This thesis aims to enhance rainfall nowcasting by improving motion field predictions and ensemble generation within PySTEPS using machine learning techniques. Accurate nowcasting is crucial for flood early warning, agriculture, transportation, and public safety. The steady-state assumptions and ensemble generation in PySTEPS, and many other nowcasting methods, face significant challenges in maintaining accuracy over time, especially during convective weather events characterized by rapid changes in precipitation patterns and their movement. The research focuses on three objectives: identifying the current errors and uncertainties in PySTEPS motion field estimations, constructing dynamic motion field predictions with the development of a new deep-learning model, MotioNNet, and developing motion field ensemble generation methods for MotioNNet.

We introduce a new dataset containing 10,000 motion field events, consisting of events with high rainfall and motion errors in the NL. MotioNNet, trained/validated/tested on 75%/12.5%/12.5% of the dataset, improves motion prediction accuracy by approximately 13%. It provides accurate temporal adjustments (e.g., advecting the motion), especially in structured and stable events. For ensemble generation, probabilistic techniques such as SpatialDropout and Monte Carlo dropout are employed within MotioNNet. The results indicate better ensemble predictions in structured and stable events, accurately identifying problematic areas and generating realistic members. However, it presents a small spread compared to PySTEPS. Challenges remain in accurately capturing uncertainty during highly dynamic convective events, but new paths are open for future enhancement.

The study concludes that integrating machine learning models into traditional nowcasting frameworks can substantially enhance motion field predictions and ensemble generation. Future work should focus on further refining these models or exploring more advanced architectures (e.g., ConvLSTM, diffusion models, or transformers), improving context-dependent perturbations in ensemble generation, and enhancing the references and assessments. These advancements hold potential for improving motion field nowcasting accuracy.

This thesis project is centered around the retrieval of meteorological parameters using a fast-scanning phased array radar, specifically targeting precipitation-like objects such as raindrops. The main objective is to transform radar data into variables that accurately characterize precipitation. This endeavor involves addressing various challenges, including extracting meteorological object-related variables, mapping targets from noisy radar measurements affected by phase noise, and resolving the issue of Doppler aliasing.
In this report, these challenges are tackled by introducing a range of techniques and algorithms designed to enhance radar data analysis and validate the proposed methods. The most commonly derived radar parameters for meteorological targets, known as Doppler moments, are extensively discussed, including reflectivity, mean Doppler velocity and Doppler spectrum width. An exploration of the circular calculation of Doppler spectra moments is conducted, providing valuable insights into the velocity distribution of radar targets. By analyzing higher-order moments, the characteristics and dynamics of the targets can be better understood, leading to improved target identification and tracking. Additionally, a novel circular variance-based target mapping technique is proposed to map targets from noisy radar measurements effectively. This approach proves particularly well-suited for scenarios where traditional reflectivity-based methods fall short. Apart from the circular variance-based technique, this thesis explores reflectivity-based target mapping approaches that offer enhanced methods for identifying and classifying different target parts. Two pipelines are proposed: the morphology-based pipeline and the entropy-based pipeline. Through simulation and real-world data analysis, the pros and cons of each pipeline are carefully evaluated. The second pipeline demonstrates significant benefits in classification, allowing for a more detailed and accurate representation of radar returns. It effectively distinguishes point targets, extended targets of interest, global noise, and phase noise, enabling a more comprehensive analysis of radar data and enhancing the interpretation of detected targets.
Furthermore, the critical issue of Doppler dealiasing is thoroughly investigated, addressing the problem of velocity ambiguity caused by the Nyquist limit. Three approaches are compared: gradient-based, existing advanced technique UNRAVEL, and optimizer-based. Based on the evaluation of the simulation model, the "ParticleSwarm" optimization approach is selected as the most effective in enhancing velocity measurements in severe weather conditions.
Throughout this report, experimental results are presented, findings are discussed, and recommendations and suggestions for future research are provided. The proposed techniques and algorithms have undergone rigorous evaluation and validation using both simulated data and real-world radar measurements. Overall, this research contributes
valuable insights into radar data processing for meteorological applications, offering improved accuracy and reliability for various weather-related analyses and forecasting tasks.

A personal weather station (PWS) is a set of weather measuring instruments that is owned and operated by an individual, club, association, or business. Even tho these stations might be less advanced as the professional stations, the amount of amateur stations that are active is much higher. With these PWS networks in full effect and still growing, the potential to use PWS data for weather forecasting or research purposes has gained popularity in the scientific field for its potential to ”fill data gaps”. However, since the stations are managed by “amateurs”, the “usefulness” and “value” of these stations are under questions. In this thesis, we tried to solve this by proposing a new way of assessing the “Value” of an amateur weather station in a citizen network. We utilized bots to extract all information about stations in the Netherlands and for the first time did a national analysis of PWS networks. Significant progress has been made in developing a framework to assess the value of the Dutch PWS network. The research process involved overcoming numerous obstacles, which have provided valuable insights and lessons for the future development of a more refined PWS network assessment system.

Machine learning models offer promising potential in precipitation nowcasting. However, a common issue faced by many of these models is the tendency to produce blurry precipitation nowcasts, which are unrealistic. Previous research on the deep learning model - TrajGRU (Shi et al., 2017) indicated that data imbalance in radar images and the double-penalty effect of pixel-wise loss functions are underlying causes for this blurriness.
In this thesis, we continue to explore various approaches to improve the predictive performance of TrajGRU. Our research has first investigated spatially and temporally enhanced loss functions to address the two remaining issues: data imbalance and double penalty. The second part of our research focuses on manifold optimizations within the model network, such as incorporating additional model inputs or increasing batch size, to understand the model’s limitations.
Our results reveal that enhanced loss functions did not lead to predictive improvements and even resulted in undesired checkerboard patterns. Changes to the model network make a difference in the image sharpness and predictive rain evolution. Our visual analysis indicates that a larger batch size generates sharper rain field edges; predictions by using multiple parameter groups exhibit more rain dynamics. The incorporation with other transformed datasets introduces finer structures within rain fields. Although the blurriness has not been completely resolved, our study recommended future work can continue exploring the optimization in the TrajGRU network.

BaCla is a new stochastic parametric precipitation model to estimate rainfall associated with Tropical cyclones (TCs) in a computationally efficient way. It is validated for a number of calibration cases along with the current benchmark IPET deterministic method. The results of these models are compared with the observed StageIV-based rainfall. Two predictive parameters, the pressure deficit [hPa] (Δ𝑃) and maximum sustained wind speed [m/s] (vmax) are tested for the BaCla model. Frank copula is used by the BaCla model to predict the peak amount of rainfall. It employs the adapted Holland wind profile to create a 1D rain profile. Asymmetry may be added to make a 2D rain profile. The calibration study challenges the assumption that the radius of maximum wind speed (Rvmax) is equal to the radius of maximum precipitation (Rpmax) in the BaCla model. Additionally, it is observed that the radial fit overestimates rainfall in scenarios where the peak amount of rainfall is less than 2.8mm/hr. These observations lead to modifications in the relationship between Rvmax and Rpmax, the radial fit threshold, and the fitting coefficients of the adapted Holland wind profile for rainfall distribution in the improved BaClHa model. The new BaClHa model is verified with new and calibrated sets of TCs. The BaClHa model shows potential in achieving good accuracy along with global applicability at a limited computational expense.

Accurate short term rain predictions are important for flood early warning systems, emergency services, energy management and other services that that make weather dependent decisions. Recently introduced machine learning models suffer from blurry and unrealistic predictions at longer lead times, causing poor performance on the rarer heavy rainfall events. The objective of this research was to explore how the loss function in a recurrent, convolutional neural network (TrajGRU, X. Shi et al. (2017)) can be modified, to get sharper and more realistic predictions, without worsening its performance. Six perceptive loss functions, which should better represent how an image is perceived, are thoroughly analysed to understand the reasons behind their functioning in the context of precipitation nowcasting. It was shown that these perceptive losses can lead to an improvement in sharpness for the first lead time, but that the blurriness for longer lead times arises due to the imbalance in the dataset and the uncertainty of the model with respect to the location of the rain. This thesis gets concluded with recommendations on how to deal with these two problems.

The Dual-Frequency Precipitation Radar on board the Global Precipitation Measurement (GPM) mission core satellite has been providing precipitation products across the globe for over 6 years, thereby even supplying precipitation estimates for areas on Earth where surface-based precipitation measurements are not possible, like in remote regions, as well as seas and oceans. In this study, the GPM DPR Ku-band Level 2A instantaneous observations are compared with continuous measurements from a surface network in the Netherlands for high-intensity precipitation events during summer months. The aim of the research is to gain more insights in the temporal and spatial correspondence of the DPR’s measurements with precipitation occurring on the surface. The surface network consists of 2 C-band weather radars and an automatic gauge network of 33 gauges (approximately 50 km apart). Radar reflectivity factor is one of the prime parameters used for precipitation retrievals from radars and is thus the parameter used for anal- ysis of both the DPR and the weather radar network. Precipitation intensity is the parameter retrieved from the automatic gauge network. Space-borne and surface measurements have vary- ing characteristics, for instance in spatial and temporal resolution, and thus many challenges arise to perform an accurate and qualitative comparison. In this research, data is funneled by selecting high-intensity gauge data with a threshold of 15 mm/h within 40 minutes from GPM scan time within a 2.5 km range from the nearest DPR footprint. Ultimately, 26 high-intensity gauge measurements spread over 18 dates were used in the comparison analyses, selected out of data ranging from May - October for the years 2018 - 2021. The analyses showed that DPR and WR data correspond best in time (average correlation of 0.785) when DPR scantime is within 3 minutes of WR scantime. Also, results showed that instantaneous DPR data can be related to high-intensity gauge observations within a 40 minute time range by use of wind direction and wind speed. Furthermore, it was observed that precipitation performance of DPR is found satisfactory for intensity ≥ 0.5 mm/h (POD between 0.7 - 0.9 based on distance WS-IFOV), but unsatisfactory for intensities ≥ 15 mm/h (POD of 0.18 - 0.25). The results imply that DPR measurements don’t have an obvious temporal and spatial lag with respect to the observations made from the surface and thus form a prospect for global usage, if more follow-up research is conducted.

Torrential rain from tropical cyclones can have a devastating impact, causing loss of life and billions in damages. To better understand the risk faced by coastal communities, it is important to estimate how often a tropical cyclone could occur and how much rainfall it will produce. One way to do this is by analyzing past storms and building parametric models of rainfall rates during tropical cyclone events. While many parametric precipitation models –such as the Bader model– exist, their accuracy remains limited and many challenges still need to be overcome. The most important challenges are output overestimation and a poor representation of rainfall over land. Therefore, this thesis aims to reduce
these biases by answering the following research question:

"How can the bias in the radial rainfall distributions of a tropical cyclone in Bader’s
parametrized model be reduced and be used for reliable rainfall estimates both above land and the ocean?"

To answer this question, several new data sources were introduced from the TRMM/GPM satellites and Stage IV to improve the Bader model. While this original model only predicted precipitation based on maximum wind speed (vmax), the updated model also considers pressure deficit ΔP. The results suggest that ΔP can be a useful parameter to reduce bias and improve accuracy. However, it also
leads to larger uncertainty ranges. Next, four precipitation profiles were proposed. A profile where precipitation is constant for low maximum precipitation values based on the predicted total rainfall (area under the graph) was selected for further exploration.

The new models are explored during a case study of Hurricane Florence. Both the ΔP and vmax based models produced satisfactory results, compared to the benchmark IPET model. Moreover, an alternative fit above land has been proposed, where the highest precipitation is simulated at the eye. The proposed land fit improved the median of the predictions based on both vmax and ΔP. The ΔP
based model performed the best in the case study, however, no definitive conclusion could be reached upon which model is most suitable overall as more case studies would be required.

Finally the updated model has been compared to the original Bader model. The new data ensured better representation over land, the overestimation of precipitation was reduced, and the model was applied with more confidence outside of the training data set. Consequently, results showed an improvement
on its prediction capabilities. As a concluding remark, this research project highlights the importance of having insightful data to enhance the decision-making and risk management of natural hazards: a model that accurately quantifies uncertainty and the risks associated with a TC, representing a valuable tool for better understanding flood risk. Nonetheless, there are still several ways to further improve the modeled profiles (e.g., by including more data, introducing asymmetry or adding temporal autocorrelation).

Accurate short-term forecasts, also known as nowcasts, of heavy precipitation are desirable for creating early warning systems for extreme weather and its consequences, e.g. urban flooding. In this research, we explore the use of machine learning for short-term prediction of heavy summer rainfall showers in the Netherlands. We explore the use of a recurrent, convolutional neural network (TrajGRU, Shi et al., 2017) with lead times of up to 100 minutes. We aim to train the model to predict the formation and dissipation of dynamic, heavy, localized rain events, a task for which traditional Lagrangian extrapolation based nowcasting methods still come up short. The network is trained, validated and tested on a 13-year archive of radar images with 5-min temporal and 1-km spatial resolution from the precipitation radars of the Royal Netherlands Meteorological Institute (KNMI). We report on different ways to optimize predictive performance for heavy precipitation events through two experiments. In the first experiment different training dataset compositions are explored. The large dataset available provides many possible configurations for training. To focus on heavy rainfall intensities, we use different subsets of the radar dataset through using different conditions for event selection and varying the ratio of light and heavy precipitation events present in the training data set. In the second experiment we change the loss function used to train the model. To assess the performance of the model, we compare our method to a current state-of-the-art deterministic Lagrangian extrapolation-based nowcasting system from the pySTEPS library, S-PROG (Seed et al, 2003). The results of the experiments are used to discuss the pros and cons of machine-learning based methods for precipitation nowcasting and possible ways to further increase performance. Model behaviour was found to be significantly influenced by the formulation of the loss function. It was concluded that there is always a trade-off between performance at low rainfall intensities and performance at high rainfall intensities: (1) If a model makes smaller errors at low rainfall intensities this results in a low total error, but also in the failure to detect high rainfall intensities. (2) If model performance is improved at detecting high rainfall intensities, this results in a decreased performance at low rainfall rates and increases the total error.

Large parts of the world rely on rainfed agriculture for their food security. In Africa, 90% of the agricultural yields rely only on precipitation for irrigation purposes and approximately 80% of the population’s livelihood is highly dependent on its food production. Parts of Ghana are prone to droughts and flood events due to increasing variability of precipitation phenomena. Crop growth is sensitive to the wet- and dry spell phenomena during the rainy season. To support rural communities and small farmer in their efforts to adapt to climate change and natural variability, it is crucial to have good predictions of rainfall and related dry/wet spell indices.
This research constitutes an attempt to assess the dry spell patterns in the northern region of Ghana, near Burkina Faso. We aim to develop a model which by exploiting satellite products overcomes the poor temporal and spatial coverage of existing ground precipitation measurements. For this purpose 14 meteorological stations featuring different temporal coverage are used together with satellite-based precipitation products.
Conventional machine-learning and deep-learning algorithms were compared in an attempt to establish a link between satellite products and field rainfall data for dry spell assessment. The deep-learning architecture used should be able to efficiently manipulate spatial data. Hence, Convolutional Neural Networks were used in order to detect spatial patterns in the satellite data.
Using these models we will attempt to exploit the long temporal coverage of the satellite products in order to overcome the poor temporal and spatial coverage of existing ground precipitation measurements. Doing that, our final objective is to enhance our knowledge about the dry spell characteristics and, thus, provide more reliable climatic information to the farmers in the area of Northern Ghana.

Radar rainfall nowcasting stands for the prediction of rainfall amounts and intensities over the next 6 hours by means of statistical extrapolation of radar measurements. It is the principal ingredient for modern flood forecasting and early warning systems. Radar forecasts are generated by identifying and tracking rainfall cells in radar images and extrapolating them along the main direction of motion based on the assumption of Lagrangian persistence. The latter means that the rainfall cells do not physically evolve over time. Therefore the key to improving predictions is to anticipate the growth and decay of rainfall cells during the next few hours. Studies have shown that estimating growth and decay based on past radar images is nearly impossible. Therefore, in this thesis, a new approach for predicting growth and decay based on physical guidance from a numerical weather prediction model is investigated. The idea is that numerical weather prediction models are better at anticipating changes in the atmosphere than radar and thus, should also contain information about the growth and decay of rain cells. To test this hypothesis, two simple machine learning models have been trained to learn the relationship between the observed growth and decay in radar images and the output parameters from the Dutch mesoscale numerical weather prediction model HARMONIE. The models were trained on data in the summer of 2019 around the city of Rotterdam for two area sizes, $20\times 20$ km$^2$ and $60\times 60$ km$^2$. The first model is static and assumes that all predictions from HARMONIE are correct. Unfortunately, this turns out to be too optimistic as HARMONIE often places the rain cells at the wrong locations with the wrong intensities and trends. As a result, no meaningful relationship between HARMONIE outputs and growth/decay in radar could be learned. The dynamic model overcomes this issue with the help of an additional classifier. The classifier predicts whether the information from HARMONIE can be trusted or not. Then, a regression model predicts the magnitude of growth and decay (only for the trusted cases). Out of 308 analysed cases, the classifier labelled 237 HARMONIE predictions as being untrustworthy, thereby removing a lot of bad cases. However, 98 of these were false negatives, meaning that they could have been used to predict growth and decay. On the other hand, only 71 cases were labelled as containing useful information, from which 27 cases were false positives. Despite these errors, the dynamic model was able to improve the root mean square error on the predicted growth/decay by $27$\% for these 71 cases. Also, the correlation between the predicted and actual growth/decay increased from near zero to $0.488$. This is encouraging and clearly highlights the potential of this approach. Still, some important challenges remain. In particular, results show that the dynamic model often underestimates the magnitude of growth and decay. Also, when the entire validation data set is considered, including all unusable cases, the improvement with respect to the static model is only $6$\%. This is mainly caused by the poor performance of HARMONIE and the large number of unusable cases. However, performance could be improved further by building a better classifier capable of identifying all good cases with a low number of false negatives.

The increasing penetration of weather-dependent energy sources brings additional challenges to the operation of the power system. Wind power forecasting is a valuable resource for these power operators: a tool that aids the decision-making process and facilitates risk management. On the other hand, the progress of machine learning and their success in different fields attracted research into wind power probabilistic forecasting. The objective of the research was to enhance wind power forecasting through data-driven or machine learning models, proposing a new framework as an alternative to the existing ones, validated in different climate regions using grid-like topology data (weather images). The project focused on the European Energy Markets 2020 Conference (EEM20) Forecasting Competition. The setting emulates the day-ahead electricity market and the location consisted of all four electricity price regions in Sweden, where three main sets of data were provided. Three main challenges are present in this competition, namely the large volume of data (the curse of dimensionality), the lack of information regarding wind turbine availability, and the growing evolution of wind installed capacity between data sets. The methodology to create a framework consisted of evaluating different machine learning approaches, feature engineering, and tuning the final model for every region. The final framework is called the Smooth Pinball Hybrid Neural Network (SPinHy-NN). The results showcased a suitable CNN architecture, providing more accurate forecasts compared to MLP and k-means clustering. The quantile cross-over problem has been evaluated through the crossing loss and the number of crossings metrics. The analysis shows that a margin parameter can correct this undesired behavior. The conclusions validate the SPinHy-NN as a generalized framework for different climates through the final score achieved post-competition, managing to capture the spatial patterns in most cases. Moreover, the framework can be adapted to reach a desirable trade-off between accuracy, sharpness, and consistency. Recommendations aim to extend this research by employing satellite imaging, further feature engineering, novel neural networks, and exploring spatiotemporal models to capture atmospheric dynamics.

Eutrophication processes in coastal waters are becoming more prominent as a result of high nutrient discharges from intensive agriculture and increased urban waste. These processes can be devastating for local ecosystems and lead to dissolved oxygen depletion, which applies considerable stress on aquatic organisms. For ecosystems to preserve their status, stop and reverse the negative effects of eutrophication, regular estimation of corresponding indicators has to take place. In this direction, mostly process-driven models have been used, but the presented project argues that freely available remote sensing data can also provide useful insights for the oxygen saturation of the water. The proposed methodology uses Sea Surface Temperature and Chlorophyll-a estimations from AQUA and ENVISAT satellite sensors for the period 2003-2011 to predict the dissolved oxygen content in the Dutch coastal waters. It does so by implementing various Machine Learning models, namely Random Forest, Artificial Neural Network and Gradient Boosting Regressors, with the latter demonstrating the best results. After extensive data pre-processing, the results show that dissolved oxygen can be predicted with an average Root-Mean-Squared error of 0.8 g/m3. Important steps towards a lower error include the use of gap-filled variables and their decomposition into their temporal components as inputs for the model. Furthermore, the effect of the Sea Surface Temperature on the dissolved oxygen is documented through its contribution in the estimation of the latter’s seasonal variability, while the estimation of the maximum dissolved oxygen values is attributed to Chlorophyll-a. Further feature engineering and model development can possibly improve the estimation of the minimum dissolved oxygen values in the coast and the overall prediction in more complex intertidal areas, like the Wadden Sea.

Flooding in cities, known as Urban Pluvial Flooding (UPF), causes disruption of society, damage to cities and inconvenience for people. Cities are expected to become more vulnerable to UPF due to more frequent and more intense extreme rainfall events with high spatial variability as result of climate change. Additionally increasing urbanisation results in more impervious surfaces, inducing shorter response times of the drainage systems. A high spatial and temporal resolution of the urban rainfall network is required to forecast UPF. Professional rain gauge networks do not provide this necessary resolution. A strategy to increase this density is to use rainfall data obtained from Citizen Weather Stations (CWSs). The downside of CWS networks is the low quality of measurements compared to professional measurements. This is caused by the use of lower quality sensor, web-platform processes , set-up of the weather station, stability of data transfer and cleaning of the CWS. Undetected Rainfall (UR) is one of the errors found in rain gauge data of a CWS network. An UR error is an incorrect zero rainfall value in the rainfall data, meaning that rainfall did occur at the gauge but it was not detected. A quality-control system that creates a more reliable CWS network is required to facilitate UPF forecasting. The aim of this research is to take a first step towards a quality-control system by developing and testing different automatic quality-control filters that flag for UR errors in rainfall data of a CWS network.

For this research 753 CWSs in the Rotterdam-The Hague region are used. They were obtained from Netatmo weathermap and give a density of 1 CWSs per 5.93 km2. This is not a uniform density because more CWSs are located in densely populated areas. The availability of the CWSs grew from 369 to 668, measuring at least once per day between October 2015 and October 2017. Two filters are developed to identify UR errors in CWS rainfall data. The first filter is the CWS filter which uses neighbouring CWSs within a radius of 8 kilometre to find the UR errors. The second filter is the radar filter, which identifies UR errors using the overlying radar pixel. The performances of these filters are tested by artificially placing zeros at 10% of the rainfall sequences of 10-min rainfall data from four automatic KNMI gauge datasets within the study area. Both filters are applied on these datasets and evaluated using confusion matrices. Finally, the two filters are applied on the 753 CWS rainfall datasets for a period between October 2016 and October 2017.

While rainfall is the key input to most hydrological models, its precise characteristics are often uncertain. Runoff generation does not only depend on the measured rainfall resolution but also on the level of detail of land-use and therefore of the runoff generation. This study aims at identifying the influence of rainfall radar resolution and land-use data on the urban water balance in Rotterdam. Results show that the water balance in this study does not close properly, as more volume enters than leaves the system. This is most probably because infiltration is neglected and the reliability of the pumping data is uncertain. Furthermore, the Combined Sewer Overflow (CSO) volumes are overestimated which might be caused by the high uncertainty of the weir parameters and of the water levels at the non-monitored CSO weirs. This error multiplies with an increase in sewer district size, a higher amount of unmonitored CSOs and lower weir levels. When comparing the different resolutions, the water balance degrades remarkably with coarser land-use data detail and improves slightly with higher rainfall radar resolution, until reaching a certain threshold where the error is minimized. After this threshold the water balance closes less again. Possibly, the reduction in noise and in sensitivity to shifts in timing and location of the radar data with coarsening rainfall radar resolutions is responsible for these unexpected results. Furthermore, this study suggests that there might be a relationship between the changes in land-use resolution and the changes in rainfall radar resolution.

The Inter-Amount Times (IAT) adaptive sampling technique has been used for the analysis of hydrological response in urban areas. The analysis of hydrological response often requires the correct use and interpretation of temporal parameters such as lag time adapted to complexities and requirements of urban environments.
In this thesis, the development of a hydrological response methodology that avails itself of the statistical framework of IAT for the hydrological response analysis is suggested by proposing a new lag time definition based on IAT. This new definition was evaluated through the analysis of discharge and precipitation time series of fifteen years. The IAT lag time results were compared against classical methods used to define and compute lag time such as centroid-to-centroid (C-to-C) and peak-to-peak (P-to-P) methods. Finally, a validation procedure is carried out to assess the validity of the proposed IAT lag time definition.
Results from the computation of hydrological response values using each of the three methods showed some relevant similarities but also exhibited varying behaviours and limitations based on the proposed methodology for the definition of the hydrological events. Statistical distributions of the IAT lag time method and C-to-C method had similar distributions, in contrast to what was observed for the statistical distribution of P-to-P values. Results from the IAT lag time method suggested a more stable and flexible methodology compared to the classical methods.
The IAT lag time method can potentially be used to have a more dynamic and tailored interpretation of the hydrological response behaviour in urban areas. Experimental work is required to test the method in other urban areas to confirm the validity of the results of this approach used to study the hydrological response of urban catchment.