“IT IS TWENTY MINUTES PAST FOUR” or “IT IS QUARTER TO SEVEN” - just like you’d say it to a friend.

The clock updates every minute, automatically highlighting the words that form the correct time phrase. Notice how at different times, different word combinations light up to create readable sentences.

It’s a fun, more human way to experience time passing. Enjoy watching the words change as the minutes tick by!

Inspiration

Inspired by the Word Clock sold by Walmart - The Word Clock - Shows The Time In A Sentence

See it at Observable - https://observablehq.com/@bennyistanto/word-clock

This is my second-owned Thinkpad after the T480, and same as before I also did some upgrades on my T14. Here’s the list of the components:

1. T14 Gen 2: 14″, 3840×2160, IPS, 500 nits, 100% Adobe RGB, Anti-glare from https://www.myfixguide.com/store/screen-for-thinkpad-t14/
2. 40pin UHD cable from https://www.myfixguide.com/store/lcd-cable-for-t14-gen2/
3. 4TB NVMe SSD from https://www.crucial.com/ssd/p3-plus/CT4000P3PSSD8
4. 32GB RAM DDR4 SODIMM 3200MHz from https://www.corsair.com/us/en/p/memory/cmsx32gx4m1a3200c22/corsair-high-performance-vengeance-memory-kit-cmsx32gx4m1a3200c22
5. WWAN card and the antenna from https://thinkparts.com/products/-4970

I did all the installation myself by following the awesome guideline and use Pro Tech Toolkit https://www.ifixit.com/Store/Tools/Pro-Tech-Toolkit/IF145-307 from IFIXIT.

See some of the picture below:

[caption id=“” align=“alignnone” width=“3333”] Remove the old screen and install the WWAN antenna [/caption] [caption id=“” align=“alignnone” width=“3333”] Install the new screen [/caption] [caption id=“” align=“alignnone” width=“3333”] Put the memory in the free slot [/caption] [caption id=“” align=“alignnone” width=“3333”] Replace the old SSD with new one, and put the WWAN card in place and antenna cable too [/caption] [caption id=“” align=“alignnone” width=“4032”] This is the machine with upgrades [/caption] [caption id=“” align=“alignnone” width=“3024”] Look, the Cellular connection is available now [/caption] [caption id=“” align=“alignnone” width=“3024”] The new 4K screen is on and I just need to install the bezel 😎 [/caption]

To download the full repository, you can ccess it via this link: https://github.com/bennyistanto/drought-propagation

time_scale_combinations = [
    "spi03_ssi03", "spi06_ssi03", "spi09_ssi03", "spi12_ssi03",
    "spi03_ssi06", "spi06_ssi06", "spi09_ssi06", "spi12_ssi06",
    "spi03_ssi09", "spi06_ssi09", "spi09_ssi09", "spi12_ssi09",
    "spi03_ssi12", "spi06_ssi12", "spi09_ssi12", "spi12_ssi12"
]

I have old version of Thuraya SatSleeve for iPhone, now categorized as a legacy product on Thuraya website.

Since year ago the SatSleeve device experiencing intermittent connection, between bluetooth on iPhone SE and SatSleeve, also the GPS connection. Luckily, Thuraya providing a firmware upgrade with release v3.0.1 on their website https://www.thuraya.com/en/support/upgrades/legacy/thuraya-satsleeve-for-iphone

Before upgrading a SatSleeve, please check which firmware is installed (SatSleeve > Settings > Device Info > Firmware version). Perform the upgrade only if Thuraya releases a firmware version newer than your existing one (mine is v2.94)

To upgrade the firmware, follow steps on the website:

Step 1

• Download the below SatSleeve Upgrader program.

  SatSleeve upgrader

• Unzip and Run the setup file - the Upgrader program including the USB driver will be installed.

Step 2

• Download the latest Thuraya SatSleeve firmware release to your hard disk.

  SatSleeve iPhone firmware release v3.0.1

  (works only on the SatSleeve for iPhone Data model) - Unzip it.

  Release notes of v3.0.1: Fixed GPS rollover issues

Step 3

• Connect your SatSleeve with the PC/laptop via USB data cable.

• You can now start the SatSleeve Upgrader program (please make sure you run this software as Administrator.

  • The requirement is using PC with Windows 8/8.1, Windows 7 or Windows Vista, and I am using Windows 11 is working fine
  • Right click Thuraya SatSlevee Upgrader > More > Run as administrator

• and locate the firmware on your hard disk. The Upgrader program will help you through the upgrade process.

Now, my SatSleeve is working fine on iPhone SE

Recently, I embarked on a journey to calculate humidity data from a myriad of sources. Throughout this process, I experimented with various methods, ranging from the saturation water vapour pressure using Teten’s formula (with parameters according to Buck) to saturation over ice from Alduchov and Eskridge, and finally to Clausius-Clapeyron.

For those in need of hourly humidity data spanning from 1 Jan 1950 to the present, there’s good news! You can seamlessly extract this information from ERA5-Land Hourly data via Google Earth Engine (GEE). The Specific and Relative Humidity is meticulously calculated based on three core parameters: T2m (Temperature at 2 meters), Dew Point, and Surface Pressure.

Interested in exploring further? Check out my GEE script: https://code.earthengine.google.com/9b23f929939122fb1fdc8418d17c43f5

By the way, for those diving deep into the technicalities, the GEE script I’ve shared leans on a simpler approach, kinda like a nod to the good ol’ Magnus formula. So, it’s pretty straightforward and user-friendly

I hope this proves beneficial to researchers, data scientists, and enthusiasts in the realm of climatology. If you have any suggestions, feedback, or improvements, please don’t hesitate to reach out.

Reference

Alduchov, O. A., & Eskridge, R. E. (1996). Improved Magnus form approximation of saturation vapor pressure. Journal of Applied Meteorology, 35(4), 601-609.

Exciting news! I’m now a certified GIS Professional (GISP)

Big thanks to GIS Certification Institute (GISCI) for this recognition. To learn more about the GISP certification process, visit https://www.gisci.org

Shoutout to my mentors at the WBG, Keith Garrett and Ben Stewart, for their guidance in the last 2 years.

Looking forward to doing more meaningful work in climate analytics and geospatial technology for greater impact! 🗺️

Update: happy to received the certificate at the end of 2023

Few months ago, I wrote a post about how to calculate The modified Radar Vegetation Index (mRVI) using Sentinel-1 satellite. It was try to extract the mRVI every Dekad with study case is Ukraine.

For areas in Europe, getting the S1 data every dekad is doable, but it’s bit tricky for area outside Europe. Currently, for my work, I would like to extract the mRVI for Mpumalanga province in South Africa. The location is near -25.5 S, then according to below picture the revisit time is every 12 days.

[caption id=“” align=“alignnone” width=“3579”][](https://sentinel.esa.int/documents/247904/4748961/Sentinel-1-Repeat-Coverage-Frequency-Geometry-2021.jpg) Source: https://sentinel.esa.int/documents/247904/4748961/Sentinel-1-Repeat-Coverage-Frequency-Geometry-2021.jpg [/caption]

Getting monthly mosaic mRVI seems possible for case in South Africa, as if I keep the dekad, some of them will return an empty collection.

So, I need to modify the existing code to get the monthly list, calculate monthly mosaics, calculate monthly mean and the ratio anomaly

Full GEE code is here: https://code.earthengine.google.com/aea00cb8f3f1ccc921d5f6698b5c0c5a

Then go to Climate Data > Daily Data, choose the Station Type, Parameter, Province, Regency, Station Name and the Date Period and click Process button. You will get the data in *.xlsx format.

You can get one of the data example from this link: https://docs.google.com/spreadsheets/d/1xbBWeHhiMNs8IehHbsrMV9yeZlcu8GqR/edit?usp=sharing&ouid=104182606454912191559&rtpof=true&sd=true

In above example, I tried to get daily precipitation data for all station from 1 Jun 2000 - 31 Dec 2021. I would like to use it to correct the value and distribution daily IMERG data using bias correction method that currently I have develop.

Unfortunately, too many missing value.

I should find alternative daily timeseries precipitation data, probably gridded data will suit my objectives.

1 Introduction

In the sphere of meteorology, the significance of statistical models in comprehending and forecasting diverse weather patterns is incontestable. Within this context, the Fourier regression model has emerged as a formidable asset, specifically in generating daily time series from monthly temperature data (Wilks, 1998). The model lays a robust foundation for simulating temperature patterns, yielding crucial insights that are indispensable for weather prediction, climate change studies, and managing water resources.

The Fourier regression model has been proven to be a highly effective tool for generating daily time series from monthly temperature data, enhancing our understanding and prediction capabilities in weather forecasting, climate change studies, and water resource management. This model’s unique ability to incorporate historical context allows it to capture intricate dependencies and transitions in temperature data, which are crucial in understanding temperature patterns.

By applying Fourier series, it is possible to reduce the number of parameters involved in the process, thereby simplifying complex calculations and making the model more efficient. Moreover, the Fourier regression model can seamlessly replace missing values and handle anomalies, which are often challenges in data analysis. This enables more accurate simulations and predictions, making it a vital tool in fields such as agriculture and urban planning.

The Fourier regression model’s success in generating daily time series from monthly temperature data not only contributes to our understanding of weather patterns but also provides practical solutions for real-world challenges, making it a powerful instrument in various domains.

The original GEE script was meant to be used only for individual updates, as time progresses and the need for vegetation monitoring continually increases, I believe it’s necessary to obtain this RVI time-series data, which can be matched with monthly rainfall time-series data for monitoring food crop phenology.

For this reason, I’ve added a function to mosaic every ten days and batch downloading if the list of data is quite extensive.

All credit goes to the awesome work of Jose Manuel! Hats off to him!

[caption id=“” align=“alignnone” width=“1277”] RVI in Crimean Peninsula [/caption]

Above picture is Vegetation Indices based on Sentinel-1 (generated using the GEE script), below picture is Vegetation Indices for the same period based on Sentinel-2 (generated using Climate Engine https://climengine.page.link/sZnR)

NDVI in Crimean Peninsula

Full GEE code is here: https://code.earthengine.google.com/62f799954525c997629cefdd435c500e

1 Introduction

In the realm of meteorological studies, the use of statistical models is pivotal for understanding and predicting various weather phenomena. Among these models, the second-order Markov chain model has emerged as a powerful tool, particularly in generating time series of rainfall occurrence (Wilks, 1998). This model provides a robust framework for simulating rainfall patterns, offering valuable insights that are crucial for weather forecasting, water resource management, and climate change studies.

The second-order Markov chain model distinguishes itself from its first-order counterpart through its ability to consider not just the state of the system at the previous time step, but also the state at the time step before that. This additional layer of historical context allows the model to capture more complex dependencies and transitions in the rainfall data (Bellone et al., 2000). This enhanced capability significantly improves the accuracy of the generated time series, making it a powerful tool in the study of rainfall patterns.

Rainfall, as a natural phenomenon, exhibits a high degree of variability and randomness. The second-order Markov chain model, with its ability to incorporate historical context, is well-equipped to handle this variability (Hughes et al., 1999). By considering the state of the system at two previous time steps, the model can capture the inherent randomness in rainfall occurrence, thereby generating a time series that closely mirrors real-world rainfall patterns.

The application of the second-order Markov chain model to rainfall data is not just a theoretical exercise. The generated time series of rainfall occurrence can have practical applications in various fields. For instance, in the field of agriculture, understanding rainfall patterns can help farmers plan their planting and harvesting schedules (Rosenzweig et al., 2000). In urban planning, accurate rainfall predictions can inform the design of drainage systems to prevent flooding (Ashley et al., 2005).

The report examining the two-way relationship between cities and climate change, including valuable insights to help cities boost resilience and thrive, both now and in the future.

The team are utilizing temperature and precipitation based index: SPEI, SPI, CDD, CWD, number of annual hotdays and annual mean temperature, distance and magnitude of tropical cyclone to the city center, to support the analysis which organized into four inter-related workstreams:

• Who is affected?
• Stressor that make urban development less green.
• Stressors that make urban development less resilient.
• Stressors that make urban development less inclusive.

If you are interested to read the story and report, feel free to visit below:

1. Story: https://www.worldbank.org/en/publication/thriving
2. Recorded of Live event:https://live.worldbank.org/events/thriving-making-cities-climate-ready
3. Publication: https://openknowledge.worldbank.org/entities/publication/7d290fa9-da18-53b6-a1a4-be6f7421d937

In the implementation phase of this analysis, we utilized Python and the Scikit-Fuzzy library to develop a fuzzy logic-based flood risk assessment model. This model took into account four essential factors affecting flood risks: precipitation intensity, soil moisture, land cover, and slope. By defining the fuzzy sets and rules for these variables, the model was able to estimate the flood risk for various combinations of input values. The ultimate goal of this implementation was to identify the conditions under which low flood risks could be achieved, even in situations where precipitation intensity was at its maximum.

1.1 How-to?

In the first stage of the analysis, we defined the variables that influence flood risk. These variables include precipitation intensity, soil moisture, land cover, and slope. Each of these variables was represented as a fuzzy variable using the Scikit-Fuzzy library’s Antecedent class. Additionally, we defined the output variable flood_risk using the Consequent class. This stage set the foundation for the fuzzy logic-based flood risk assessment model by establishing the key variables that the model would use to estimate flood risk.

In the second and third stages, we focused on defining the fuzzy sets and their respective membership functions for each of the variables defined in the first stage. We used the automf() function to automatically generate triangular membership functions for precipitation intensity, soil moisture, and flood risk, each with three levels: low, medium, and high. For the land cover and slope variables, we manually defined triangular membership functions, specifying the appropriate ranges for each fuzzy set (urban, vegetation, and bare_soil for land cover, and flat, moderate, and steep for slope). These stages were critical for establishing the relationships between the input variables and the output flood risk, which would later be used to evaluate different combinations of input values in the fuzzy inference process.

In the fourth stage, we defined the rules that describe the relationships between the input variables (precipitation intensity, soil moisture, land cover, and slope) and the output variable (flood risk). We first created a list of classifications for each input variable and the output variable. Using the multiplication principle, we calculated the total number of possible combinations of these classifications, resulting in 81 unique rules.

For each combination of input classifications, we determined the appropriate flood risk level based on a set of predefined conditions. These conditions were based on expert knowledge and domain understanding, considering factors such as high precipitation and soil moisture, bare soil land cover, and steep slopes. After determining the flood risk level for each combination, we created a fuzzy rule using the Scikit-Fuzzy library’s Rule class, linking the input conditions with the corresponding flood risk level. These rules formed the basis of the fuzzy inference system that was used to evaluate different scenarios and estimate the corresponding flood risks.

In the fifth stage, we created the control system and simulation by combining the defined rules from the previous stage. The Scikit-Fuzzy library’s ControlSystem and ControlSystemSimulation classes were used for this purpose. The ControlSystem class takes the set of rules as input and initializes the fuzzy inference system, while the ControlSystemSimulation class initializes a simulation environment that can be used to compute the output based on the input values.

In the sixth stage, we provided example input values for each input variable (precipitation, soil moisture, land cover, and slope) to test the fuzzy inference system. The input values were assigned to their corresponding input variables in the simulation, and the compute method of the ControlSystemSimulation object was called to perform the fuzzy inference process and obtain the output flood risk level.

In the final stage, we output the computed flood risk level and visualize the result using the Scikit-Fuzzy library’s built-in plotting capabilities. The flood risk level was displayed as a numerical value, while the visualization provided a graphical representation of the membership functions and the defuzzified output. This allowed us to assess the performance of the fuzzy inference system and analyze the relationships between the input variables and the flood risk.

This will returned:

Flood Risk Value: 83.33333333333336
Flood Risk Category: high

And a plot below

The initial implementation of the fuzzy inference system for flood risk assessment has been completed successfully. By providing example input values for precipitation (100), soil moisture (50), land cover (25), and slope (30) in Stage 6, we have demonstrated the functionality of the fuzzy system. The system processes these inputs through the defined membership functions, rules, and defuzzification methods to produce an output flood risk value and the corresponding flood risk category.

Upon evaluating the system with the given input values, a flood risk value is generated, and the flood_risk.view(sim=flood_risk_sim) function provides a visual representation of the output. The plot displays the aggregated output membership functions and indicates the defuzzified crisp value. In this case, the plot reflects the flood risk level based on the provided inputs, and the computed flood risk category helps to understand the risk associated with the given conditions. With this initial implementation, we have set the foundation for further analyses and can adapt or extend the fuzzy system as needed to address specific flood risk assessment scenarios.

1.2 Plot the membership function of the input variables

The provided code visualizes the membership functions for each of the input variables (Precipitation Intensity, Soil Moisture, Land Cover, and Slope) and the output variable (Flood Risk Level) in the fuzzy inference system. Here’s a summary of what each part of the code does:

• precipitation.view(sim=flood_risk_sim): Plots the membership functions for the Precipitation Intensity variable, displaying how the input values are categorized into low, medium, and high precipitation levels.
• soil_moisture.view(sim=flood_risk_sim): Plots the membership functions for the Soil Moisture variable, showing how the input values are categorized into low, medium, and high soil moisture levels.
• land_cover.view(sim=flood_risk_sim): Plots the membership functions for the Land Cover variable, illustrating how the input values are categorized into low, medium, and high land cover levels.
• slope.view(sim=flood_risk_sim): Plots the membership functions for the Slope variable, demonstrating how the input values are categorized into low, medium, and high slope levels.
• flood_risk.view(): Plots the membership functions for the output variable, Flood Risk Level, indicating how the output values are categorized into low, medium, and high flood risk levels.
• flood_risk.view(sim=flood_risk_sim): Plots the final Flood Risk Level for given input values, illustrating how the fuzzy inference system computes the flood risk based on the input variable values and the defined fuzzy rules.

To interpret the plots, observe how each input variable is divided into categories (low, medium, high) based on the membership functions. These categories represent the degree to which an input value belongs to a particular category.

The output variable plot shows how the flood risk levels are determined based on the input variables’ membership values and the fuzzy rules defined in the system. The final plot, Flood Risk Level for Given Input Values, displays the aggregated output membership functions and the computed flood risk level as a single value.

1.3 2D Plot

The provided code generates a 2D contour plot of flood risk as a function of Precipitation Intensity and Land Cover, while fixing the values of Soil Moisture and Slope. Here’s a summary of what each part of the code does:

• Create grid points for input variables: Define a range of values for each input variable (Precipitation Intensity, Soil Moisture, Land Cover, and Slope) using np.linspace().
• Define compute_flood_risk() function: This function takes Precipitation Intensity (P), Soil Moisture (M), Land Cover (L), and Slope (S) as inputs and computes the flood risk using the fuzzy inference system (flood_risk_sim).
• Fix Soil Moisture and Slope values: Assign fixed values to Soil Moisture (M_fixed) and Slope (S_fixed).
• Create flood risk matrix: Initialize a matrix with the size of the combination of Precipitation Intensity (P_values) and Land Cover (L_values). Iterate through each combination of these values and compute the flood risk using the compute_flood_risk() function with the fixed values of Soil Moisture and Slope.
• Plot the 2D contour plot: Using plt.contourf(), create a contour plot that visualizes the flood risk as a function of Precipitation Intensity and Land Cover. The color map ‘viridis’ is used to represent the flood risk levels, with 20 contour levels.
• Add colorbar, labels, and title: Add a colorbar to represent the flood risk values, label the axes, and add a title that includes the fixed values of Soil Moisture and Slope.

To interpret the plot, observe how the flood risk values change as the Precipitation Intensity and Land Cover values vary. The plot shows how the flood risk is influenced by these two input variables while keeping the other two (Soil Moisture and Slope) fixed at specific values.

The contour lines in the plot represent different levels of flood risk, with the color intensity indicating the flood risk level. Darker colors represent lower flood risk, and lighter colors represent higher flood risk.

1.4 3D Plot

The provided code generates a 3D surface plot of flood risk as a function of Precipitation Intensity and Land Cover, while fixing the values of Soil Moisture and Slope. Here’s a summary of what each part of the code does:

• Create a 3D plot figure: Initialize a new figure using plt.figure() and add a 3D subplot with the projection='3d' argument.
• Create the 3D surface plot: Use the ax.plot_surface() function to create a 3D surface plot for Precipitation Intensity (Y-axis) vs Land Cover (X-axis), with the flood risk as the Z-axis. The color map ‘viridis’ is used to represent the flood risk levels.
• Add colorbar, labels, and title: Add a colorbar to represent the flood risk values, label the axes, and add a title that includes the fixed values of Soil Moisture and Slope.

To interpret the plot, observe how the flood risk values (Z-axis) change as the Precipitation Intensity and Land Cover values (X and Y axes) vary. The plot shows how the flood risk is influenced by these two input variables while keeping the other two (Soil Moisture and Slope) fixed at specific values. The color intensity on the surface indicates the flood risk level, with darker colors representing lower flood risk and lighter colors representing higher flood risk.

The 3D surface plot provides a more detailed visualization of the relationship between flood risk, Precipitation Intensity, and Land Cover compared to the 2D contour plot. You can observe the shape of the surface to identify areas with high or low flood risk and better understand the interaction between the input variables.

2 Minimizing Flood Risks under Maximum Precipitation

This chapter emphasizes the focus on reducing flood risks under the most challenging conditions (maximum precipitation) while highlighting the three main variables (soil moisture, land cover, and slope) being examined in the analysis.

Flood risk management is a critical aspect of urban planning and environmental protection. Understanding the factors that contribute to flood risks and identifying strategies to minimize these risks is essential for creating resilient communities. In this analysis, we explore the relationships between four key variables - precipitation intensity, soil moisture, land cover, and slope - to determine their influence on flood risk. Our goal is to identify the combinations of these variables that result in low flood risks, even under conditions of maximum precipitation.

Using a fuzzy logic-based simulation model, we examine the interactions between these variables and their impact on flood risk. The model incorporates expert knowledge and rule-based systems to predict flood risk levels based on various input scenarios. By analyzing the simulation results, we aim to provide insights into the conditions that can effectively mitigate flood risks, helping policymakers and urban planners make informed decisions for better flood management strategies.

The analysis includes a scatterplot matrix visualization that highlights the relationships between soil moisture, land cover, and slope under maximum precipitation conditions. By interpreting this matrix, we can identify patterns and correlations between these variables that contribute to lower flood risks. These insights will help guide future efforts in designing urban areas and implementing flood management measures that are both effective and sustainable.

The provided code performs a sensitivity analysis to minimize flood risks under maximum precipitation conditions. It evaluates flood risk categories based on all input variables, generates a dataset of data points with different combinations of soil moisture, land cover, and slope values, and finally creates a scatterplot matrix. Here’s a summary of what each part of the code does:

• Define the maximum precipitation intensity: Set the value of max_precipitation to 100, which is considered high precipitation intensity.
• Define the flood risk category function: Create a function get_flood_risk_category() that takes precipitation, soil moisture, land cover, and slope as input variables, and returns the flood risk category using the previously defined categorize_flood_risk() function.
• Generate data points for soil moisture, land cover, and slope: Create arrays of evenly spaced values for each of these input variables.
• Iterate through all combinations of soil moisture, land cover, and slope: For each combination, use the maximum precipitation value and the get_flood_risk_category() function to obtain the flood risk category. If the category is not None, append the combination to the data_points list.
• Create a DataFrame containing the data points: Convert the list of data points into a pandas DataFrame, which makes it easier to analyze and visualize the data.
• Create a scatterplot matrix: Use the seaborn library’s pairplot() function to create a scatterplot matrix of the data points, with flood risk categories represented by different colors.

Here’s how to read the scatterplot matrix:

• The diagonal plots (from the top-left to the bottom-right) are bar plots showing the distribution of each variable. These plots give an idea of the frequency of different values for each variable when the flood risk is low under maximum precipitation conditions.
• The off-diagonal plots are scatter plots showing the relationships between pairs of variables. These plots help identify any patterns or correlations between the variables. The color of the dots indicates the flood risk category associated with each data point.
• In the off-diagonal plots, if you see that dots of a specific color (in this case, low flood risk) are clustered in a particular region, it indicates that certain combinations of variables are more likely to result in low flood risk conditions.

To interpret the plots, consider the following:

• In the scatterplot between soil moisture and land cover, if there is a pattern or a specific region where low flood risk dots are clustered, it would suggest that there’s a relationship between these two variables that contributes to lower flood risks under maximum precipitation conditions.
• Similarly, in the scatterplot between soil moisture and slope, look for clusters or patterns of low flood risk dots to identify any relationships between these variables that contribute to lower flood risks.
• Finally, in the scatterplot between land cover and slope, examine the distribution of low flood risk dots to determine if there’s a connection between these variables that leads to lower flood risks.

By analyzing these plots, we can gain insights into the relationships between soil moisture, land cover, and slope that contribute to low flood risks even under maximum precipitation conditions.

3 Sensitivity Analysis

After obtaining the flood risk assessment results from the Fuzzy Inference System (FIS), we can assess the quality of the model by comparing its predictions to observed data. To do this, we’ll need a dataset containing historical flood events along with the corresponding values of the input variables (Precipitation Intensity, Soil Moisture, Land Cover, and Slope).

What if observation data on flood events never exist?

If we don’t have any observation data to compare the FIS model results, evaluating the model’s performance becomes more challenging. However, we can still follow some steps to ensure that your FIS model is reasonable and plausible:

• Expert knowledge: Consult with experts in the field of flood risk assessment to ensure that your fuzzy sets, membership functions, and fuzzy rules are realistic and based on sound principles. This can help us refine our FIS model even without actual observation data.
• Sensitivity analysis: Perform a sensitivity analysis to understand how the output flood risk varies with changes in input variables. By altering the input variables within their expected range and studying the corresponding changes in flood risk, we can gain insight into the behavior of the model and identify any unrealistic responses.
• Comparison with other models: If there are other flood risk assessment models available (either deterministic or statistical), compare your FIS model’s predictions with those from the other models. Although this is not a direct comparison with observed data, it can provide some indication of how your model’s performance compares to alternative approaches.
• Simulation data: If we have access to hydrological or hydraulic models that can simulate flood events, we can use the simulated data as a proxy for observed data. Although this approach has its limitations, as the simulated data may not perfectly represent real-world conditions, it can still provide valuable information for evaluating your FIS model.
• Temporal validation: If we have historical data for some of the input variables but not for the flood risk, we can still evaluate your FIS model by analyzing its performance over time. For instance, we can assess whether the model’s predictions of high flood risk align with periods of heavy rainfall, high soil moisture, or other conditions known to increase flood risk.

Remember that without observed data, it is more challenging to assess the performance of your FIS model accurately. However, following the steps outlined above can help us gain some confidence in your model and identify areas for potential improvement.

Let’s try Sensitivity Analysis

We’ll use the One-at-a-time (OAT) sensitivity analysis method to understand the effect of varying each input variable while keeping the others fixed. Assume that we have the FIS model already built and implemented in Python using the variables and fuzzy rules defined earlier.

This code performs a sensitivity analysis to study the relationship between input variables (Precipitation, Soil Moisture, Land Cover, and Slope) and the output variable (Flood Risk) in a FIS. It evaluates the FIS model for different values of the input variables, keeping the other input variables at their median values.

Here is a summary of the main steps in the code:

• Define the range and step size for each input variable.
• Calculate the flood risk for each input variable using the FIS model. The sensitivity_analysis function iterates over different values of each input variable while keeping the other input variables fixed at their median values.
• Categorize the flood risk levels (low, medium, high) based on the computed flood risk values.
• Plot the sensitivity analysis results, showing how flood risk varies with changes in the input variables.

The plot consists of four subplots, one for each input variable, with flood risk on the y-axis and the input variable on the x-axis. The background of each plot is filled with colors corresponding to the flood risk categories (low, medium, and high). The data points are plotted with different markers (‘o’, ‘s’, ‘x’) based on the input variable’s categories (low, medium, and high).

To interpret the plot, observe how the flood risk changes as the input variable value increases or decreases. A steep slope in the plot indicates that the flood risk is highly sensitive to changes in the input variable. If the flood risk remains relatively constant despite changes in the input variable, it suggests that the flood risk is less sensitive to that input variable.

To understand the meaning of the plot, consider that it represents how much the flood risk is affected by each input variable, given that other input variables are kept constant. By analyzing the plot, you can identify which input variables have a more significant impact on flood risk and prioritize interventions or mitigation strategies accordingly.

4 Summary

Flood risk assessment is a critical component of disaster management and urban planning. Accurate and reliable flood risk estimation helps authorities make informed decisions, prioritize resources, and implement effective mitigation strategies. With the increasing impacts of climate change and urbanization, there is a growing need for advanced techniques that can provide better insights into flood risk under varying conditions.

Fuzzy Inference Systems (FIS) offer a robust and flexible approach to model complex relationships between multiple input variables and an output variable, such as flood risk. By incorporating expert knowledge and handling uncertainties, FIS models can capture the intricacies of real-world systems, providing more accurate and reliable estimates of flood risk compared to traditional methods.

FIS models have gained popularity in the field of hydrological modeling and flood risk assessment due to their ability to handle imprecise and incomplete data, as well as their capability to incorporate human reasoning and intuition in the form of linguistic rules. This ability to integrate expert knowledge with quantitative data provides a valuable advantage, especially in situations where data availability is limited or uncertain.

The utilization of FIS in flood risk assessment typically involves defining input variables that influence flood risk, such as precipitation intensity, soil moisture, land cover, and slope. These variables are then used to estimate the flood risk level, which can be categorized into different levels, such as low, medium, or high.

To build an FIS model for flood risk assessment, the first step is to identify relevant input variables and their value domains. Next, fuzzy sets and membership functions are defined for each variable, followed by the formulation of fuzzy rules that describe the relationship between input variables and flood risk. These rules are derived from expert knowledge or empirical data and are used to determine the output flood risk level.

One of the critical aspects of FIS models is their ability to handle uncertainties and vagueness in the input data. This is particularly important in the context of flood risk assessment, where data can be scarce or subject to significant measurement errors. By using fuzzy sets and membership functions, FIS models can accommodate these uncertainties, providing more reliable and robust estimates of flood risk.

Sensitivity analysis is a valuable tool for evaluating the performance of FIS models in flood risk assessment. By varying input variables within their expected range and studying the corresponding changes in flood risk, modelers can gain insight into the behavior of the model and identify any unrealistic responses or potential areas for improvement.

FIS models can be further enhanced by incorporating optimization techniques to identify the most critical factors contributing to flood risk. This can help decision-makers focus on specific areas or interventions that have the most significant impact on reducing flood risk and improving overall resilience.

One of the challenges in applying FIS models for flood risk assessment is the lack of observed data for model validation. In such cases, the performance of the model can be evaluated using expert knowledge, sensitivity analysis, comparison with other models, or the use of simulated data from hydrological or hydraulic models.

In conclusion, FIS provides a promising approach for flood risk assessment, offering a flexible and robust framework for modeling complex relationships and handling uncertainties. By incorporating expert knowledge and quantitative data, FIS models have the potential to significantly improve our understanding of flood risk and support more effective decision-making in disaster management and urban planning.

The Weather Research and Forecasting (WRF) model is a powerful numerical weather prediction system used to simulate atmospheric phenomena at various scales. WRF produces a significant amount of output data that can provide valuable information for meteorological and climatological research, weather forecasting, and environmental management.

WRF output data is typically stored in netCDF files, which contain multiple variables with different units and dimensions. However, the netCDF files generated by WRF are not always following the Climate and Forecast (CF) metadata convention, which can make it difficult to interpret the data.

The CF metadata convention provides a standardized way of describing the metadata and units of variables in netCDF files. Adhering to this convention makes it easier to interpret the data and compare it with other datasets. However, the WRF output files are often not fully CF-compliant.

Visualizing the output from the WRF model can help researchers and practitioners to gain insights into various meteorological and climatological phenomena, including temperature, wind, precipitation, cloud cover, and atmospheric pressure. Visualization is an important step in understanding the data and extracting meaningful insights. Visualization can help identify patterns, trends, and anomalies in the data that might not be apparent from raw numerical output.

• Time series plots are a common way to visualize the temporal evolution of variables such as temperature, precipitation, and wind speed. These plots can reveal patterns and trends in the data and help to identify anomalies or outliers.
• Contour plots can be used to visualize the spatial distribution of variables such as temperature, pressure, and precipitation. These plots can show the magnitude and direction of the variables and help to identify patterns and features such as fronts, ridges, and troughs.
• Maps are a common way to visualize the spatial distribution of variables over a region of interest. Maps can be used to display variables such as temperature, precipitation, wind speed, and cloud cover, and can help to identify patterns and features such as mountains, coastlines, and rivers.
• Animations can be created from the WRF output data to visualize the temporal and spatial evolution of variables over a specific period. Animations can be useful for identifying trends, patterns, and anomalies in the data and for communicating the results to a wider audience.
• 3D visualizations can be used to represent the three-dimensional structure of atmospheric phenomena such as clouds and fronts. These visualizations can provide a more detailed and realistic representation of the data and help to identify features such as updrafts, downdrafts, and vortices.

There are several tools available for visualizing WRF output, ranging from simple plotting libraries to more advanced graphical user interfaces.

• One of the most popular tools for visualizing WRF output is the NCAR Command Language (NCL). NCL is a programming language designed specifically for scientific data analysis and visualization. NCL provides a powerful set of tools for working with NetCDF files, including the ability to subset and manipulate data, generate contour plots and maps, and create animations.
• Python is a popular programming language for data processing, analysis, and visualization. Python libraries such as Matplotlib, Cartopy, and Basemap can be used to create a wide range of visualizations from the WRF output data.
• Xarray is another popular Python library that can be used to handle and visualize multi-dimensional datasets. Xarray provides a powerful set of functions for data manipulation, analysis, and visualization and can be used to create a variety of plots and maps.
• R is a statistical programming language that can also be used for data processing, analysis, and visualization. R provides a wide range of packages for creating static and interactive visualizations, including ggplot2, lattice, and Shiny.
• Commercial software packages like ArcGIS and open-source Geographic Information System (GIS) software such as QGIS can be used to visualize WRF output data on maps and perform spatial analysis. QGIS provides a user-friendly interface for creating maps, visualizing data, and conducting geospatial analysis.

Regardless of the tool used, there are some key considerations to keep in mind when visualizing WRF output. These include choosing appropriate color maps, selecting appropriate contour intervals, and ensuring that the data is presented in a clear and understandable way.

It is also important to consider the spatial and temporal scales of the data when visualizing WRF output. For example, high-resolution data may require different visualization techniques than coarser resolution data, and data spanning multiple time scales may require animations or time series plots.

Visualization of WRF output is not only important for understanding the data but also for communicating results to stakeholders and decision-makers. Effective visualization can help convey complex scientific information in a way that is easily understood by non-experts.

Overall, visualization is a crucial step in the WRF modeling process, enabling scientists and researchers to extract meaningful insights from the vast amounts of data generated by the model. While there are many tools and techniques available for visualizing WRF output, it is important to choose the most appropriate tool for the specific task at hand and to consider best practices for data visualization to ensure clear and accurate representation of the data.

2 New CF based NetCDF files from native WRFOUT NetCDF files

The WRFOUT files created by the WRF model are not the most straightforward to interpret, and pose several challenges. These files include a series of three-dimensional fields that cover a specific region over a specific period of time, which can be used to analyze different meteorological phenomena, such as temperature, pressure, wind speed and direction, clouds, and precipitation.

However, WRF output data does not always follow the CF convention, especially when dealing with large datasets. This makes it difficult to interpret and analyze the data in its raw format, without the ability to visualize it. Additionally, the files can be very large and contain many variables related to the WRF simulation that may not be necessary for a research project.

Modifying the WRF registry can help to address some of these issues by allowing the user to change the variables included in the WRFOUT files. This can be a tedious and potentially messy process though, as some variables may be included in WRFOUT for reasons that the user may never know. As such, it is important to consider the potential unintended consequences of removing certain variables. Furthermore, the variables will still be on the staggered grid and on the model vertical levels, so visualization can still be challenging.

To address this issue, users can use the NCL to translate the netCDF output into a format that follows the CF convention. This can help to standardize the metadata and units of the variables, making it easier to interpret and compare the data.

wrfout_to_cf is an NCL based script designed to create CF compliant NetCDF files with user selectable variables, time reference, vertical levels, and spatial and temporal subsetting. This script is designed to be a simple, user-flexible post-processing utility and is particularly useful for research projects because it produces output files that are more convenient to work with.

However, wrfout_to_cf can be a relatively inefficient post-processing utility and may not be suitable for all applications. It is important to consider the pros and cons of each post-processing utility to determine which one is the best fit for any given research project

1. After running the simulation, WRF will produce three output files, you can check it by typing command in your simulation folder:

It will return

Download wrfout_to_cf.ncl from https://sundowner.colorado.edu/wrfout_to_cf/wrfout_to_cf.ncl and put it in the same folder with WRFOUT file. If you check one of file output using below:

You will get responses and see the netCDF structure, but it is not easy to understand.

Start to convert the native WRFOUT netCDF file to new CF based netCDF file using below command:

You will find a new file wrfpost.nc in the folder. Let check it using ncdump command below:

After you get the wrfpost.nc in place, you are ready to visualize it using various tools.

3 Visualizing the Wind

With 3 days of hourly information, there are several ways to visualize wind speed and direction to best illustrate the information.

Let’s start writing the code using Python.

After importing the library and defining the data, we can start visualizing the wind data.

Here are some options:

3.1 Time series plot

We can create a time series plot showing the variation of wind speed and direction over the three-day period. This type of plot is useful for identifying patterns or trends in the data over time. We can use Python libraries such as Matplotlib or Seaborn to create time series plots.

Above code will produce a map below.

As an alternative, we can have other visualizations for each grid as a line over time.

Above code will produce a chart below.

3.2 Wind rose plot

A wind rose plot can be used to show the distribution of wind direction and speed over the three-day period. This type of plot is useful for identifying the prevailing wind direction and speed. You can use Python libraries such as Windrose or Matplotlib to create wind rose plots.

Above code will produce a chart below.

3.3 Contour plot

A contour plot can be used to show the spatial distribution of wind speed and direction at a particular time during the three-day period. This type of plot is useful for identifying areas with high or low wind speed and direction. You can use Python libraries such as Matplotlib, Cartopy, or Basemap to create contour plots.

Above code will produce a map below.

3.4 Streamline plot

 A streamline plot can be used to show the flow of wind direction and speed at a particular time during the three-day period. This type of plot is useful for identifying the flow of wind over a geographic area. You can use Python libraries such as Matplotlib, Cartopy, or Basemap to create streamline plots.

Above code will produce a map below.

Overall, the choice of visualization method will depend on the specific research question or application, and the intended audience. It may be useful to try different visualization methods and compare the results to determine which method is most effective for the task at hand.

4 Notebook

The compilation for all the above code available as a notebook and hosted here: https://gist.github.com/bennyistanto/3f7877b44eebaf0db5e37fa8e7b8603a