The Lempa River is one of the rivers with biggest flow that end at the Pacific Ocean. It spans for 422 Km, making it the longest river in Central America, and discharges an average of 362 m^3 of water per second. Its watershed is about 17,790 square kilometers. 56% of its watershed is located in El Salvador which constitutes 49% of El Salvador’s territory including the capital, San Salvador. It provides drainage for the area where 77.5% of the Salvadoran population. It is by far the most important river in El Salvador.
San Salvador is the capital of the small country of El Salvador in Central America. During the civil war that occurred from the late 70s until 1992 thousands of people moved from the countryside smaller towns into the capital where they felt the war was less intense. San Salvador as a city previously isolated began to merge with the municipalities around it. Today the metropolitan area of San Salvador (A.M.S.S.) is formed by 14 municipalities, 2 of them located in a different “state” or “departamento”. Antiguo Cuscatlán and Santa Tecla belong to the state of La Libertad, and Apopa, Ayutuxtepeque, Cuscatancingo, Delgado, Ilopango, Mejicanos, Nejapa, San Marcos, San Martin, Tonacatepeque, San Salvador y Soyapango belong to the state of San Salvador. Each municipality has its own government as well as each state has a different representative in the Legislature. According to the 2007 census the population of the A.M.S.S. was 1,566,569 where in 1992 the population was 1,223,472 and in 1971 before the formal civil war only 581859.
Due to this rapid increase of the population and the political distribution of the city, the drainage and flood control techniques were always matters of big discussion but poor results. Being located in Central America the city of San Salvador receives rainfall, tropical storms and sometimes low intensity hurricanes during a period approximately of 6 months causing severe flooding, property damage, loss of homes as well as deaths. One of my experiences while I was in my hometown was in 2005 when severe storms flooded the barrios of La Vega, Modelo and Candelaria in San Salvador. Another event I remember occurred in 2008 when a Bus with 31 passengers was dragged by the strong current into the river Acelhuate in colonia La Malaga, killing all of them but a boy that managed to climb to the roof of a house nearby. The Acelhuate River is an important drainage for the city and usually runs dry, but when it rain it is common that it overflows.
The objective of this project is to use ARC GIS computer tools to develop a digital image of the main drainage rivers of the city of San Salvador, identify main flood spots as well as inputting rain data from local rain-gages in order to estimate the amount of rain that would make these places to flood and cause significant damages or threat the lives of citizens.
In order to do a watershed analysis I had to delineate the Lempa river basin and its river network on ARC GIS desktop. I started by looking on the Arc GIS website for files that would be useful for my purposes. I was able to find a GIS polygon shape file and a river network shape file that contained all the watersheds and rivers for Central America, Mexico and the Caribbean as shown in figure 2 and 3. I also found the slope direction raster for Central America, Mexico and the Caribbean shown in figure 4.
After obtaining these shape files I had to isolate the watershed as well as the rivers inside of it from the rest of the file and save it as a feature class on a new geodatabase that I created. First I had to manually identify the Lempa River Basin, to do this I uploaded a Bing Basemap into ARC GIS and selected the river basin with the select tool. After selecting the basin I saved it with the name Lempawatershed. After doing this, I will no longer need the big file containing all the watersheds so I can dele it from the project. Using the process on Execise 2 (Select by Location) I selected all the streams inside the watershed and saved them as “Lempa Streams” feature class. The result of this process is shown in figure 7.
The next step was to isolate the rivers that drain water from the city into the Lempa River. To do this I added a basemap into my project, which allowed me to know where the city was located and be able to manually select the streams that connect the city of San Salvador with the Lempa River. These two rivers are the Acelhuate River and the Las Cañas River as seen in figure 8. I decided to stop the selection before the Acelhuate meets the Guazapa River because the stream gage is located right before their junction.
To upload the major flood spots in San Salvador I used information provided by the National Service of Territorial Studies S.N.E.T. (Servicio Nacional de Estudios Territoriales). This agency is in charge of managing the alert system for the country and in its webpage collects a timeline of flood events, drainage collapses and other water related events. The events provided by the SNET do not contain any coordinates that can be interpreted by ARC GIS; it only provides a description of streets, landmarks and neighborhoods. In order to obtain the latitude and longitude of these locations I used the latitude and longitude application of Google maps. Since I am familiar with the streets I was able to pinpoint where these events occurred and copy into excel the latitude and longitude coordinates as shown in figure 9. Along with the latitude and longitude I also included in the spreadsheet the date of the event, a brief description of the event and the exact location as described by the SNET. I did this process from the latest events in August 2010 to the first event on June 2007.
With the latitude and longitude coordinates, I uploaded the spreadsheet and created a point layer using the display ‘X,Y Data’ option and later exporting the spreadsheet as a layer. When the layer was ready I was able to give a specific symbol for the different events as shown in figure 12.
The data acquisition for this project was the hardest part, not because it was hard to upload but because I had to wait a long time for it. After a month and a half of requesting it to the SNET I got daily flow averages for the stream gage on the Acelhuate River just upstream of Aguilares. I was not able to obtain any other values downstream from here because according to the SNET it was not available; on the other hand the data from this gage was quite useful since it is the one closest to the city of San Salvador. Using the coordinates of this gage provided by the SNET I was able to upload the stream data and confirm that effectively this gage is located on the Acelhuate River. I changed the symbology of the stream flow station to change of size for a given flow and enabled the time line, since these values had a date for each record. >
The rain data was obtained in two different sets. The first set of data was sent to me by the SNET and it contained continuous daily average rainfall from 3 rain gage stations they manage: Nueva Concepcion, Sensuntepeque and Chorrera del Guayabo. Unfortunately these rain gages were a far from my study area of San Salvador, but since it was high quality data considering it is difficult to obtain, I uploaded it to ARC GIS. The second set of data was obtained directly from the webpage of the SNET, for each major flood event they provide the rainfall value from a rain gage nearby. I was curious about why these values were not available continuously and I found out that these were values for each storm event. The value reported is the actual amount of rain it fell during each storm, and these were operated by private companies or ONGs. Following the same steps as for the flood spots I uploaded the rain-gage data to ARC GIS and created a layer for each station.
It was more difficult to manage three different layers for the SNET stations and another three layers for the private stations so I decided to merge them into two separate layers using the Merge tool. Once this was done I was able to give the same symbology to the three rain-gages close to the city and the same symbology to the ones provided by the SNET. The result is provided in figure 13 below, we can appreciate the three continuous rain gages with a uniform simbology, the intermittent gages also with a uniform symbology and the stream flow station.
The first thing I did with the data is to do an interpolation with the theissen polygon and find out which rain gage value would be more appropriate for the flood event given its proximity. Most of the events were on middle ground between the three as shown in figure 14, so I decided to use all three gages for my analysis. According to the theissen polygon I could delete the Procafe gage in the left lower corner since none of the events fall into its area. I decided to keep this gage since its closer to the mean elevation of the city as opposed to the Boqueron gage which is located at the top of San Salvador’s volcano.
After deciding to use the three of my gages I joined each of the layers with the flood layer in order use the attribute table to find a pattern between rain quantity and the floods. As shown in table 1 it is difficult to make a relationship using a table since we do not know what station is closer to each location, moreover it is important to notice the big gap in the data for 2010. Gaps in the data are one of the many problems I encountered during the development of this project. These gaps in the data occur because the webpage is not updated automatically and other duties are a priority before making this data available.
In order to overcome the spatial reference problem, I enabled the timeline on arc map and created a video that can be seen here . With this video I can fast-forward or rewind and analyze any given event. With this method I was able to see some relationship between quantity of rain and flood events, for example in July 4th 2008 Pro-café and Illopango stations recorded more than 100 mm of rain during the storm event that caused the overflow of river Acelhuate, in colonia La Malaga, dragging an entire bus with 30 passengers into its stream. Although there is some correlation between quantity of rain and water related disasters in San Salvador, there are some events that are caused with very low quantities of rain. For example the same river that killed 30 people in July, again overflew in August 28 this time no fatalities were registered but the rainfall during the storm event by the three stations were 1.6, 1.8 and 1.6 mm. One could argue that maybe the river was already at a high level when it experienced the new rain and that is the reason for the overflow, but this cannot be confirmed nor refuted because there are no stream gages inside the city of San Salvador. Even if we cannot look at the actual stream flow in the location we can look at the stream flow gage downstream and try to find our answer. Recall that the stream gage is located downstream where the flow is greater so the numbers following are recorded where the river transports more water than what can be appreciated figure 2 show. During July 4th 2008 a flow of 343 m^3/s was recorded and during August 28th 2008 only 200 m^3/s making the argument of the river having a higher flow at the moment less plausible. I can then counter argue that the flood of July 4th filled the stream path with debris that make the river more prone to overflow because of water accumulation. All these ideas are possible explanations that cannot be proven without any data. Another example of this uncertainty on flood prediction is the city of Soyapango, which is very close to the Illopango station and we could assume that it perceives the same rainfall as the rain gage indicates. If we look at the flood events in downtown Soyapango we can only find 1 in three years which occurred in July 12th 2009 and was provoked by a rainfall of 12.1 mm. But this city previously withstood storms with more than 100 mm of water and later withstood a storm with more than 200 mm of water without any flood.
This is a complicated issue that cannot be answered with the data I have, so I decided to do literature search for the answer. I found several documents stating that “Floods in San Salvador are not always provoked by big quantities of rain because in many occasions with low quantities of rain it could experience a big flood.” In my opinion part of the problem is the poor design of the drainage system and the reduction of the impervious cover. But all these problems have its roots in the lack of data; I was not able to find the current 100 year or the 50 year rain for San Salvador. According to official studies by the SNET these values can only be calculated with data from 1950 to the 1980s when the civil war started. After this period most of the systems were shut down until the mid 2000s which gives us a 20 year gap and estimations of 100 year rains that urgently need updates.
In conclusion floods in the San Salvador metropolitan area are not dependant only in the quantity of rainfall that it receives. There are many variables that come into play that cannot be accounted for such as levels of the rivers at any given moment, clogging of the waterways with garbage, contamination of river streams with debris etc.
In my opinion to do a good analysis of the reasons for constant flooding, one would need to first install stream gages on the main streams inside the city, calculate the increase of the pervious cover in the city, maintain an automatic 24 hour rain data collection system and make the data available to engineers other than the ones at the government’s agencies.
Aside from the difficulties presented to me I was able to effectively locate the main drainages of San Salvador although we could expand this into a bigger network using a high resolution DEM which might be very difficult to obtain for the region. I was able to pinpoint expected flood spots such as La Malaga, La Vega and Barrio Santa Anita, which flood several times every year, but it was also interesting to discover that there are several spots in the city which are in no way near a flood plain that get flooded at least once a year such as the Northern side of san Salvador, close to Mejicanos. If I expand my time study to a couple more years back I would be able to know if these northern areas have been in constant danger or it is just a coincidence of my sample.