Elementary GIS-Based Hydrologic Analysis using Remote Sensing Data

1.0. Introduction

Permit me to begin by stating that if you are a 21st century professional involved in environmental sciences and water resources engineering, and you lack functional knowledge of Geographic Information Systems (GIS), you are essentially working with one eye blind. As a matter of fact, anyone who is interested in, or deals with the spatial placement and relationships of data, objects, and services on Earth needs to know how to work with GIS. This article is intended for intermediate level GIS users, I however encourage beginners to tag along, you will definitely find useful information here.

Recently, I have been privileged to review several civil infrastructure designs for a World Bank funded project in Nigeria, and I have realized that there is an overwhelming deficit in expertise regarding hydrologic analysis. This is quite understandable considering that unlike other fields such as hydraulics, geotechnics, and structures; hydrologic analyses and designs require a lot of recurrent and spatially varying datasets which can be quite expensive to obtain and typically require decades of observed records. The relevant national & state agencies do not help matters because these datasets, even if available (which oftentimes they are not), are quite costly to obtain.

Recent scientific and technological breakthroughs, coupled with the benevolence of some international research institutions and organizations have however made this much needed data freely available to aid hydrologic analyses in any part of the world, especially in data scarce locations like Africa. The only challenge regarding this plethora of free datasets is how to visualize and manipulate them; this is where GIS comes to the rescue.

The main aim of hydrologic analysis in flood and erosion control is to estimate the peak rate or amount of runoff which a hydraulic structure has to convey or retain, for a given "design rainfall" depth or intensity. I will be conducting this analysis with reference to a Culvert structure; notwithstanding, this procedure can also be applied to drainage channels, dam spillways, and other relevant flood and erosion control infrastructure used for conveying storm runoff. I should also note at this point that this article is not a tutorial; I will be trying as much as possible to comprehensively introduce the basic procedures necessary to determine peak runoff for practically any site in the World using data gleaned from the internet. I will also be including direct links to all my internet sources so anyone can easily adapt the data for their locale; please drop a comment if any of the various links do not work.

2.0. Necessary Tools

• If you need to brush up your knowledge on hydrology, the processes involved, how hydrologic models work, and what hydrologic design entails, you will require a good reference text on hydrology. I recommend “Applied Hydrology” by Ven Te Chow et al; hereinafter referred to as Chow_1998.
• Next you will require a technical reference manual for your locality, this will contain codes, specifications, and data relevant to the region or country. I will be using the Highway Drainage Design Manual for Nigeria, hereinafter referred to as FMWH_2013. Hopefully you can obtain one relevant to your location if the site is outside Nigeria.
• GIS software is absolutely necessary for storing, analyzing, and displaying the plethora of data involved in hydrologic analysis. I personally prefer and use Global Mapper for quick work, and Arcgis?for serious stuff, both are proprietary software, but QGIS is an open source alternative that is quickly gaining prominence.?
• Finally you will need a spreadsheet program like Microsoft Excel for all your calculations. 21st century engineers serious about working with complex and iterative computations no longer use handheld calculators.

3.0. Location! Location!! Location!!!

It is expected that you are familiar with the proposed site and know the geographic coordinates at least. Please refer to this earlier article of mine if you need some tips on 21st century site reconnaissance. I currently reside in the Federal Capital City of Nigeria, so I will be selecting a particularly troublesome location that has been experiencing floods in recent years, it is a culvert close to Galadimawa roundabout, located precisely at coordinates 9° 00' 11.7754" N and 7° 25' 32.0251" E. Before I proceed further, I would also like to state that there is really no other way to depict your project location and immerse your client in your reports without employing visual aids, and what better way to do this than with maps and images; this is one of the many capabilities and advantages of GIS.?

4.0. Data

Once done with identifying your site location, the next step is to get the necessary site-specific data. This is usually where the problem with hydrologic analysis and design in Nigeria becomes cumbersome.

Topography

One of the first approaches to hydrologic modelling is terrain analysis, you need to understand the shape of the land because this is what dictates where and how fast water will flow. Normally a land surveyor is required to survey the site and obtain “levels” for generating a digital terrain model (DTM). However for large areas, which usually require proper terrain analysis, this is highly cost prohibitive, especially if there is no plan to develop the entire land area; this is where Remote Sensing using aircrafts, UAVs, and satellites come in.?

Large-area topographic data for Nigeria were initially available as topo sheets generated by aerial photography (circa 1954), and you are required to part with some good money to get your coverage area.

Recently, the Office for the Surveyor General of the Federation (OSGOF) has digitized these maps and further acquired a 20m resolution Digital Elevation Model (DEM) for Nigeria, the cost of purchase for personal use however still remains a significant hurdle. Thankfully, the United States Geological Survey (USGS) has made global digital terrain data, with a maximum resolution of about 30m available for free, this product is called the Shuttle Radar Topography Mission (SRTM). The 90m resolution data, which I will be using, can be downloaded here.

Land-use

Next, there is a need to know the development pattern of the site, estimate how much of the area is developed, and what type of development and vegetation cover exists there. For small areas, Google Earth imagery will suffice, but for much larger areas, the USGS also provides free lower-resolution satellite imagery via the LANDSAT missions.

Rainfall Records

The last dataset required for our hydrologic analysis is local rainfall data in the form of Intensity Duration Frequency (IDF) curves. These curves are generated by performing statistical frequency analysis on several years of maximum rainfall series of various durations. I will definitely not be going into the nitty gritty of IDF curve development here, but the key issue is that IDF curves are location or region specific because the characteristics of rainfall vary spatially. In plain language, you cannot use the same IDF curves for two locations unless they are in the same region and share similar rainfall characteristics. In Nigeria for example you can see from this figure available in FMWH_2013, that average annual rainfall depth generally increases from North (200mm) to South (2800+mm), you will also notice the Isopleths are not horizontal, this means rainfall depth also varies from West to East but not as starkly as from North to South.

There are IDF curves for 35 locations available in FMWH_2013 as depicted below

However, the challenges with these IDF curves are:

• They were generated using rainfall data from between 1948 to 1978, we know now that Climate Change would have had an appreciable effect on their accuracy.
• The IDF curves are just curves; no tables or equation coefficients included. It is necessary to distill these curves into an equation so engineers can more readily and accurately calculate the intensity values instead of simply eye-balling the curves.
• Lastly, 35 locations for a country as large as Nigeria is highly insufficient, there are many major cities currently without IDF curves.?

My current project mentioned here is aimed at addressing these deficiencies by using more up-to-date data, and increasing the available IDF curves to 72 locations. For now however, we will use the available curves in FMWH_2013.?

To address the spatial limitations, I have created Thiessen polygons to demarcate the “zone of influence” of each IDF curve.

I use this to aid in selecting the closest IDF curves to my site, as you can see, my city does not have an IDF curve so I will be using the closest one which is Minna.

Secondly, I had to digitize and distill each component of the FMWH_2013 curves to equations using nonlinear regression, so it is a lot easier and more accurate to look up the values.

The formula below is the general form of the IDF equation

where:

i, is the intensity to be determined (mm/hr),

d, is the duration (minutes),?

A, B, and M are the site-specific regression coefficients

Table below contains the regression coefficients for each return period at Minna site

5.0. Modelling and Analysis?

Now to the main business.

Terrain Analysis

Using GIS and the acquired DEM, the first step involves generating the drainage pattern and streams, after which we locate our design outlet and delineate the catchment area. Next we extract the geometric attributes of the catchment area and longest flow path (LFP).

Design Storm Frequency

Selection of a design return period is based on a careful balance between safety and economy, the higher the return period, the safer the structure, but obviously costlier because it will be required to convey a larger flow. We will be adopting the guidance provided in Chow_1998 below which is a 50yr return period, or more technically, a flood with 2% probability of recurrence.

The Model

We will be employing one of the simplest rainfall-runoff models in Hydrology, the Rational Formula. This empirical model depicts the relationship between rainfall and runoff as a simple linear equation?

where;

Q, is the peak discharge to be determined (m3/s),

k, is the unit conversion factor (0.278 for metric, and 1 for imperial),

C, is the dimensionless runoff coefficient with a value between 0 and 1,

i, is the average rainfall intensity (mm/hr) for the selected return period and duration equal to the “time of concentration”,

A, is the catchment area (sq. km).

Determining the parameters on the right hand side of this equation may seem deceptively easy, but bewilderment awaits you.

The C coefficient determines how much of the rainfall turns to runoff, a value of C = 1 indicates that no infiltration, interception, or retention occurs within the catchment, and all incident rainfall becomes runoff. A value of C = 0 on the other hand means Q = 0 and no runoff occurs. Both extremes are not feasible in practice but it is clear to see that an undeveloped and well vegetated catchment will have low C values, and an urbanized development with a high occurrence of impervious surfaces (roofs and roads) will have high values of C. There are tables provided in major hydrologic literature that give C values for possible combinations of land-use, land cover, and terrain slope; we will be adopting this table as found in Chow_1998.

The rainfall intensity, i, is selected from the IDF curves or equation based on a critical duration called the "time of concentration" (tc), this duration is defined as the time it takes for runoff to travel from the "hydraulically farthest point" in the catchment, to the outlet discharge point. That is quite a mouthful, but it essentially means that peak flow at the discharge outlet is achieved when flow from all parts of the catchment, especially from the farthest point, has reached this outlet. This catchment's "hydraulically farthest point" is depicted by the "longest flow path" delineated during the terrain analysis procedure.?

There are different methodologies and formulas for computing tc, but i will be using one of the simplest methods called the Kirpich formula which says,

where;

tc, is the time of concentration (minutes),

L, is the Longest flow path length (m),

S, is the slope of the longest flow path (m/m), this is essentially the elevation difference at the start and end, divided by the length.

The computed tc value is plugged into the IDF equation, along with the regression coefficients, to obtain the design rainfall intensity. Finally, the catchment area is taken from the attributes of the GIS delineated catchment.

6.0. Runoff computation

Considering this catchment is rather large and complicated, the runoff coefficient will be computed in a "weighted" manner by considering all the component land uses. For smaller more homogeneous catchments, a single value could be selected from the relevant tables. Note however that the land-use proportions below are gross approximations.

Finally, the peak flow is computed based on the aforementioned parameters as tabulated below.

There you have it, the design peak flow required to size the culvert is 922.52 m3/s.

Before we conclude, I'd like to introduce a caveat; the Rational model used for this introductory analysis has several limitations, chief amongst which is, it is not a reliable method for large catchments (> 80ha) because it over-estimates the peak discharge. I personally apply it for "quick and dirty" analyses when reviewing designs or cross-checking other methods. I encourage you to read up on the limitations in relevant hydrologic literature so you are aware of the application range. Further reading should also include other more sophisticated hydrologic models, it is my hope that I get to demonstrate another in the near future, thank you for following.