Calculating Impervious Surface Area

Introduction

In this lab we learned how to calculate pervious and impervious surface area through the "Calculate Impervious Surfaces From Spectral Imagery" ESRI tutorial.

Impervious surfaces or surfaces that are impenetrable to water pose a serious threat to the environment due to contaminated run-off and flooding. The environmental impact of impervious surfaces causes some civil institutions to charge landowners with high amounts of impervious surfaces on their properties extra fees. Using aerial images we can calculate the surface area of impervious surfaces by segmenting and classifying the imagery.

Method

Segment the Image 

We began this lab by first segmenting the imagery, this will allow us to easier identify different features that are pervious and impervious. By segmenting the imagery we avoid errors in classifying the different surfaces. First, we extracted spectral bands to distinguish specific features more easily and tell if the are pervious or impervious. The resulting image will look like Figure 1 below.

Figure 1: Image with Extracted Spectral Bands. 

As you can see from Figure 1 the image looks much different than what a normal aerial image would look like. We chose a spectral band combination of 4 1 3, with vegetation being red, roads grey and roofs shades of blue and grey. The significant different in colors of man-made and natural resources allows us to classify them easier. Next, we segment the image by grouping pixels with similar spectral characteristics together. Segmenting the image will generalize and simplify groups of pixels to classify a smaller number of segments. Since we are only looking to classify pervious and impervious surfaces we do not need that many segments. 

Classify the Image

Now that the image is segmented we can classify the image. We classified this image into the two main classes of pervious and impervious, within these two classes we have smaller classes to represent types of land cover. We create training samples to classify the imagery. Training samples are polygons that represent specific areas of different land cover type. We created seven groups of training samples for our image. These groups were grey roofs, roads, driveways, bare earth, grass, water and shadows. Figure 2 below shows our classified image with the colors we have chosen for each training class. 

Figure 2: Classified Image

Figure 3: Training Samples

After classifying the image we then reclassify any errors in classification. If the data shows misclassifications we can go back and change them manually. I did not change any of the classifications in my data set. The resulting image is shown in Figure 4, this image shows only pervious and impervious surfaces. 

Figure 4: Pervious and Impervious Surface

Now we can begin to calculate surface area. We start by creating accuracy points, we made 100 points. These are shown in Figure 5 and 6. 

Figure 5: Accuracy Points

Figure 6: Accuracy Points Attribute Table

In the attribute table we must manually go in and change ground truth to 20 or 40 depending on the location of the point with 20 being pervious and 40 being impervious. Next, we computed a confusion matrix and tabulated the area. A confusion matrix will determine the accuracy of the classified and GrndTruth attributes. The confusion matrix results are shown in Figure 7 below. 

Figure 7 Confusion Matrix

Tabulating the area will show the pervious and impervious area within each parcel, this is shown in Figure 8 below. 

Figure 8: Tabulated Area

Now that we have calculated the impervious surface are we need to symbolize the parcels. We chose a yellow to red gradient to symbolize surface area. The symbolized parcels are pictured in Figure 9 below. 

Figure 9: Symbolized Parcels

Discussion 

Figure 10: Map of Impervious Area

In the map we are able to analyze the amount of impervious surface area per each individual land parcel. Local governments and civil institutions can utilize this data to determine storm water bills. We can determine how much impervious surface are each parcel has by comparing the color on the map with the corresponding color in the legend. As we can see the road contains the most impervious area of all the parcels which is what we had expected.