MAPPING AND GIS ANALYSIS OF HOUSEHOLD PONDS IN THE HALDA RIVER WATERSHED

ABSTRACT


INTRODUCTION
The watershed of the Halda River is compacted with large number of water bodies.Many natural water bodies behind manmade water bodies also exist in the watershed.But here the research is mainly focused on ponds.Ponds are mainly dug by the local people for different purposes.The main purposes are aquatic food production, house hold water supply, agricultural water supply, retaining water for dry season etc.Large number of fresh water ponds in a particular area also contributes to the expansion of aquaculture in that area.The wild catch of fish is decreasing day by day.In the year of 1983-84 the contribution of inland capture fisheries to total fish production was 62.59% whereas in 2016-17 it is only 28.14% (DoF, 2017).So from the trend it can be assumed that, in future, probably the culture of fish will mainly depend on artificial water bodies like ponds.
So, proper management of these ponds has much importance and the management is greatly depended on the correct information about these ponds.An intensive study of these ponds like this research work can reveal the proper baseline information and will be able to suggest policy maker and other related organization to take the proper steps for the best outcomes from these ponds.The watershed of the Halda River considered between 22° 23' 44.16" N to 22° 56' 8.16" N latitude and 91° 36' 59.04" E to 92° 01' 10.56" E longitude in this research.The total area of the Halda River watershed is approximately 1683 km 2 .It falls in three districts of Bangladesh named Chittagong, Rangamati & Khagrachhari (Chowdhury, 2015).In the south east region of Bangladesh, Halda is one of the major rivers.After originating in Ramgarh upazilla (sub district) and flowing through several upazilla, it finally falls into the Karnaphuli River (Akter and Ali, 2012).Watershed of Halda is comprised of many large and small ponds.
The hilly region of the Halda River watershed rarely has any pond but the valley area is densely packed with numerous ponds.Mapping and GIS analysis are very important tools for generating information on geographic feature.If the aerial view is unobstructed, the remote mapping is very much useful technology generating accurate and continuous maps of different objects across entire watersheds (Marcus and Fonstad, 2007).Water bodies can be mapped accurately by using satellite data viz.Landsat TM data (Frazier, 2003).In case of freely available high resolution satellite images, Google earth is providing quality image.By zooming into the target location on satellite images, shape and clear aerial view of ponds are available.By precise onscreen digitization of these ponds, their geometric and other spatial parameters can be analyzed in GIS software.Obtaining this information without remotely sensed image and GIS analysis would be expensive and somewhere impossible.

Digitizing
High resolution satellite image of Google earth is used (Figure 1).All ponds have been digitized using "Google earth" application.As ponds are polygon, so polygon tool is used to digitize ponds one by one across the whole watershed.During digitization, only ponds above 40 m length were digitized, those ponds are approximately more than 1000 m 2 .At the later stage, for GIS analysis, less than 1000 m 2 ponds were excluded and a separate statistic was generated for those small ponds based on intentional sampling of ponds by selecting 6 different sample areas in the whole watershed.In case of bad quality images, time slider was used to see the previous year's clear images.Among different year's images the best one was used.After completing digitization, the file of all digitized ponds was then saved as KML (key hole markup language) file.Keyhole Markup Language (KML) is a language which is based on XML and used to express geographical annotation and visualization (Samrat, 2012).Total 36 places were surveyed in the watershed for ground truthing.

Processing GIS data
For processing GIS data, QGIS 2.8 wien version was used, however it can be done in other GIS software also.Previously saved KML file of ponds was used to do further analysis.After loading the KML file in the GIS software the file is then saved as shape file (SHP file) which was then loaded in the GIS software as a vector layer.First of all the GIS software configured in a certain way such as the CRS (coordinate reference system) was set at WGS 84/UTM zone 46N and also all shape file was saved at CRS: WGS 84/UTM zone 46N.This was done because the study area is belongs to UTM zone 46N.

Calculating different geometric parameters of ponds in the attribute table of shape file (SHP file)
In the attribute table of shape file of digitized ponds, different geometric parameters of ponds like area, perimeter, length, width, aspect ratio etc. were calculated using different formulas.

Excluding small ponds (ponds less than 1000 m 2 )
As we have only dealt with the large ponds (greater than 1000 m 2 ) in the first stage.We had to get rid of small ponds (less than 1000 m 2 ).To eliminate small ponds from the database, the rows of small ponds in attribute table was selected by expression and then deleted.

EDA (Exploratory Data Analysis)
Exploratory data analysis could be done in both QGIS and MS Excel applications.But for convenience MS Excel application was used.For doing EDA, the attribute table was exported from the GIS application as excel file with desired fields.After opening the exported excel file, descriptive statistics was generated based on desired column that is shown in the result.

Elevation statistics
Elevation data of ponds were extracted by ENVI software and then further analysed by excel file.The statistics based on the elevation of ponds are then included in the result.

Creating RNG (Relative Neighbourhood Graph) of ponds
Relative Neighbourhood Graph (RNG) of ponds shows the relationship of ponds with their neighbouring ponds.It shows the distance among neighbouring ponds and also identifies the isolated ponds which have no neighbour ponds around them.Here, for the RNG graph, the cut off distance 500 meters is used, that means if any ponds stay 500 meters away from a pond, there will be no relationship between those ponds.

Mapping
To create a map, at first the shape file of Halda river, It's basin polygon file, it's stream line file and other shape file such as shape file of ponds and Relative neighbourhood graph etc. (necessary files that will be shown in the map) was loaded in the GIS software.From the properties of those layers, the colour, border width, fill colour etc. were changed according to their categories as needed.After changing all the properties, the desired layers which were shown in a particular map were activated.Such as for pond map, the layer of ponds and Halda River basin, main stream and its tributaries were activated and shown in the map.Legends, map title, scale, direction, latitude and longitude were included in the map as map component.

Method of small pond analysis
Six sample places of 500 m x 500 m approx.were selected intentionally in the whole watershed and digitized all small ponds (less than 1000 square meters) existed in those sample areas.Then the KML file of small ponds was imported into the QGIS and then to the Excel for further analysis.Descriptive data analysis tool was used for analysing the attribute data of small ponds that was exported from QGIS.

RESULT AND DISCUSSION
After analyzing all GIS data, a descriptive statistics table was generated (Table 1)

Area statistics of large ponds
From the descriptive statistics the average area of large ponds is found around 2670 square meters.The median most ponds are nearly 2276 square meters and the most abundant ponds are of nearly 2310 square meters.As we only studied the ponds more than 1000 square meters in the first stage, the minimum area of ponds among the studied ponds is 1000.20 square meters and the maximum is 54723.95square meters.The total area occupied by the large ponds is 17.97 km 2 that is 1.07% of the total watershed area and 3 % of the valley (here <20m elevation and relatively plain land that is 598.85 km 2 ).Total number of ponds those are above 1000 square meters is 6730.A bar diagram (Figure 2) and a line diagram are given below showing the frequency and cumulative percentage of large ponds according to their area:  The cumulative percentage graph (Figure 3) is showing that, about 90% large pond is less than 4100 square meters and the median area is about 2300 square meters.
A map of all large ponds is given here.Here we can see almost all ponds are in the valley area of Halda River watershed.

Pond map
Figure 4: The map of all large ponds in the Halda River watershed

Aspect ratio statistics of large ponds
Here aspect ratio is the ratio of length and width.High aspect ratio denotes the less square shape of a pond.If the aspect ratio is 1 then that pond will be a perfect square size pond.The average aspect ratio is found 1.46 and the median most aspect ratio is 1.45.The highest number of large ponds has 1.42 aspect ratio.The minimum aspect ratio is 1.01 that means it is the squarest pond among the studied ponds.And the maximum is 3.92.As it is a ratio of same dimension so it has no unit.A bar diagram (Figure 5) and a line diagram (Figure 6) are showing the frequency and cumulative percentage of large ponds according to their aspect ratio.In the bar diagram, the Highest number of large ponds (2905) lies between 1.4 and 1.6.Ponds having more than 3.6 aspect ratio is 2. Another important thing is that, number of ponds with aspect ratio "-1" is 50.This is because of nan values.These ponds had more than 4 sides.So their length and width were incalculable and ultimately their aspect ratio became nan values, that means "-1".The cumulative percentage graph is showing that, 95% large ponds have aspect ratio less than 1.9 and the median aspect ratio is 1.45.

Neighbourhood distance statistics of large ponds
The average distance among neighbouring large ponds was around 237 meters.The highest occurring distance among those ponds was 168 meter.The minimum distance among them was 30.07 meters and the highest (as 500 meters is used for cut off distance) was 499.60 meters.The total number of neighbourhood under 500 meters distance was found 16851  146) has less than 50 meters neighbourhood distance.The bar diagram (Figure 7) is showing that there is no significant difference in the number of ponds at 50 meters intervals.The cumulative percentage graph (Figure 8) is showing that, more than 90% large ponds are in between 400 meters neighbourhood distance and the median neighbourhood distance is around 220 meters.A relative neighbourhood graph (RNG) of all large ponds in the Halda River watershed is given here.This neighbourhood graph of all large ponds shows how close or far these ponds are situated and also show their density.Points are the representative of large ponds and connecting lines are the distances among them.

Relative Neighbourhood Graph (RNG)
Figure 9: Relative Neighbourhood Graph (RNG) of all large ponds If we zoom into the RNG map it will be looked like figure 10.

Length statistics of large ponds
From the descriptive statistics of large ponds the average length is around 60 meters.The median most length is nearly 57 meters and the highest number of large pond is with around 50 meters length.The minimum length found around 32 meters and the maximum length found around 307 meters.And the total number of ponds of which length is measured is 6680.Because ponds with more than 4 sides are not accounted for measuring length and their number is 50.So (6680 + 50) = 6730, the total number of ponds.A bar diagram and a line diagram are given below showing the frequency and cumulative percentage of large ponds according to their length:  The cumulative percentage graph (Figure 12) above showing that about 90% large ponds have length less than 90 meters and median length is about 57 meters.

Width statistics of large ponds
From the descriptive statistics, the mean width of large ponds is about 41 meters.The median most width of these ponds is about 39 meters.The line diagram (Figure 14) is showing that, 95% large ponds have length less than 60 meters and the median width is about 39 meters.

Elevation statistics of large ponds
By generating descriptive statistics on elevation data of large ponds, the average elevation of these ponds is found 11.75 meters.The minimum and maximum elevation is 1 and 40 meters respectively.The highest number of large pond is found at 7 meters elevation.But in case of elevation range, the highest number of pond is found in the range of 10-15 meters.Another Range of elevation with 2nd highest number of ponds is 5-10 meters.
Combining this two elevation ranges, it can be said that, most of the ponds are in between 5 to 15 meters of elevation from the sea surface.And if we see the pond map, can see that almost all of the ponds are in the valley of the Halda River watershed.Here is a bar diagram (Figure 15) showing the number of large ponds according to their elevation.From the cumulative percentage graph (Figure 16) it can be said that, 90% of large ponds are below 20 meters elevation and the median most elevation is 11 meters.

Result of small pond analysis and combined estimation
Six areas were intentionally selected in the watershed area and digitized all ponds existing in those sample areas.Here small ponds means, ponds having less than 1000 m 2 area.  2 showing that the number of small ponds approximately 5.5 times than large ponds (as large ponds found 6730), That is 5.5 small ponds for one large pond.The estimated total area occupied by the small ponds 26.23 km 2 which is 1.46 times than large pond's area, where the total area of large ponds is 17.97 km 2 .The area of small pond is about 1.56% of the whole watershed and 4.38% of the valley area (area <20 m elevation and relatively plain, which is 598.85 km 2 in total), where the area of large pond is about 1.07% of the whole watershed and 3% of the valley area.Collectively, these ponds occupy 44.19 km 2 and that is 2.63% of the whole watershed and 7.38% of the valley area, where the whole watershed is 1683 km 2 .The total number of ponds (both small and large) was then estimated 43745.

Figure 1 :
Figure 1: An overview of digitized ponds in the Halda River watershed

Figure 2 :
Figure 2: Frequency of large ponds according to their areaHere, the highest number of large ponds (1441) lie between 2000 to 2500 square meters.Combining top 4 bars in the bar diagram, it can be said that, most of the large ponds (above 1000 m 2 ) are in between 1000 to 3000 m 2 .The highest scale in the graph is 10000 m 2 and above this scale, there are only 74 ponds.That means extremely large ponds are very few in number.

Figure 3 :
Figure 3: Cumulative percentage of large ponds according to their area

Figure 5 :
Figure 5: Frequency of large ponds according to their aspect ratio.

Figure 6 :
Figure 6: Cumulative percentage of large ponds according to their aspect ratio

Figure 7 :Figure 8 :
Figure 7: Frequency of large ponds according to their distances

Figure 10 :
Figure 10: Zoomed view of the RNG (with the zoomed location in inset)

Figure 11 :
Figure 11: Frequency of large ponds according to their length Here (Figure 11), most of the large ponds (6390) have length in between 30 to 90 meters and the lowest number of ponds (1) having length in between 210 to 240 meters.

Figure 12 :
Figure 12: Cumulative percentage of large ponds according to their length

Figure
Figure Frequency of large ponds according to their widthHere (Figure13), most of the large ponds (6388) have width in between 20 to 60 meters and the lowest number of ponds (2) having width in between 160 to 180 meters.

Figure 14 :
Figure 14: Cumulative percentage of large ponds according to their width

Figure 15 :
Figure 15: Frequency of large ponds according to their elevation.

Table 1 :
Descriptive statistics of all important geometric and spatial parameters of large ponds

Table 2 :
Result of small pond analysis