Analysis of Future Weather event Possible application for the region Amsterdam

(1)

Analysis of Future Weather event

Possible application for the region Amsterdam

Kevin van Diepen (K.H.H.) UvAnetID: 10198105

kevin.vandiepen@student.uva.nl

01/04/2014 – 30/06/2014 External supervisor: Bart van den Hurk (Royal Netherlands Meteorological Institute) External supervisor: Ben Wichers Schreur (Royal Netherlands Meteorological Institute) First supervisor: John van Boxel (IBED, University of Amsterdam)

(2)

significantly affect urban areas. The aim of this research is to investigate the behavior of a selected extreme events to projected changes in temperature and relative humidity and to investigate the influence of the Urban Heat Island on this extreme event. This is done by using the weather forecast model HARMONIE at a resolution of 2.5 km. The selected case is based on the observed weather event at Herwijnen on the 28th of June 2011. The main focus lies on the size of the area with precipitation and extremes of hourly precipitation. However, the HARMONIE model is not able to shape extreme local conditions for this event. The results of related researches are is taken over to investigate future circumstances and to investigate if precipitation is affected by urban areas. The increase of hourly precipitation intensities is approximately 15 mm/h or 30% as response to a 2°C warmer atmosphere for this case.

Furthermore, the area, which experiences high precipitation intensities, significantly increases at higher temperatures. Furthermore, it can be concluded that extreme precipitation events frequently occur above urban areas in the Netherlands. The net-effect of urban areas worldwide on precipitation remains unclear.

Keywords

Meteorology, extreme precipitation, Urban Heat Island effect (UHI), HARMONIE, climate change

(4)

1. Introduction

The climate is changing across the planet, largely due to human activities (IPCC, 2013 – AR5-WG1-SPM). Convincing evidence from observations and ice core records shows that atmospheric concentrations of important greenhouse gases have increased over the last few centuries. Likewise, the global mean surface temperature over land and oceans have increased over the last 100 years. Besides surface temperature, there are many more important

indicators of climate change, for example atmospheric water vapor, precipitation and extreme events (IPCC, 2013).

Precipitation records extend throughout the 20th century, including large variations which occur from year to year and on decadal time scales (Trenberth et al., 2007). Confidence in precipitation change globally averaged over land areas is low for the years prior to 1950 and medium afterwards because of insufficient data (IPCC, 2013).

Since the Third Assessment Report (IPCC, 2007), climate change studies have especially focused on changes in the global statistics of extremes. Later, the Fourth Assessment Report (IPCC, 2013) highlighted the importance of understanding changes in extreme climate events because of their excessive impact on society and ecosystems compared to changes in mean climate (IPCC, 2013). Events of extreme precipitation have a large impact on society, as they can cause floodings, disruption of infrastructure, erosion, agricultural crop damage and even loss of life (Attema et al., 2014). Many studies have found a statistically significant increase in the number and intensity of extreme precipitation events of durations ranging from hourly to a few days (Kunkel et al., 2013). The most consistent trends towards heavier precipitation events are found in central North America and assessment for Europe shows likely increases in more regions than decreases (IPCC, 2013).

Precipitation is an essential part of the global water cycle and also an important indicator for the changing climate. Therefore, proper understanding of the effect of the urban environment on precipitation will be increasingly important, as extreme precipitation events can result in significant floods in urban areas (Shepherd, 2005; Overeem, 2014). The meteorology of urban areas significantly differs from the weather over rural areas, which is called the Urban Heat Island effect (UHI). This is specifically experienced during summer time with clear skies and low wind speeds. Nightly temperatures are relatively high during these periods. Consequently, urban areas are likely to influence precipitation around the city (Steeneveld and Van Hove, 2011).

The main objective of this research is to investigate the behavior of a selected extreme event to projected changes in temperature and relative humidity. In addition, the influence of the Urban Heat Island on this extreme event is investigated by using Amsterdam as location of interest. This is done by using the weather forecast model Harmonie at a resolution of 2.5 km. The selected case is based on the observed weather event at Herwijnen on the 28th of June 2011. The main focus lies on the size of the area with precipitation. However, also the extremes in hourly precipitation and consequences for the Amsterdam area are considered in this research.

This is investigated with the aid of the following research question: Will future climate change lead to an increase in the area, which experiences a critical precipitation intensity, for the region Amsterdam?

(5)

2. Methodology

On the 28th of June 2011, the KNMI measured 79 mm of precipitation at Herwijnen between 20 and 21 local time. Never did a KNMI observation station measure an hourly precipitation that high (Lenderink et al., 2012). For 10-minute interval measurements at Herwijnen, an even higher intensity is found of 88.9 mm/hour. Due to its exceptional conditions, the Herwijnen event is chosen as case experiment for this research. In order to conduct the research and collect data, the Harmonie-model is used for this research.

2.1. Research design

Harmonie is currently used at the KNMI for short term weather forecast (0-48 hours) (KNMI, 2013). It is a mesocale weather prediction model, which operates at a resolution of 2.5 km. Harmonie consist of 540 by 600 grid cells and is centered at 3.0° E and 53.0° N (Attema et al., 2014). In most circumstances, also extreme events, the model provides proper forecasts (KNMI, 2013). The model is also equipped with the UHI effect (figure 1). It calculates the influences of walls, roofs, shadow and industry. Also heat is absorbed and released by concrete during summer (Tijm, 2010). For those reasons, the Harmonie-model is assumed to be an appropriate model for simulating extreme weather events like the Herwijnen case.

Figure 1. Simulated temperature in Harmonie. The model calculates slightly higher temperatures for urban areas, like Amsterdam and Rotterdam. As a result, the UHI-effect becomes visible (Tijm, 2010).

In order to investigate both the influence of future circumstances and the UHI effect, the Herwijnen case is simulated under three different circumstances within the Harmonie-model.

(6)

The first simulation (P0) within Harmonie projects the Herwijnen event as it is observed on

the 28th of June 2011. To create the event in the model observational data of the 28th of June are used and the boundary conditions are generated by ERA-interim. ERA-interim is a reanalysis of past weather using a general circulation model (GCM), that operates at a resolution of 1.5°, which is about ~166 km. Every 6 hours the grid cells within the GCM are updated. The Harmonie-model is used to interpolate the coarser resolution of ERA-interim to 2.5 km resolution and hourly updated of all grid cells (Attema et al., 2014). The observed circumstances of the Herwijnen event in 2011 are considered as present-day circumstance. P0

represents current precipitation.

The second simulation (P0 + Δt) consist of an extreme Future Weather event. For this

simulation perturbed data is used of temperature and humidity as predicted by the KNMI’06 scenarios for 2050. This was done by using an idealized approach with a +2°C warmer atmosphere assuming the relative humidity remains constant. The second part of the original plan was to investigate the response of the first simulation to the perturbed data. However, the HARMONIE-model failed to create similar conditions as the observed Herwijnen event. Consequently, the proposed Future Weather event in HARMONIE is canceled. The response of case 6 of Attema et al. (2014) is taken over to represent the Future Weather event. The research performed by Attema et al. (2014) investigates the dependency of the specific humidity on dew point temperature. Therefore, the response of a selection of 11 cases (table 1) over the Netherlands, characterized by intense precipitation, to perturbations in temperature and humidity is investigated in the HARMONIE-model.

Table 1. Extreme precipitation events analyzed by Attema et al. (2014). The highest measured intensity per hour are represented by prmax. The values for temperature, dew point temperature, and relative humidity are measured 2 m above the surface. They are averaged over the 9x9 grid points surrounding the precipitation maximum, 1 hour before the maximum occurred.

Case 6 is chosen because the precipitation field has clearly defined structures, making it easier to identify the general features of the precipitation response to the perturbations. Furthermore, this case has a relatively strong response to the perturbations compared to the other cases (Attema et al., 2014).

To investigate the effect of cities on precipitation, climatic radar data sets (Overeem, 2014) are used. The analysis of two recently build, climatic radar datasets provides more insights into extreme precipitation in the Netherlands, particularly in urban areas. The results can be

(7)

useful regarding the functioning of sewer systems in Dutch urban areas (Overeem, 2014). For each of the radar datasets is counted how often certain intensity thresholds were exceeded at a random location in the Netherlands. A pixel is marked as urban area when at least 10% of the surface consist of cultivated area. The first radar dataset covers the period 1998-2012 and provide 5-minute precipitation sums for grid cells with an area of 6 km². The second radar dataset contains precipitation data for the period 2009-2012 and consist of grid cells with an area of 1 km².

2.2. Data analysis

Harmonie projects the first simulation within the area of 49.0° - 55.9° and 0.0° - 11.1° longitude. The field in which Harmonie operates consist of 300x300 grid cells, which is an 750 x 750 km² area as the resolution is 2.5 km. Harmonie retains the original simulation field of Attema et al. (2014) during the second simulation. The first 100 km (almost 1°) from the edges is cut off in order to avoid boundary complications. The area of one grid cell is 6.25 km². The intensities of precipitation are calculated at each clock hour for all grids.

Consequently, ratios between intensity and area can be extracted from the HARMONIE model results.

As extreme precipitation events are often very local phenomena (IPCC, 2007), maybe spanning even less that one grid cell, it is important to find a relation between the mean precipitation over an area and the size of that area. This can be done by taking the average precipitation over multiple grid cells (1, 4, 9, 16, 25, 36 and 49 grid cells) and plotting the relationship between mean precipitation and area size. By means of interpolation and/or extrapolation, also the ratio between very small/large areas and intensity can be estimated

3. Results

The results of the described methodology are shown below. The model results of the created event in Harmonie (P0) are given first. Subsequently, the results of the Future Weather event

based on case 6 of Attema et al. (2014) are shown. Later, the analysis of Overeem (2014) is interpreted.

3.1. Present-day (Po) simulations versus Herwijnen observations

The observed conditions on the 28th of June are simulated within the HARMONIE model. 3 different grid cell locations are chosen to compare the model results with the observations. An overview of these locations is shown is figure 2.

Figure 3. The simulated temperature (left) and wind speed (right) for location A (red), B (magenta) and C (orange). Both the local (solid) and averaged (dotted) results are displayed for each location. The dotted results are averaged over 3x3 grid points surrounding the location of interest.

The precipitation measured at location A,B and C during the simulated period is shown in figure 4. The horizontal axis is divided into bins of hourly time intervals at which the accumulated precipitation of 1 hour is displayed for location A,B and C. For example, the measured intensity at 19 PM is the accumulated precipitation between 18:00 and 19:00 PM.

(8)

Figure 4. Local precipitation intensity (left) and the averaged precipitation intensity (right) for location A (green), B (cyan) and C (blue) plotted against observed precipitation intensities (red).

The simulated event shows high distribution of precipitation in time and location. The grid cell, at which the highest amount of precipitation accumulated, experienced high intensities of approximately 10-20 mm/h spread between 19 and 23 PM. Furthermore, the grid cell, which experienced the highest hourly intensity and has the most northern location of the 3 grid cells, received 22.1 mm between 15 and 16 PM. These two grid cells received precipitation from different showers, as the showers approached from the south west. This indicates

12 13 14 15 16 17 18 19 20 21 22 23 24 0 10 20 30 40 50 60 70 80

Time UTC [hour]

P re c ip it a ti o n [ m m ]

Simulated vs Observed (single) Herwijnen (A) Max Intensity (B) Max Accumulation (C) Observed 12 13 14 15 16 17 18 19 20 21 22 23 24 0 10 20 30 40 50 60 70 80

Time UTC [hour]

P re c ip it a ti o n [ m m ]

Simulated vs Observed (averaged) Herwijnen (A) Max Intensity (B) Max Accumulation (C) Observed 10 12 14 16 18 20 22 24 0 1 2 3 4 5 6 7 8 9 10 Simulated windspeed Time [hour] W in d s p e e d [ m /s ] Herwijnen (A) Max Intensity (B) Max Accumulation (C) 10 12 14 16 18 20 22 24 15 20 25 30 35 Simulated temperature Time [hour] T e m p e ra tu re [ o C ] Herwijnen (A) Max Intensity (B) Max Accumulation (C)

(9)

Figure 2. Geological map of the Harmonie simulation area. The locations of interest are marked by the red crosses; (A) the grid cell which represents the topographic location of Herwijnen in the model, (B) the grid cell which received the highest precipitation intensity and (C) the grid cell with the highest

cumulative precipitation. The x- and y-axis represent respectively the longitude and latitude.

In addition, the model results at these locations are averaged over the 3x3 grid points

surrounding the location of interest. The total of 9-grid cells represents an area of 56.25 km². In this way, the distribution of the simulated conditions around these locations is taken into account.

Figure 3 shows the simulated temperature (left) and wind speed (right) over a period of fourteen hours (10:00– 24:00). The simulated event shows a peak temperature at 16 PM close to 32°C and then slowly decreases to 20°C at 21 PM for the topographic location of

Herwijnen in the model (location A). For location B, high temperatures are simulated between 10 AM and noon with a peak temperature of 35°C at noon. The overall temperature at

location C is lower and cools down from 27°C to 15°C after noon. The simulated event shows high wind speeds of approximately 9 m/s between 19 and 22 PM for location A and around 17 PM for location B. The overall wind speed at location C is lower with a maximum of 7 m/s around 13 and 15 PM. All locations experiences different temperatures and wind speeds over the simulated period, which can be explained by the significant distance between the location of the grid cells within the model. The temperature and wind speed hardly changes when averaged with its surrounding area (dotted lines in figure 3), which applies to all locations.

0 2 4 6 8 10 49 50 51 52 53 54 55

Geological map of HARMONIE simulation area

Longitude [degrees] L a ti tu d e [ d e g re e s ] A B C

(10)

multiple showers with intense cores during the simulated event. The weather radar shows this was also observed during the showers on the 28th of June 2011 (figure 5 (left)).

Figure 5. Total precipitation for the observed (left) and simulated (right) event at 20:00 PM. Location A, B and C in the simulated event are marked by red crosses.

The simulated precipitation intensity at location B at 16 PM drops significantly when averaged with its surrounding area. This indicates local changes in the conditions which are indeed present as the temperature drops 5 °C before the shower took place.

The showers that occurred on the 28th of June 2011 were exceptional. Figure 6 shows the temperature, dew point temperature, wind speed and precipitation observed at Herwijnen that day. Just before the shower took place, the temperature was high with ~30°C. The dew point temperature of approximately 23°C, which is unusually high for the Netherlands, really characterizes the shower. The temperature dropped almost 10°C at the start of shower. Also the dew point temperature dropped with several degrees. The drop in dew point temperature, or absolute humidity, is likely due the transport of dry air from high altitudes to the surface caused by the upward motion of the cloud (Lenderink et al., 2012).

Figure 6. Time lapse of observations at Herwijnen on the 28th of June 2011. The observations consist of 10-minute precipitation (upper), temperature and dew point (middle) and wind speed (lower) (http://www.knmi.nl/klimatologie/nieuws/o nweer_juni11.html). Longitude [degrees] L a ti tu d e [ d e g re e s ] A B C 0 2 4 6 8 10 49 50 51 52 53 54 55 In te n s it y [ m m /h o u r] 0 5 10 15 20 25

(11)

The exceptional conditions were caused by a frontal zone, which approached from the south west. The high absolute humidity and vertical instability of the atmosphere resulted in heavy convection and strong precipitation. Between 20:00 and 21:00 local time high 10 minute intervals of precipitation of approximately 20 mm were observed, which is shown in figure 6 (Lenderink et al., 2012).

There is an approximate match between the observations and model results for temperature and wind speed at location A. However, the sharp temperature drop observed at Herwijnen around 20:00 local time only took about 30 minutes. Apparently, this is spread over several hours during the simulated event at location A. Furthermore, some severe wind gusts were observed of approximately 20 m/s at Herwijnen. The simulated wind speed at location A lacks these wind gusts. They also remain absent at locations (B and C) of high accumulative

precipitation. It is no surprise that different temperatures and wind speeds are simulated during the simulation period for location B and C as the distance between the grid cells is significant.

Any similarities between the observations and the model results remain absent regarding precipitation. Several high intensities (~20 mm) of 10-min precipitation were observed between 20:00 and 21:00 local time at Herwijnen. The same amount of precipitation only occurred twice at two different locations and only for hourly time intervals (figure 4). The highest intensities found for the model results occurred at 16 PM and 23 PM for respectively location B and C. Location A, which is the topographic location of Herwijnen in

HARMONIE, only 1.2 mm accumulated during the whole simulation. Nevertheless, it is not relevant and not likely that the simulated shower would occur above Herwijnen in the model. The lack of wind gusts and sharp temperature drop indicate that Harmonie is not able to simulate extreme local conditions for this event. The absence of high precipitation intensities, which are often local phenomena, enhances this statement. Furthermore, the overall

precipitation during the simulation is not proportional to the observations. High 10-minute intensities (~20 mm) were observed at Herwijnen, whereas the simulation only reached ~20 mm/hour. The purpose of this research is not to reproduce the observations precisely in time and location, but the model clearly failed to create extreme conditions as observed at

Herwijnen (June 28th, 2011). Therefore, it is irrelevant to investigate the response of this simulation in a warmer climate.

3.2. Future Weather event

Case 6 of Attema et al. (2014) consist of a band of precipitation which is caused by a shower crossing the Netherlands from the east to the north-west. The simulated event within

(12)

Figure 7. Total precipitation for 2 experiments: Left is the reference case (Eref) and right the perturbation case based on a homogeneous temperature shift of plus 2°C (ETplus).

The shower is observed at Lettele, which is located is the Dutch province of Overijssel. The simulated shower location has shifted a bit to the south-east, as the prominent area with rain is located in south-western Germany. Although it reflects the observations well. The structure of the precipitation field remains very similar in the perturbed experiment compared to the reference experiment. As expected by the literature, the ETplus experiment displays an increase

in maximum precipitation amounts. Furthermore, some areas with light rain in the reference simulation disappeared as a result of the perturbations, for example over the North Sea (Attema, 2014).

Figure 8 shows the peak intensities per hour for Eref and ETplus. Each particular grid cell is

averaged over certain areas of surrounding grid points. There is significant difference between the reference experiment and the perturbation experiment for individual grid cells (green). At certain time steps the reference experiment produces the highest intensities. This indicates that the occurrence of extreme precipitation at some locations may diminish or shift spatially or in time response to the perturbations. There is a strong overall response to the perturbations as hourly precipitation increases with +20 mm/hour at several time intervals for all scales. Locally, the intensity of precipitation per hour increases up to +100 mm/hour and decreases at larger scales. Nevertheless, for ~55 km² areas, which is roughly a quarter of the size of Amsterdam, still high intensities per hour are reached of approximately 80 mm/hour.

Accumulated precipitation over 24 hours [ETplus]

Longitude [degrees] L a ti tu d e [ d e g re e s ] -4 -2 0 2 4 6 8 10 12 48 49 50 51 52 53 54 55 56 57 58 A c c u m u la te d p re c ip it a ti o n [ m m ] 0 20 40 60 80 100 120 Accumulated precipitation over 24 hours [Eref]

Longitude [degrees] L a ti tu d e [ d e g re e s ] -4 -2 0 2 4 6 8 10 12 48 49 50 51 52 53 54 55 56 57 58 A c c u m u la te d p re c ip it a ti o n [ m m ] 0 20 40 60 80 100 120 0 5 10 15 20 25 0 20 40 60 80 100 120

Peak intensities per hour

Time [hour] In te n s it y [ m m /h o u r] Eref 4grid = 25km2 Eref 9grid = 56.25km2 ETplus 4grid = 25km2 ETplus 9grid = 56.25km2 0 5 10 15 20 25 0 20 40 60 80 100 120

Time [hour] In te n s it y [ m m /h o u r] Eref 1grid = 6.25km2 ETplus 1grid = 6.25km2

(13)

Figure 8. Case 6. The intensity of precipitation, which is averaged at multiple scales; 1x1 grid (upper left), 2x2 grid, 3x3 grid (upper right), 4x4 grid, 5x5 grid (lower left), 6x6 grid and 7x7 grid (lower right).

All time evolutions of hourly precipitation at different spatial scales contain a maximum peak intensity, which lies between 2 PM and 12 PM for smaller scales (figure8, upper) and between 2 PM and 6 PM for larger scales (figure 8, lower). For each scale, the maximum one-minute, hourly and daily intensity is displayed in figure 9. It contains a solid relationship between intensity maxima and the size of the area as the results are averaged. The difference in intensity maxima, between Eref and ETplus, increases when the time interval becomes smaller.

The difference is approximately 15 mm/h for hourly intensities. The increase, as response to the 2°C warming, is about 30% at all scales.

Figure 9. Maximum peak intensities for minutely (blue), hourly (red) and daily (magenta) intensities for the perturbation (dotted) and Eref (solid) and ETplus (dotted). The minutely and daily intensities are

converted to hourly intensities in order to compare the results.

0 5 10 15 20 25 0 20 40 60 80 100 120

Time [hour] In te n s it y [ m m /h o u r] Eref 16grid = 100km2 Eref 25grid = 156.25km2 ETplus 16grid = 100km2 ETplus 25grid = 156.25km2 0 5 10 15 20 25 0 20 40 60 80 100 120

Time [hour] In te n s it y [ m m /h o u r] Eref 36grid = 225km2 Eref 49grid = 306.25km2 ETplus 36grid = 225km2 ETplus 49grid = 306.25km2 0 50 100 150 200 250 300 350 0 50 100 150 200 250 300

Maximum peak intensities

Area [m2] In te n s it y [ m m /h o u r]

Max Minute Intensity Max Hourly Intensity Max Daily Intensity

(14)

Figure 10 shows relationships between intensity and area when the grid cells are considered as individual; no averaging is applied. It is clear that the frequency for extreme intensities shifts to the right for ETplus (red). Furthermore, it is found that for the total simulation period

the amount of grid cells which received no precipitation at all increased for the perturbed simulation. This may confirm the statement of Attema et al. (2014) that some areas with light rain in the reference simulation disappeared as a result of the perturbations. At last, the amount of grids, which is area, increases sharply for high intensity thresholds for the perturbation case. Not only the precipitation becomes more extreme, but for the total simulation field also the number of grids, which experiences certain intensities, increases sharply at higher temperatures.

Figure 10. Relationships between area and intensity displayed by a frequency table (left) and a bar graph (right) for the reference (blue) and perturbation (red) case. The frequency table includes only measurable precipitation (>1.0 mm/h) for the simulations. The horizontal axis is divided into 110 bins, where the markers (blue and red squares) are located at the middle of each bin. Gaps occur where no frequency is present for that particular intensity interval. For example, no frequency exist for Eref between the 59-60

mm/h interval. The bar graph shows the amount of grids which exceeds the 50, 60, 70, 80 and 90 mm/h threshold over the total simulation period. Each grid is considered as individual, no averaging is applied. The amount of grids are converted to area (km²).

3.3. The influence of the Urban Heat Island effect

The rural and urban areas used in the analysis of the climatic radar data sets for the periods 1998-2012 and 2009-2012 are shown in figure 11. This results in 9600 km² (27%) and 7900 km² (23%) for respectively the 6 km² and 1km² datasets on the total surface of 35.000 km².

50 60 70 80 90 0 500 1000 1500 2000 2500 3000

Intensity threshold [mm/hour]

A re a [ k m 2 ] Eref ETplus 10 20 30 40 50 60 70 80 90 100 110 100 101 102 103 104 105 Intensity [mm/hour] F re q u e n c y

Frequency table for measurable precipiation (>1.0 mm/h) Eref ETplus

(15)

Figure 11. Urban (grey) and rural (green) areas in the Netherlands for both radar datasets (Overeem, 2014).

Figure 12 shows the mean frequency of exceedance per year of the 6 km² dataset for 15- and 60-minute precipitation sums during the period 1998-2012. For example, the 60-minute precipitation sums of 60 mm (lower right) have an average occurrence of seven times per year within a 6 km² grid cell, of which two occur above urban areas.

Figure 12. Mean frequency of exceedance per year for 15- and 60-minute sums for the 6km² radar dataset during the period 1998-2012. The frequencies apply to a random place in the Netherlands (yellow) or urban area (grey) (Overeem, 2014).

(16)

Figure 13. Mean frequency of exceedance per year for 15- and 60-minute sums for the 1km² radar dataset during the period 2009-2012. The

frequencies apply to a random place in the Netherlands (yellow) or urban area (grey) (Overeem, 2014).

Figure 12 and 13 both display a red square at frequencies of extreme precipitation at urban areas. It represents the frequency which is expected when the observed extremes in the Netherlands would be divided in proportion to urban areas and rural areas. If the frequency at urban areas is higher than the red square, than precipitation is relatively more often measured at urban areas as would be expected based on the area of urban areas. The frequency at urban areas is approximately equal for the 6 km² dataset. This also applies to the less extremes of the 1 km² dataset (figure 13, left). However, the frequency at urban areas is clearly higher for the most extreme events (figure 13, right). For some thresholds the amount of observations is low. Nevertheless, the most extreme showers seem to occur more often at urban areas (Overeem, 2014).

4. Discussion

This section starts with explaining why the model results significantly differs from the

observation. Then, a number of general features of the results regarding extreme precipitation and precipitation with area are illustrated. Lastly, possible implications for the region

Amsterdam are discussed.

4.1. Present-day (Po) simulation versus Herwijnen observations

For the Herwijnen case analysis showed significant differences between model results and the observed conditions. This can partly be explained by the chaos theory of Lorenz (1963). The chaos theory consist of general laws, which would describe complex systems, such as the weather, more understandable and allows to predict more accurately how these systems will

(17)

behave. A chaotic system consist of working parts which are sensitive to even the slightest change. These slight perturbations can become greatly amplified through positive feedback (McKinney et al., 2012). Lorenz showed precisely how the evolution of weather could be changed by minute perturbations to the initial conditions (Slingo & Palmer, 2011). This is often called the “Butterfly effect”, which is the idea that the flapping of a butterfly in South America could eventually affect weather in North America (McKinney et al., 2012). In other words, when forcasting chaotic systems there is not a single solutions, but there are multiple possible realisations and all realisations are possible futures.

The KNMI uses the ECMWF forecasting system to forecast the weather in the Netherlands. The operational ECMWF forecast represents the expected weather in the Netherlands and runs at 16 km² resolution. This forecast is repeated 50 times at a lower resolution of 32 km². To each of these repetitions tiny perturbations are applied to the initial conditions. By this method, the behavior of the weather is explored (Perrson, 2011). An example of this so called ensemble forecast is shown in figure 14.

Figure 14. Ensemble forecast for precipitation (upper) and wind speed (lower) on the 6th of March 2006 for the next 10 days. It shows the operational forecast (red) at 16km2 resolution, the control run (blue), which is the operational forecast at lower resolution (32km2), 50 ensemble forecasts (green lines) and the ensemble mean (brown).

Figure 14 illustrates significant variety in the ensemble forecasts, which indicate that many solutions are possible with substantial differences between them. This is in agreement with the chaos theory. The extreme event observed at Herwijnen is one of the many solutions which could have occurred that day. It is very unlikely that HARMONIE would simulate exactly the same extreme solution as observed (John van Boxel, personal communication, June 11, 2014). At the resolution of 2.5 km the HARMONIE-model may not be able to create certain conditions as the size of a single grid cell already covers 6.25 km². Furthermore, interpolation within grid cells and with surrounding grid cells may suppress the emergence of extreme local conditions and the chaotic behavior of the weather during the simulation. At a resolution of 1 km² the model could be more capable of simulating the processes which form these exceptional conditions, for example turbulence, vertical temperature differences and

(18)

stability indices of the atmosphere (J.H. van Boxel, personal communication, 14 may, 2014). The cases dealt by Attema et al. (2014) show a wide spread in response to the perturbations. This could be explained by the chaotic behavior of these showers at small scales (Attema et al., 2014), which may be suppressed for certain cases.

4.2. Future Weather event

The amount of water vapor, that can be present in the atmosphere, increases about 7% per 1°C temperature rise. As the relative humidity is expected to remain constant for the Netherlands, the amount of water vapor in the atmosphere will increase at the same rate (7%) per degree temperature rise (Lenderink et al., 2012).

However, extreme events observed in the Netherlands clearly follow the dependency of approximately 14% per 1°C rise in temperature. The same dependency is found for Hong Kong (Lenderink & van Meijgaard, 2010). This is possible because convective forces (upward motion) within showers can increase. During condensation of water vapor, latent heat is released which results in additional warming of the air. As a result, the upward motion of convective showers trough the atmosphere is increased (Trenberth et al., 2003). This causes water vapor to transform into raindrops more quickly (Klein Tank & Lenderink, 2009). For case 6, a 2°C warming leads to anincrease of hourly intensities of approximately 15 mm/h or 30%. This roughly matches the dependency of 14% found for extreme events.

This is also found by Attema et al. (2014). Case 6 has a relatively strong response to the perturbations as the mean response of all cases is 11%. However, for cases with a strong response to the perturbations the mean response is 14%. The results found in this research may therefore apply for more extreme events described by Attema et al. (2014).

According to the KNMI’06 climate scenarios, the increase in mean summer temperature in 2050 ranges between +0.9°C and +2.8°C (Van den Hurk et al., 2006). A warming of 2 degrees would theoretically imply that extreme precipitation events in the present climate become (1.14)² -1 = 30% more intense in the future climate. If the observed event at Herwijnen would occur under future circumstances, the peak intensity could increase up to 88.9 mm/h * 1.30 = 115.6 mm/h. However, the projected 2 degrees warming for 2050 is relatively to 1990, whereas the Herwijnen event took place in 2011. Nevertheless, even for a 1 degree warming the maximum intensity would theoretically still exceed the 100 mm/h

threshold. The results found in this research prove that 100 mm/hour intensities can be reached even for a less extreme event than the Herwijnen event.

However, the intensity of precipitation does not give any information about the area which experienced this intensity. For example, it is not known if the measured 88.9 mm/hour intensity at Herwijnen applies to an area of 5 km2 or 50 km2. From the results is found that the area, which experiences a certain intensity, definitely increases at higher temperatures. Despite the grid cells may be spread over several clusters at the total simulation field, the results show significant increase for the highest intensity thresholds (figure 10). This indicates that the size of the intense core of the shower increases at higher temperatures.

4.3. The influence of the Urban Heat Island effect

Based on the results of this research it can be conclude that extreme precipitation events frequently occur above urban areas in the Netherlands. Furthermore, this frequency is much

(19)

higher for the 1 km² dataset compared to the 6km² dataset. This can be explained by the higher spatial resolution of the 1 km² dataset and the fact that intensity of precipitation can become more extreme for smaller areas (Overeem & Buishand, 2012).

However, the net-effect of urban areas worldwide on precipitation is unclear (Shepherd, 2005). Inoue and Kimura (Inoue and Kimura, 2004) showed that low-level clouds occur more often over the Tokyo area in the early afternoon. They suggested that low-level clouds form at the top of columns of rising air, which are enhanced by sensible heat flux in the urban area. Changnon ( 2003) found that freezing-rain occurrences in large cities are decreased by10%– 30% because of the urban heat island. The cities in his study included multiple major American cities. More directly related to precipitation processes, Takahashi (2003)

provided evidence more related to precipitation. He confirms that heavy rainfall in the Tokyo area occurred more often during recent decades. Fujibe (2003) connected the increase of surface convergence over the urban area to stronger convection processes over large Japanese cities like Tokyo. Diem and Brown (2003) found that anthropogenic activities in the Phoenix area appear to enhance summer precipitation in downwind areas, especially the Lower Verde basin. As possible cause they suggest increased convergence due to urban roughness or more urban aerosols in the atmosphere serving as condensation nuclei.

There are several other studies which show similar results (Shepherd, 2005). However, some other studies remain skeptic about the possible influence of urban environments. Lowry (1998) discussed several potential problems with the methodology, which is

used in many studies of urban-induced precipitation. Additionally, Tayanç et al. (1997) found no evidence of urban effects on precipitation in their study of four large cities in Turkey. Robaa (2003) suggested in an analysis of Cairo even that an inverse

relationship existed between the size of urban areas and precipitation.

Shepherd (2005) suggest that the modeling of climate / weather must include urban land surface and aerosol processes to better understand the influences of built-up land and urban aerosols on weather (short term) and climate (long-term). As case 6 has clear precipitation structures, it is a proper case to investigate whether urban areas, like Amsterdam, would affect the event. It would be interesting if the event could be spatially shifted within Harmonie, in such way that the intense core of the shower would pass the urban environment of Amsterdam for example.

4.4. Implications for the Amsterdam area

Until now, events of extremely high precipitation intensities are simulated in order to discuss climate adaption in the Netherlands. For example, Waternet performed simulations including precipitation intensities of 100 mm/hour – 150 mm/hour. These intensities have no physical support or scientific background (Eljakim Koopman, personal communication, April 28, 2014). The 100 mm/hour simulation performed by Waternet uses the WOLK-model to

simulate an idealized, static shower for the region Amsterdam. The shower is implemented by processing 100 mm of precipitation evenly spread over time and place for one hour above Amsterdam (Bas de Nijs, personal communication, June 11, 2014). The WOLK-model, or flood map, digitally displays where accumulation of precipitation occurs during the event. Despite that the shower shows no variability in time and place, the simulation provides insights in the consequences for Amsterdam when this shower would occur in future. The possible implications of a 100 mm/hour shower are shown in figure 15. There are many locations in Amsterdam at which floods would occur during the 100 mm shower. This applies

(20)

especially to the city center and surrounding districts, which are relatively old and compactly built. Waternet found more or less 30 serious bottlenecks in total (marked in red). At these locations approximately 200 mm accumulated in front of several buildings (Bas de Nijs, personal communication, June 11, 2014).

Figure 15. Map overview of the Amsterdam center area (Bas de Nijs, personal communication, June 11, 2014). Spots which are named in red indicate locations of high accumulative precipitation (>200 mm) during the event.

The bottlenecks found for the Amsterdam area are possible realizations for 2050, as the results show local intensities of +100 mm/hour.

The WOLK-Analysis provide proper analysis of accumulative precipitation throughout the city. In order to acquire a more realistic overview of possible floods, these extreme events require more detailed investigation. A proper solution would be the use of the 3Di-model. This model will be able to calculate dynamics of water at high resolution and on large scale. Together with the high computing speed of the model, 3Di is very suitable to deal with complex issues regarding the management of local floods (Wytze Schuurmans, personal communication, June 13, 2014). With the aid of 3Di, extreme events which are simulated within weather forecast models can be analyzed on high detail. On the basis of these analysis, municipalities can decide which areas need adaptations if a certain event would occur for that region. Linking simulated weather events to a dynamic waterflow model requires great precision as it concerns the safety of people and infrastructure. It is not possible to shape the observed event precisely within weather models. However, the distribution of precipitation in time and space is important. For example, 100 mm of precipitation evenly spread over one hour can be acquired by a single shower. This can also be reached in one hour by two intense showers including a 10-minute break of no precipitation. For the latter, the 3Di-model calculates drying processes and takes account of a wet surface after the first shower. The second shower will be drained more quickly, which decreases the flood risk. Furthermore,

(21)

high intensity showers of short durations are the most hazardous precipitation events

regarding urban areas. Proper simulation of certain observed intensities at certain locations is important for these events (Floor Speet, personal communication, June 25, 2014).

The 3Di-model is still under construction and not able to process the research scenarios for the region Amsterdam. If the model is ready, the interaction between weather forecast models and 3Di can be tested by processing weather events through 3Di. In this way more insights in linking simulated weather events to the 3Di model can be acquired.

5. Conclusions

In this research, the behavior of an observed extreme event at Herwijnen (28/06/2011) is investigated to projected changes in temperature and relative humidity. In addition, the influence of the Urban Heat Island on this extreme event is investigated by using Amsterdam as location of interest.

The HARMONIE model is not able to simulate the extreme local conditions for the

Herwijenen event. The significant differences between model results and the observations can partly be explained the chaotic behavior of extreme showers on local scale. At the resolution of 2.5 km the HARMONIE-model may not be able to create certain conditions as the size of a single grid cell already strikes a 6.25 km². At the resolution of 1 km² the model could be more capable of simulating the processes which forms these exceptional conditions.

The response of a case also studied by Attema et al. (2014) is taken over to investigate future circumstances. The increase of hourly intensities is approximately 15 mm/h or 30% as response to a 2°C warmer atmosphere for this case. Furthermore, the area, which experiences high precipitation intensities, significantly increases at higher temperatures. Results from Overeem (2014) are interpreted to investigate if precipitation is affected by urban areas. Based on these results it can be conclude that extreme precipitation events more frequently occur above urban areas in the Netherlands. The net-effect of urban areas worldwide on precipitation remains unclear.

Waternet performed simulations including precipitation intensities of 100 mm/hour. There are many locations in Amsterdam at which approximately 200 mm accumulated in front of several buildings during the 100 mm shower. This applies especially to the city center and surrounding districts, which are relatively old and compactly built. In order to acquire a more realistic overview of possible floods, these extreme events require more detailed investigation. With the aid of the 3Di-model, extreme events simulated by weather forecast models can be analyzed on high detail. In this way Amsterdam is able to deal with complex issues regarding the management of local floods.

Acknowledgements

For this research, a small internship is done at the Royal Netherlands Meteorological Institute in order to write the bachelor thesis at the University of Amsterdam. I would like to express my thanks to Bart van den Hurk and Ben Wichers Schreur, for the guidance and supports through this project and for enlighten me with new knowledge and skills regarding climate, weather and modeling. Also, I would like to thank John van Boxel for all the questions I had during this research and for the many thoroughly and helpful feedbacks I received. Finally, my thanks go to Eljakim Koopman for the informative support and helpful contacts.

(22)

References

Attema, J.J., Loriaux, J.M., Lenderink, G. (2014). Extreme precipitation response to

climate perturbations in an atmospheric mesoscale model. Environ. Res. Lett., 9, 12 pp.

Chagnon, S.A. (2003). Urban modification of freezing-rain events. J. Appl. Meteor., 42, 863–

870.

Diem, J. E., Brown, D. P. (2003). Anthropogenic impacts on summer precipitation in central

Arizona, U.S.A. Prof. Geogr., 55 (3), 343–355.

Fujibe, F. (2003). Long-term surface wind changes in the Tokyo Metropolitan area in the

afternoon of sunny days in the warm season. J. Meteor. Soc. Japan, 81, 141–149.

Inoue, T., Kimura, F. (2004). Urban effects on low-level clouds around the Tokyo

metropolitan area on clear summer days. Geophys. Res. Lett., 31, L05103.

IPCC, (2007). Climate Change 2007: The Physical Science Basis. Contribution of Working

Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change [Solomon, S., D. Qin, M. Manning, Z. Chen, M. Marquis, K.B. Averyt, M.Tignor and H.L. Miller (eds.)]. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA, 996 pp.

IPCC, (2013). Climate Change 2013: The Physical Science Basis. Contribution of Working

Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change [Stocker, T.F., D. Qin, G.-K. Plattner, M. Tignor, S.K. Allen, J. Boschung, A. Nauels, Y. Xia, V. Bex and P.M. Midgley (eds.)]. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA, 1535 pp.

Klein Tank, A.M.G. & Lenderink, G. (2009). Climate change in the Netherlands;

Supplements to the KNMI’06 scenarios, KNMI, De Bilt, The Netherlands.

KNMI. (2013). Betere kansverwachting extreem weer. Beschikbaar via

http://www.knmi.nl/cms/content/113873/betere_kansverwachting_extreem_weer

Kunkel, K.E., Karl, T.R., Brooks, H., Kossin, J., Lawrimore, J.H. and ## other authors. (2013). Monitoring and Understanding Trends in Extreme Storms: State of

Knowledge. Bull. Amer. Meteor. Soc., 94, 499–514.

Lenderink, G. & Van Meijgaard, E. (2010). Linking increases in hourly precipita-tion

extremes to atmospheric temperature and moisture changes. Env. Res. Lett., 5, 025208.

Lenderink, G., Attema, J., Kew, S., Selten, F., Ter Maat, H. (2012). Future Weather. KfC

report number 83. Nationaal Onderzoeksprogramma Kennis voor Klimaat (KvK), 48 pp.

Lorenz, E. N. (1963). Deterministic nonperiodic flow. J. Atmos. Sci., 20, 130–141.

Lowry, W. (1998). Urban effects on precipitation amount. Prog. Phys. Geogr., 22, 477–520. McKinney, M.L., Schoch, R.M., Yonavjak, L. (2012). Environmental Science: systems and

(23)

Overeem, A., Buishand, T.A., Holleman, I., Uijlenhoet, R. (2012). Statistiek van extreme

gebiedsneerslag in Nederland, H2O, 45 (12), 29-31.

Overeem, A. (2014). Inzicht in extreme neerslag in de stad op basis van langjarige

radardatasets met veel ruimtelijk detail. Ervaringen met de aanpak van regenwateroverlast in bebouwd gebied, 284-305.

Persson, A. (2011). User guide to ECMWF forecast products. ECMWF, 119 pp.

Robaa, S. M. (2003). Urban-suburban/rural differences over Greater Cairo, Egypt. Atmosfera, 16 (3), 157–171.

Shepherd, J.M. (2005). A review of current investigations of urban-induced rainfall and

recommendations for the future. Earth Interactions. 9 (12), 1–27.

Slingo, J., Palmer, T. (2014). Uncertainty in weather and climate prediction. Phil. Trans. R.

Soc. A., 369, 4751-4767.

Steeneveld, G.J. & Van Hove, L.W.A. (2011). Een eerste inschatting van het Urban Heat

Island effect voor Rotterdam en omgeving. Wageningen University Meteorology and air quality, 14 pp.

Takahashi, H. (2003). Secular variation in the occurrence property of summertime daily

rainfall amount in and around the Tokyo Metropolitan area. Tenki, 50, 31–41.

Tayanç, M., Karaca, M., Yenigün, O. (1997). Annual and seasonal air temperature trend

patterns of climate change and urbanization effects in relation to air pollutants in Turkey. J. Geophys. Res., 102 (D2), 1909–1920.

Trenberth, K.E., Dai, A., Rasmussen, R.M. & Parsons, D.B. (2003). The changing

character of precipitation. Bull. Am. Met. Soc., 84, 1205-1217.

Trenberth, K. E., Smith, L., Qian, T., Dai, A., Fasullo, J. (2007). Estimates of the global

water budget and its annual cycle using observational and model data. J. Hydrometeor., 8, 758-769.

Tijm, S. (2010). De nieuwe generatie weermodellen. KNMI. Beschikbaar via

http://www.google.nl/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&ved=0CCI QFjAA&url=http%3A%2F%2Fwww.knmi.nl%2Fklimatologie%2Fweeramateurs%2F 20100605%2FPresentaties%2FTijm.pdf&ei=zqqwU4ymCcG7PdGvgPAL&usg=AFQj CNElMbV6r1oG8o8VaI7Q3QR9Fvfp6A&bvm=bv.69837884,d.ZWU

Van den Hurk, B., Klein Tank, A., Lenderink, G., Van Ulden, A., Van Oldenborgh, G.J. and ######## other authors. (2006). WR 2006-01: KNMI Climate Change Scenarios

Analysis of Future Weather event Possible application for the region Amsterdam