiated the claim that, free and open-source available packages are capable of producing results which are as good as proprietary alternatives as shown by the RMSE analyses. A novel approach to generate a seamless bathymetry through merging and volumization was successfully tested. Results presented in this chapter encourage future studies to investigate the impact of variations in the number of Ground Control Points (GCPs) on discharge estimations in a hydraulic model with different hydrodynamic boundary conditions. This follow up was instituted in Chapter 6.

In this sixth chapter we accept that uncertainties in the data acquisition may propagate into uncertainties in the relationships found between discharge and state variables. This uncertainty prompts the need to understand the impact of varying geometries on hydraulic models. Specific attention is placed on variations caused by differing GCP numbers since the task of GCP placement is time consuming, potential dangerous and resource intensive in certain location and instances. We are successfully able to determine the minimum number of control points required to reproduce geometry. Overall, we successfully develop and test a workable method for water resources authorities to estimate river flows accurately through the application of advanced, low-cost technologies with minimal contact with measured variables.

The development and application of low-cost technologies for river flow monitoring has led to the following important conclusions:
• For the purpose of flow estimation, there is no need to use more than seven GCPs to establish accurate UAV-based geometry. Rather, it is more crucial to distribute the available markers to be maximally representative of the terrain elevations. Furthermore, it may be necessary to place more markers in close proximity to locations where one may expect the largest challenge for photogrammetry software (e.g.: water, thick forest/vegetation)
• In order to limit the impact of the “doming” effect on terrain geometry measurements, one of the most effective, yet easily implementable mechanisms is to measure a river line using Real Time Kinematic (RTK) Global Navigation Satellite Systems (GNSS) equipment. This data can then be used to correct the terrain post photogrammetry processing.

Weather generators (WGs) based on temporal resampling algorithms are not able to generate new extremes with the same or a higher temporal resolution as the historical data series of the thus far
observed precipitation. This is problematic for catchments with a time of concentration shorter than the resolution of the used data and conflicts with the increasing occurrence of more local, short duration extremes not yet observed.

In this thesis, it was researched if spatial permutation of precipitation could provide a solution to these problems by introducing historical events from related locations into the area of interest. The generated precipitation series were expected to have a larger variety of precipitation events compared to the historical data, thereby representing the current changing weather patterns better and being more suitable for small basins. This is beneficial for insurance companies, governments and aid organizations which rely on long term precipitation series to generate event catalogues and risk predictions.

To develop, improve and widen the knowledge about the effects of spatial permutation, four different questions were formulated for a case study on spatial permutation in the Rhine basin. A literature study showed that precipitation regimes in Europe can be defined based on spatial and temporal variability, precipitation amounts and the influence of controlling factors like atmospheric circulations, topography and climate change. This information was used to built three permutation models. The first model shifted historical precipitation fields over fixed distances and directions. In the second model this fixed approach was replaced by semi-random vectors including spatial and temporal correlation. The last
model used a vector approach with vectors conditioned with historical wind data. The effect of each model on the Generalized Extreme Value (GEV) distributions, the cumulative distribution functions (cdfs) and the main characteristics of precipitation for different basins and aggregation times was determined. The July 2021 Meuse flood was used as example to show the working method of each model visually and to better understand the effect of each permutation model on individual extreme events.

The results showed that spatial permutation did influence precipitation patterns, characteristics and statistics. Strongest changes in extreme precipitation events were seen for small basins and short aggregation times. The permutation direction and distance were important determinants for the outcome of each model. Precipitation permutation with semi-random vector fields was shown to be a promising method which allowed for the inclusion of both spatial and temporal correlation. However, the model had a high sensitivity to the initial and boundary conditions. Wind based vector fields were able to replicate the most important historical precipitation characteristics while at the same time generating new extremes. Yet, a clear trade off was visible between similarity of the historically observed and modelled precipitation characteristics and the number of new extremes introduced.

With the knowledge obtained, it can be concluded that spatial permutation is a promising method to generate more divers precipitation time series for the Rhine basin. Both semi-random and wind-based vector permutations can already be used to generate new precipitation series as long as the initial and boundary conditions are chosen carefully. To improve the results, a fusion of spatial and temporal relations and a physically wind-based vector generation method is advised. In addition, possibilities are seen in a combination of the currently used temporal resampling algorithms and a spatial permutation approach. The outcomes of these suggestions are not known yet. Nevertheless, it is expected that a better understanding of spatial permutation, on top of the results presented in this thesis, can advance the current methodologies used to generate long term precipitation series. Therefore, it is hoped that this research provides the incentives to explore spatial permutation of precipitation patterns in more detail and in such, contributes to a more accurate risk profile for the livelihoods of people worldwide.

Rivers have long since exceeded their natural purpose of discharging excess water, by becoming subject to many practical applications demanded by present day society [61]. In order to comply withthis variety of needs and demands, the necessity for proper water management arises, which in turn requires data and knowledge of hydrological parameters like water levels, water quality and river dis-charge [50]. This research focuses on the hydrological data demand and specifically on the measurement of riverdischarge. Discharge is generally estimated with intrusive measurement methods [64], this means that the measurement device is in physical contact with the water which can be difficult in strong current or high discharges and even dangerous during floods . Furthermore, in remote and low-resource settings, collecting discharge data is compromised by accessibility problems and difficulties maintaining and acquiring monitoring equipment. When numerous measurements are performed, it is common practice to establish a stage-discharge relationship [66] to facilitate discharge determination, i.e. by shifting to stage measurements. However, due to the empirical character of the method and the sporadic occurrence of high discharges, the relationship can contain considerable uncertainties for these higher discharges [67]. The aim of this research is to provide a sustainable and low-cost data collection and processing method in order to establish a rating curve based on a three dimensional hydraulic modelling approach. One of the main processing methods is Large-Scale Particle Image Velocimetry (LSPIV). LSPIV is a computer based technique that computes flow velocities at the river surface based on video images. Hence, with the development of such a model a more physically based stage-discharge relationship can be determined based on non-intrusive measurements, meaning that measurements can be taken during safe (low flow) conditions in a restricted amount of time. Furthermore, due to the sole use of relatively simple methods and the limited amount of observations needed, this method is particularly suitable for remote and low resource settings. The study is based on data collected during a two month field trip at the Luangwa river in Zambia. The dataset consists of point clouds collected with the aid of photogrammetry, sonar and RTK GPS which are used to create a bathymetric chart, videos recorded with a drone for the computation of the surface flow velocities, surface flow velocities measured with a current meter for LSPIV validation and discharges measured with an ADCP. The bathymetric chart is used as bed level for the three dimensional discharge model created with Delf3D D-Flow FM which is calibrated with the surface flowvelocities (LSPIV) and ADCP discharge measurements. The discharge model represents approximately 9.2 kilometres of the Luangwa river in length and can reach a maximum width of about 390 metres. The model is calibrated at a discharge of 191 m3/s by minimising the difference between measured and simulated values of ten surface flow velocities and five water levels. This resulted eventually in a Manning friction coefficient of푛= 0.014 s/m1/3. The calibrated model resembles the actual river in location, depth, width and surface flow velocity. The LSPIV velocities are approached to a mean average deviation of 0.07 m/s (1.1 m/s average) and the water level deviates 0.06 m at the research area (1.3 m average). The model is used to establish a stage-discharge relationship which is subsequently compared to two existing relationships, one based on a similar approach using a 1D model and one based on stage-discharge data measured at a conventional gauging station. The three stage-discharge relationships are in the same order of magnitude although the geometry of the river at all sites is likely to be different. Since a stage-discharge relationship is heavily dependent on the geometry [66] this comparison is only a rough indication of the accuracy. Ideally, the discharge, water level, and surface flow velocity should be measured for different discharges and compared (using the model) to the established relationship. The stage-discharge relationship could, if needed, be adjusted based on the new measurements.

To develop warning systems for flood events, create precipitation-runoff relationships, validate runoff models, or to understand the behaviour of rivers, understanding of the amount of water flowing through rivers is needed. The low-cost and novel gauging method using Large-Scale Particle Image Velocimetry (LSPIV) could complement discharge measurements at locations and stages where the possibilities of using traditional gauging methods are limited. This report investigates the feasibility of using LSPIV to quantify river discharges in an equatorial urban stream with fast response time.

LSPIV uses videos to extract surface flow velocities by tracing movements of seeds on the water's surface. Combined with the local bathymetry and water level, an estimation of the river's discharge can be made. This study consists of two sets of experiments. The first set of experiments were performed at the Dommel regarding processing software, image preparation, seeding densities, and point of views and discussed by assessing their accuracy relative to benchmark measurements -- using the mean error and root mean squared error -- and the method's precision – using the relative standard deviation.

The second set of experiments were performed along the Chuo Kikuu, Dar es Salaam, Tanzania. A flood wave was monitored through the capture of 73 5 second videos. These videos were turned into separate frames and corrected for lens distortion and perspective distortion. Thereafter the frames were gray scaled and gamma correction was applied. After the LSPIV process additional filtering removed unrealistic low flow velocities, and through substitution missing velocities were replaced with flow velocities based on the vertical logarithmic progression relationship between the surface flow velocities and water depth. The surface flow velocities found using this method match optical observations. Discharges were estimated using the empirical depth-average coefficient and local bathymetry. Results showed that the post-processing reduces the uncertainty bandwidth with 37% and increases the mean flow velocities with 96%.

The found discharges were compared with precipitation measurements observed at a nearby TAHMO meteorological station. The total volumetric precipitation was determined by estimating the contributing catchment using a digital elevation map and the locations of man-made drainage systems. When comparing the volumetric precipitation with the flood wave, a runoff coefficient of 53% [35-68] is found. This coefficient falls within the ranges found in literature, but is probably an underestimation of the true runoff due to an overestimation of the catchment area and underestimation of the discharges.

This study shows that the LSPIV method is feasible for continuously monitoring flood waves in an urban environment. Especially during peak flows LSPIV proves to be valuable, as observations using conventional gauging methods are labour intensive, unsafe, or not executable. Because of the possibility to monitor streams from a distance -- which ensures access to power and safety against vandalism -- there is a possibility to observe complete flood waves at regular intervals without the need for direct contact with the water. For Dar es Salaam, this method opens doors for continuous and secure stream monitoring, at low costs and with local devices.

Humans have always populated in the vicinity of river systems, where thesupply of water, nourishment and transportation is obtained from the river.However, inundation is a re-occurring problem and impact of floods are ex-pected to increase due to climate change. Accurate flood forecasting andearly warning is critical for disaster risk management. Tackling the problemof forecasting, in data scarce environments, has become increasingly impor-tant due to the changing climate. Remotely sensed river monitoring can bean effective, systematic and time-efficient technique to monitor and forecastextreme floods. Conventional flood forecasting systems require extensivedata inputs and software to model floods. Moreover, most models rely ondischarge data, which is not always available and is less accurate in a over-bank flow situations. There is a need for an alternative method which de-tects riverine inundation, using open-source data and software. This thesisaims to research the use of passive microwave radiometry for the detection,classification and forecasting of inundation.Brightness temperatures are extracted from the passive microwave radiom-etry and are converted in a discharge estimator: the C/M-ratio. Surfacewater has a low emission, thus let the C/M-ratio increase as the surfacewater percentage in the pixel increases. Sharp increases are observed forover-bank flow conditions. The research combines the identification of in-undation with a probability analysis via a quantile regressional fit. Floodforecasts can be obtained from an upstream catchment area. In the mostideal situation with a delay of2,5hours. This allows for probabilistic earlywarning decision making, with a lead time up to14days. (location specific)Strong Spearmans correlation coefficients between the discharge and C/M-ratio are found (>0.883). Allowing the model to forecast floods as gaugeddischarge records do. The model used has a comparable skill to the localGloFas forecast. This research investigated the impact the remote sensedtechnology could have on the flood forecast, response and warning system.An added model to an Early Action Protocol has the ability to lower uncer-tainty within decision making and enlarges the intervention window. Theadvice is to use such a model in combination with other forecasting modelssuch as GloFas.The challenge using this technology is the integration of hydrological com-plexity. The method allows for automated, global-covered creation of gridbased flood forecasts, independent to cloud coverage. Creating low spatialresolution flood forecasts combined with a probability bound in hours aftersatellite detection. The method has a high potential for data scarce flood-prone river basins around the world. The future for this technology lies inthe global daily availability of the data. With satellite sensors improving,spatial resolution is expected to increase. Allowing for even better floodforecasting ability.

Big data sources can play an important role in revolutionizing the field of water resources research. Time series data with high temporal and spatial dimensions encapsulates with itself numerous factors essential for coming up with robust decisions. In this thesis, we assess one such big data source for efficient water management in the Oum Er Rbia basin, Morocco. The surface water detection technique furnished used in this thesis is found to be accurate in detecting the surface water sources and its temporal and spatial dynamics.
The remotely sensed time-series data of reservoir area was used to come up with Level-Area-Storage(LAS) relationships for the five main reservoirs in the Oum Er Rbia basin. These curves were able to approximate the present set of LAS curves well. Hence, were used in place of the local LAS curves in a water allocation decision model called RIBASIM. Thus, we had two scenarios one where the local LAS curves were used to optimize reservoir operations and the other where remotely sensed LAS curves were used instead of the local LAS curves.
The operating rule curves in the water allocation decision model were then optimized for the two scenarios. The optimization was done to maximize the performance of the system across three objectives: a)public water supply, b)irrigation and c)hydroelectricity generation. A trade-off between the three objective functions was then shown using parallel and scatter plots. It was observed that for the same set of LAS curves the performance across all three objectives improved post-optimization of the operating rule curves. This showed that there were rooms for improvement in the existing reservoir operating rule curves. The operating rule curves for the water allocation decision model with remotely sensed LAS curves were then optimized. The best set of operating rule curves that we got from the second optimization were then used with the local LAS curves to see how the system would perform with these operating rule curves. This gave us an idea of the feasibility of using remotely sensed data to come up with water management decisions and also to assess the benefits of using remotely sensed time-series data. Though the performance over the three objectives was not as good as the results we got by optimizing the system with local LAS curves, it was better than the system performance across the three objectives with the existing set of operating rules and local LAS curves. Thus, it can be used when there is a dearth of proper LAS curves.
Besides, optimizing operating rule curves, the remotely sensed time-series data of reservoir surface area was used to assess the effects of sedimentation in the reservoir storage. It was observed that for larger reservoirs the percentage change is not much as compared to the smaller reservoirs. Apart from the size of the reservoir, more study is required to make a detailed analysis of how factors like topography and soil texture influence the rate of sedimentation. Despite its limitations, the remotely sensed time-series data of reservoir surface area can be used to perform a qualitative analysis of the rate of sedimentation and can give reservoir authorities an idea of the need for bathymetry. This can help in avoiding unnecessary bathymetries which are infeasible both economically and physically.

With climate change increasing its mark on all aspects of the hydrological cycle, societies all over the world living in flood-prone areas are increasingly exposed to flood hazards. In many parts of the world, especially in less developed areas, societies lack knowledge and data to predict future flood events. By predicting a future flood event, an organization creates a time frame in which it can implement a mitigating action that reduces the financial damage inflicted. In recent years, development in new measuring techniques has significantly lowered the cost of collecting data and information on different aspects of the hydrological cycle. These developments enable organizations in regions restrained of knowledge and data to establish methods to analyze aspects of the hydrological cycle and thereby predicting the probability of a flood hazard several hours or days in advance. This thesis explores various possibilities of designing and implementing an \ac{EWS} for the \ac{BRT} in Dar es Salaam. The EWS design is based on the forecasting requirements, investigated with the BRT-system. Several operational forecasting methods are available. The EWS designed in this thesis makes use of rainfall data obtained from rainfall stations located in the Dar es Salaam region, installed and managed by the \ac{TAHMO}. This forecasting data is chosen because it provides the needed lead-time with the lowest margin of error. This forecasting data is processed and analyzed by the designed EWS and subsequently produces a probability level on a flood event. It thereby provides an advice on if the BRT-system should implement a mitigating action based on the principle of pursuing an optimal economic outcome. The designed EWS produces the flood probability in real-time, updated every hour with a lead time of one hour. This time frame enables the BRT-system to implement a mitigating action, thereby reducing the inflicted cost.

The probability level of a flood event is determined by training the EWS with historic flood and rainfall data. In addition, the implementation of both a hydrological and relational model in the EWS was tested. The results show that the hydrological model is the better option. The results also show that the implementation of an EWS ensures a decrease in financial damage endured by the BRT-system. The produced outcome of the EWS was validated by a 'leave one out' method. This validation was done by consecutively leaving one flood event out of the historical data frame and analyzing the variability of the resulting outcome. Finally, the designed EWS is best implemented in the BRT-system alongside the EWS-systems currently in place.

Most people of Malawi are dependent on rainfed agriculture for their livelihoods. This leaves them vulnerable to drought and changing rainfall patterns due to climate change. Over time, farmers have adopted local strategies and knowledge that help reducing the overall vulnerability to climate variability shocks. One other option to increase the resilience of rainfed farmers to drought, is providing forecast information on the upcoming rainfall season. Forecast information has the potential to inform farmers in their decisions surrounding agricultural strategies. However, significant challenges remain in the provision of forecast information. Often, the forecast information is not tailored to farmers, resulting in limited uptake of forecast information into their agricultural decision-making. Therefore, this study explores whether drought forecast information can be linked to existing farmers strategies and local knowledge on predicting future rainfall patterns. During a period of three months in Malawi, participatory research approaches are used to create an understanding of what requirements drought forecast information should meet to effectively inform farmers in their decision-making. Consequently, a sequential threshold model was established that relates annually monitored meteorological indicators before the rainy season, to the occurrence of dry conditions during the season. Dry conditions were expressed in the drought indicators that farmers require for their agricultural decision-making. Additionally, using interviews among stakeholders and a visualisation of the current information flow, further insights on the current drought information system were developed. Although farmers have their own strategies and timing of decision-making, this research has generalized some of the opinions and strategies to develop the ‘requirements’ which a contextualized forecast should meet. In August farmers require a prediction of the onset of the rainy season, typically starting mid-November. In addition, an update on the timing of the onset of rains is required in beginning of November. An overall indication of the ‘dryness’ of the rainy season is required in September. Here, ‘dryness’ is characterized by the number of dry spells, a composite ‘drought index’ of associated rainfall variables by the farmers. The forecast should be on a scale that is locally relevant (EPA level). This research consequently established a forecasting model, based on meteorological variables from local knowledge which can complement the forecast variables from the DCCMS. The results of forecast verification show that meteorological indicators based on local knowledge have a predictive value for forecasting drought indicators. Subsequently, skill analysis of forecasting incorporating all the above dimensions shows that the accuracy of the forecast differs per location with an increased skill to the Southern locations. In addition, it is also location dependent whether the contribution of wind, temperature or ENSO indicators gives the most predictive value. The results show that a combination of all indicators have the best predictive value. In addition, the results show that local knowledge indicators have an increased predictive value in forecasting the locally relevant critical events in comparison to the currently used ENSO-related indicators by the DCCMS. Additional research is needed to further analyse certain aspects of this research, such as research on the robustness of the model used. Research on the risk farmers are willing to take in their respective decisions could act as another requirement the forecast skill should meet. This highlights the importance of having continuous feedback from the farmers, since farmers may experience adverse impacts from wrongly informed decisions. Despite these limitations, it is argued that the inclusion of local knowledge in the current drought information system of Malawi may improve the provision of forecast information for farmers and shows that it is possible to capture local knowledge in a technical approach. The findings have relevant implications for other stakeholders, such as humanitarian and meteorological organisations, that are implementing drought-risk reduction approaches and climate services.

The current intensification of the hydrologic cycle, in combination with expanding settlements in flood prone areas, makes an increasing share of the global population exposed to flood risks. Many parts of the world are, however, still lacking the data needed for flood risk management and risk reduction. The recent development of information and communication technologies has remarkably lowered the costs to collect data for flood resilience, which has accommodated the rise of community mapping projects to fill data gaps in resource-strained environments. This thesis utilises drainage data collected by the Ramani Huria community mapping project in Dar es Salaam, Tanzania, to investigate if community mapped drainage data can improve flood predictions on neighbourhood scale. A coupled 1D-2D hydrodynamic model is developed of Kijitonyama ward, and is run with and without Ramani Huria’s drainage data implemented in the 1D schematisation. The simulated flood depth for the scenarios is validated with citizen’s observations on flood depth during a rain event on 3 March, 2019. The developed model is then applied to investigate the impact of solid waste accumulation in the drainage system on floods, by closing the drainage segments that were recorded as blocked in Kijitonyama ward by Ramani Huria staff the morning after the simulated event. An experimental scenario is also run, to examine the impact of blocked culverts. The results show that community mapped drainage data indeed can enhance the performance of hydrodynamic models, as the model output corresponds better with the validation data when implementing Ramani Huria’s drainage data in the 1D schematisation, compared with a scenario run with only a 2D schematisation. The scenarios run with solid waste blockages do not influence the model output when comparing with citizen’s observations, but increase the water level in the drainage segments located upstream of the blockages.

River data on discharge and characteristics is essential for water management and water supply, as well as for flood prediction and flood control (Pan, Wang, and Xi 2016). In practice, many watersheds are ungauged due to high costs, inaccessibility and even due to political instability (Pan, Wang, and Xi 2016). For this reason, measuring remotely without the need of being physically present, for instance by remote sensing satellites, can be interesting for many applications. The large amount of satellite data can result in the ability to extend short observation series into larger series with satellite missions.
Discharge is one of the conditions in a river, which is relevant to have data on during regular periods but in particular during or after extreme events. This thesis focussed on an approach, by using remote sensing, to obtain data that can be used for further research to determine discharge. River-width is one of the current variables researched to be used as a substitute for river stage data. River stage is currently used to obtain estimations for river discharge via earlier obtained river stage-discharge relations, which can be transformed into river width-discharge relations.
The objective of this thesis was to develop a method to obtain sub-pixel accurate river-width estimations by remote sensing. The objective to estimate river-widths on sub-pixel base originates from the need of river-width estimations with higher accuracy than the freely available optical satellite resolutions of 10 to 20 metres. The study contains the improvement of the current water classification methods by including analyses for discriminating band combinations, to construct site-specific indices. This was noticed to be needed, due to the conventional indices, like the NDWI, performing differently with the presence of certain land types.
By having multiple indices based on uncorrelated satellite bands transformed into probability bands, it is possible to combine indices, via Bayes theorem. Based on the site-specific indices and index combinations, the aim is to develop relations between spectral information and water fractions of pixels that could lead to a more detailed river-width estimation by including sub-pixel information.
The resulting method was able to show discriminating abilities in satellite bands and band combinations, specifically for an area of interest, other than the conventional NDWI and MNDWI. With the use of river edge information, the spectral bands could be transformed into spatial water probability bands, indicating a probability for the present pixels to be water. The probability indices and index combinations showed to reduce a large part of the occurring misclassifications. With the use of ROC curves, to assess the classification performance of the indices and combination of indices, variation in misclassification of certain land types between days were observed for certain indices.
The probability bands, which are based on the river’s edge value distribution, also seemed to be useful, especially for the pan-sharpened MNDWI and the Bayes 0-3 indices, to obtain the needed water fraction relations for sub-pixel base estimations. A comparison in river-width estimation of a conventional automated water classification method; Otsu’s thresholding method and a Supervised training map classification method were made against the use of probability indices with sub-pixel water fraction relationships. It was found that the use of sub-pixel information resulted in a significant improvement of the accuracy for river-width estimations. For the first and second fieldwork day, the average river-width deviations of the pan-sharpened MNDWI and Bayes0-3 decreased, respectively, from 16 and 9 metres, to under 5 and 7 metre deviation by including the found water fraction relationships.

In 2008 the Red Cross Red Crescent (RCRC) started with Forecast-based Financing pilots to improve existing Early-Warning Early Action systems. Forecast-based financing is a new methodology to prepare, deliver and respond in a more effective and efficient manner, based on hazard forecasts. Actions are triggered when a forecast exceeds a danger level in a vulnerable intervention area. Forecast-based financing consists of several implementation steps, of which the first three aim at impact-based forecasting. Therefore, In this study we investigate how forecast skill of agricultural drought forecasts can be achieved. More specifically, the aim is to identify the contribution of machine learning and satellite-derived products in early warning early action systems improving the forecast skill of agricultural drought forecasts. We explore this through a machine learning model for a case-study area of the Lower Shire River Basin in Malawi. Several experiments with different sets of predictors and predictands are conducted to test which data adds to the skill and at what spatial detail. As predictors, the following agro-climatic indices are used: cumulative precipitation, soil moisture anomalies,mland surface temperature anomalies, El Niño Southern Oscillation in July and four different dry spell categories within the growing season (0-2 days dry spell, 3-4 day dry spell, 5-10 day dry spell and larger than 10 day dry spell). As drought predictand, the normalized difference vegetation index (NDVI) and the vegetation optical depth
(VOD) in March are used, the latter obtained from satellite data company VanderSat. The final set of predictors and predictands is narrowed down based on which data is available and with which quality (timeliness, reliability, accuracy). Initial results, show higher accuracy and weighted accuracy values for the models including soil moisture data compared to the ones without soil moisture, expect for the last month in the growing season, where it give opposite results. The outcome of the model can support humanitarian organisations to increase the lead time necessary to act upon a drought trigger and reduce the impact of such event.

The livelihood of the Maasai pastoral communities in Longido District of Northern Tanzania are impacted by droughts regularly, with expectations of increasing variability in rainfall patterns the coming years due to climate change. The goal of this research is to explore if weather forecast and remote sensing data can be tailored to existing coping strategies and decision-making. Furthermore, it is assessed if this tailored information provides enough skill to effectively complement local knowledge and drought management strategies. The study generated important methodological and theoretical findings, both of which have practical implications for policy and technological development. An ethnographic and participatory approach, including four months of immersion with local families, was used to document local knowledge and strategies, and understand what specific, weather information may benefit pastoralists. The study focused on alamei periods, which refers to times of drought and scarcity in the Maasai language. It revealed that weather information around particular important ‘decision junctures’ is most relevant. On the one hand, decisions to move livestock during vulnerable times are based on current water and grass availability; on the other hand, families also consider expectations of rainfall in their decisions. The research determined that at very specific junctures throughout respective seasons, key, timely decisions must be made to maintain household resiliency. It is at these junctures that rainfall predictions become crucial. Using NDVI data and the ECMWF weather model, it was assessed if the onset of rains at such junctures can be predicted with enough skill to support livestock movement decisions. It revealed both optimism and scepticism about the role of current remote sensing and weather prediction technologies vis-à-vis variable, dryland ecologies and pastoral livelihoods.

Heavy rainfall, combined with expanding (unplanned) urban settlements in flood prone areas, expose Dar es Salaam (Tanzania) to the risks of flooding. The urbanisation is so rapid in many areas that it is not beneficial to carry out expensive surveys which are quickly out of date. The work carried out by community-mapping project Dar Ramani Huria (Swahili for "Open map") aims to make a detailed map of Dar es Salaam, to enable the hydrologic models to approach the real situation more closely. However, the surveying methods used until recently are not sufficiently accurate. However, an alternative emerges in the form of community members using a low-cost, dual-frequency global navigation satellite system (GNSS) receiver during surveys. However, before this receiver can be implemented a detailed research has to be done. In this thesis the horizontal and vertical performance of the U-blox ZED-F9P receiver in Delft (the Netherlands) and Dar es Salaam is studied. The research is divided into two parts: performance and case study. For the performance study a series of post-processed kinematic (PPK) experiments were conducted in Delft and Dar es Salaam. The experiments have been designed in order to provide a variety of location, antenna-performance, baseline length, software package and movability. In addition, two re-initialisation experiments were conducted to measure how fast the interrupted GNSS signal is regained by the receiver. The case study focused on the desirability and feasibility, mainly focussing on accuracy, of implementation in the project of Dar Ramani Huria. Structured and unstructured interviews with employees of the Humanitarian OpenStreetMap Team (HOT) Tanzania were held to find out the requirements of implementation. The positioning performance of the receiver varies between the different experiments. The conclusions regarding the positioning performance are based on the scatter plots in the horizontal plane and the positioning over time for the three separate directional components; East, North and Up. The values for the horizontal performance (RMS East, RMS North) and for the vertical performance (RMS Up) of the fix solutions insofar as they fall inside the 95% confidence ellipse are decisive. Only the relevant experiments, namely those who can map a larger area with a single reference station are taken into consideration. The horizontal positioning performance ranges from 1.13 till 16.83 However the latter, high value is from the 9 baseline Dar es Salaam experiment with a very low percentage of fixed solutions. If we disregard the experiments with low percentage of fixed solutions then the horizontal positioning performance ranges van 1.13 till 9.42. The vertical positioning performance shows less accuracy ranging from 3.56 till 14.75. If we compare this performance with the requirements for Dar Ramani Huria’s project, even the strictest of 2cm, the performance is more than adequate according to the "few cm accuracy" requirement. The experiment with the high-end antenna shows with values 2.44mm (RMS East) and 3.42mm (RMS North) the best horizontal and with the value 3.75$mm(RMS Up) the best vertical performance. Another factor influencing the performance is the location, in particular the aspect of atmospheric delay that varies between Dar es Salaam and Delft. This research thesis concludes that the implementation of the receiver in Dar Ramani Huria's project is well possible and that the performance of the receiver is adequate. This conclusion is confirmed by what is actually occurring in the field: HOT Tanzania and Dar Ramani Huria already started using the GNSS receiver and carrying out surveys with this receiver.

Direct measurement of river discharge is difficult, time consuming and costly. Therefore a rating curve is often used to estimate the river discharge. Limited measurements under extreme conditions result in extrapolation of the rating curve for high flow conditions. This induces uncertainties and errors in the stage-discharge relation. Recently there has been a gradual shift to more physically based rating curves, where the geometry of a river is included and no extrapolation is needed. This seems to be a promising shift to improve traditional river rating. However, the challenge now is to accurately determine the parameters bed roughness and hydraulic slope. The aim of this research is to develop and evaluate a method to better estimate the hydraulic parameters bed roughness and hydraulic slope. To do so a case study has been carried out in the Luangwa river catchment in Zambia.