Application of Geographic Information Systems in Groundwater Prospecting : A Case Study of Garissa County , Kenya

Groundwater prospecting in Kenya has been haphazard and expensive due to lack of information on the appropriate areas for hydrogeological exploration and drilling of boreholes. Drilling in areas without prior knowledge about their groundwater potential has been leading to the drilling of numerous dry boreholes. In this study, we explored the use of Geographic Information System as a pre-analysis tool to identify zones with groundwater potential for Garissa Country. Factors that contributed to groundwater occurrence were identified as landcover, soil type and rock formation. The groundwater potential zones were generated by analysing thematic data of the three factors and integrating the musing Weighted Index Overlay Analysis (WIOA) method. The groundwater potential zones were validated by comparing the predicted potentials with actual yields of existing boreholes drilled within those areas. Results indicate that, whereas the model correctly predicted areas with low or no groundwater potential, it performed sparingly well when predicting areas with good groundwater potential. The study conclusively identified areas where groundwater prospecting should not be attempted and other alternative methods of surface water provision should be explored.


Introduction
Kenya is classified as a water scarce country, characterized by high spatial and temporal variability in rainfall leading to extreme droughts and floods.Kenya's renewable fresh water supply is estimated at 647 m 3 per capita, almost half the United Nations' recommended bench mark of 1000 m 3 per capita.This compares dismally with its neighbours namely Uganda with 2940 m 3 and Tanzania with 2696 m 3 per capita respectively [1].Kenya's fresh water supply is reducing due to declining rainfall, increase in population, and degradation of existing water catchment/conservation forest cover, and is projected to drop to 245 m 3 per capita by the year 2025 [1].Among the economically underdeveloped areas of the country, northern Kenya is the most vulnerable since water, arable land and pasture are scarce resources [2].Famine and drought are common in this region and coupled with underdeveloped water supply facilities, water sources are a major cause of conflict between local communities [2] [3].
Northern Eastern Kenya covers the largest part of the country but has the greatest scarcity of water.This problem is as a result of many factors.Traditionally, water security has been achieved by harvesting surface water through construction of river flow obstruction/storage structures such as dams and water pans [4].However Northern Kenya lacks suitable embankment materials and sites for construction of dams.Construction of dams would require transportation of suitable embankment materials from borrow sites in far regions which is an expensive exercise due to the bulky nature of these materials.High temperatures and poor vegetation cover that characterise the region lead to high evaporation and siltation rates respectively greatly reducing the lifespans and storage capacities of the water pans.
Groundwater source provides a viable alternative to surface water harvesting, and has proven useful in dry areas [5] [6].However, groundwater resources in Kenya are underdeveloped with only 0.18 billion cubic meters extracted annually from a total estimated yield of 1.08 billion cubic meters [7].Therefore there is need to identify and map potential groundwater harvesting zones in the Northern region as well as in other Arid and Semi-Arid Land (ASAL) areas in Kenya.
Garissa Country experiences water supply problems when surface water sources dry up during dry seasons.All the hinterland rivers are seasonal (Figure 1) and only River Tana flowing along the southern border offers perennial water source to the nearby communities and towns.
For many years, ground water harvesting has been tried in various parts of the country by the national government and non-govern mental organizations as an alternative water source.However, the exploration has been haphazard due to lack of information regarding groundwater potential areas.Overtime, drilling has relied on hydrogeological estimates and data from nearby boreholes, if any, which has led to the drilling of many dry or low yielding boreholes.Drilling of dry boreholes is a waste of time and precious resources.This negatively affects the livelihoods of the local community.Therefore there is an urgent need to utilize efficient pre-exploration methods to enhance use of all valuable resources.
In recent years the use of Geographic Information Systems (GIS) and Remote Sensing (RS) has made it easier to define the distribution of different groundwater The overall objective of the study was to develop a groundwater potential zones map for Garissa Country, using Weighted Index Overlay Analysis (WIOA) modelling, for selection of areas suitable for drilling of boreholes.
The specific objectives of the study were: 1) To identify factors that influence occurrence of groundwater in an area; 2) To establish suitable locations for exploration of groundwater for Garissa Country; 3) To test the validity of the generated groundwater potential map.

The Study Area
Garissa Country is comprised of the former Garissa and Ijara districts.The Country covers an area of about 34,952 km 2 and has a population of more than 623,060 [8].It borders Wajir Country in the North along Habasweni swamp and Lamu Country in the East.In the South, Tana River runs from west to east and for ms its boundary with Tana River Country.On the western side it borders Mt.Kenya game reserve and Isiolo Country.It lies in between latitudes 2˚01'30"S & 0˚59'36"N and longitudes 38˚40'20"E & 41˚34'40"E (Figure 2).Among the counties in the Northern region, Garissa Country was chosen as a priority for this case study because of three main reasons.First, it is one of the economic gateways to the region.Second, Garissa has been characterised by insecurity for many years which curtailed water infrastructure develop ment for long.Third, Garissa Country is projected to have a significant increase in population and economic development due to the proposed Lamu Port South Sudan Ethiopia Transport (LaPSSET) corridor infrastructure development which will pass through the country (Figure 3).

Existing Boreholes Data
The Ministry of Water and Irrigation drilled boreholes nationally at a high rate  Many boreholes records were incomplete and out of 218 boreholes drilled, only 111 borehole sites were georeferenced (Table 1).

Garissa Country Boreholes
From the regional data the country data was arrived at by carrying out a comparison of the data for the seven counties.The comparison was based on the number of boreholes drilled and the number of boreholes with GPS coordinates.Garissa Country was found to have the largest number of boreholes with complete data records.The georeferenced country data was further filtered to re Journal of Geographic Infor mation Syste m move repeated, erroneous and inconsistent records to obtain the data records that constituted the study validation data (Table 2).Woodland, Dense bushes, Sparse bushes, Grassland, and Swamps (Figure 4).

Soils Data
Data on the soils type in the study area was downloaded from the International Livestock Research Institute (ILRI) GIS portal (http://www.ilri.org/gis).
The data was initially generated from a study done by the Kenya Soil Survey (KSS) in 1982, and thereafter revised in 1997.The soil data was classified into 4 types; Clay, Very Clay, Loamy and Sandy Soils (Figure 5).The data was classified into the following groups: Igneous, Metamorphic, Sedimentary and Unconsolidated rocks (Figure 6).

Existing Boreholes
The data in Table 1 was used to generate a map layer showing the locations and yields of existing boreholes (Figure 7).

Groundwater Factors Conversion to Raster
To make the Factors layers integration possible the factors data was converted from vector for mat to raster for mat using ArcGIS10.1 Arc Toolbox (conversion tools-to raster and feature to raster).The thematic factors raster layers are shown in Figure 8, Figure 9 and Figure 10.

Weighting the Factors
After reclassification, the three thematic (factor) layers were weighted using the Analytical Hierarchy Process (AHP) method.AHP is a logical framework that is used to deter mine the relative input of each factor towards accomplishing a certain output [10].AHP involved pairwise comparison of the three variables (fac-Journal of Geographic Infor mation Syste m tors) with respect to each individual variable's relative influence on groundwater potential.The comparison was done on a scale of 1 -4 as follows; lithology is 2 times as important as soils; soil is 3 times as important as landcover and      lithology is 4 times as important as landcover.The comparison was expressed as a ratio and tabulated (Table 6).
The pairwise comparison generated a matrix that was manipulated to produce its Eigen vector.The computation stopped when the difference of Eigen vectors in two consecutive calculations was smaller than 0.001 a prescribed value.The Eigen vector gives the factors weights (Table 7).

Integration of the Factors Layers
After weighting, the three Factors (Vegetation, Soils and Lithology) were integrated (added) using ArcGIS10.1 Arc Toolbox (Spatial Analyst Tools-Overlay-Weighted Overlay) to produce the final output (Results) which is the Groundwater potential zones.The Output produced two classes of groundwater potential zones namely; medium and low yield zones (Figure 14).

Overlay with Existing Borehole Yields
This was done by overlaying the groundwater potential zones layer (Figure 14) with the existing boreholes layer (Figure 7) and evaluating the predicted ground water potentials against the actual borehole yields.The overlay produced the validation map shown in Figure 15.

Classification of Existing Borehole Yields
After the thematic data was integrated (added) the output (groundwater Figure 14.Garissa Country groundwater potential zones. Figure 15.Overlaid groundwater potential zones with existing boreholes.Total 21 potential zones map) produced two classes of groundwater potential zones namely low and medium.In this regard the existing boreholes yields were classified into two classes; 0 -7 low and 8 -20 good in order to enable graphical and statistical analysis (Table 8).

Ranking of Borehole Yields
To enable evaluation of the predicted groundwater potential zones against the yield values of the existing boreholes, the existing boreholes classes were ranked on a scale of 1 to 2. Good yield boreholes were ranked 1, and low yield boreholes were ranked 2 as shown in Table 8.

Ranking of Groundwater Potential Zones
The groundwater potential zones were ranked on a scale of 1 and 2. Good po-Journal of Geographic Infor mation Syste m tential was ranked 1 and low potential 2 as indicated in Table 9.

Validation Process
The names and yields of all the existing boreholes were tabulated.For each borehole its rank (1 or 2) Table 9 was noted and indicted as Actual Rank and its corresponding class Good or Low) noted and indicated as Actual Potential from Table 9.
From the validation map (groundwater potential zones and existing boreholes overlay) (Figure 15), the potential zone (Low or Good) in which each borehole was located was noted and indicated as predicted potential and the zone rank ( 1or 2) from Table 10 noted an indicated as predicted rank.The comprehensive data on the model and boreholes Yields, Actual and Predicted ranks/potentials were tabulated as shown in Table 10.

Graphical Comparison
The comparison between the predicted and actual potentials was demonstrated graphically by plotting the predicted rank alongside the actual rank (Figure 16).

Statistical Comparison
It can be observed from the graphical comparison that one can't make a quick conclusion of the validation.In this regard it was necessary to exude the validation statistically.The actual and predicted potential scores were expressed inter ms of low and good, and analysed (Table 11).

Groundwater Potential Zones Map
From the above analysis the predicted groundwater potential zones (model results) were validly confirmed to for the Groundwater Potential Map for Garissa Country (Figure 17).

Groundwater Potential and Existing Data
In this study, Weighted Index Overly Analysis and Analytical Hierarchal Process     Groundwater Factors The analysis indicated that lithology of the area had the biggest influence on groundwater potential, accounting for 56% of the generated groundwater potential.Soils had the second largest influence on groundwater potential accounting for 32% of the potential and vegetation accounted for only 12% of the groundwater occurrence (Table 13).
Despite the lithology in the country showing great potential for groundwater, overall, the influence of the other factors contribute to the poor groundwater potential experienced in Garissa.

Past Studies and Limitations
Numerous studies have been carried out to map groundwater potential in many regions where consistent supply of surface water is not guaranteed.Water resources managers have taken advantage of the ability to quickly create GIS models, making GIS the "go to" tool when looking at problems dealing with water management, and in particular groundwater exploration.
Few studies corroborate the findings of their model with actual data on the ground, mainly because such data is difficult to obtain or has not been generated.Where possible it is recommended that GIS modelling results are validated with ground data.

Conclusions
• It is established that provided the model is used as a pre-analysis tool.The result of the model can give useful information to planners in that whereas the map generated here does not accurately indicate sites where drilling is to be done, it accurately predicts areas where the groundwater potential is poor and drilling of boreholes should be avoided.• The model and actual borehole yields showed similar results when overlaid.About 57% of the borehole yields confirmed that the model correctly predicted the zones with good groundwater potential and 86% of boreholes confirmed zones with poor potential and an overall accuracy of 76%.
• The model clearly delineates areas with poor ground water potential where drilling of boreholes will not be used as a method of water supply and other methods of water provision should be explored thus saving time and other resources.

Recommendations
• Prospecting for groundwater in areas predicted to have good potential require caution since groundwater is not uniformly distributed underneath.
Exploration for suitable sites will require use of other available supplementary information such as yields and depths of existing boreholes to evaluate the suitability of a site before borehole drilling work commences.
• The output can be improved by improving the quantity and quality of the study validation data by carrying out field visits to confirm the GPS coordinates and yield values of the all the 37 mapped sites.Therefore a complete mapping of all the existing boreholes will ensure the use of the model in a more conclusive pre-analysis excise.
• The user of the model need to be aware that the actual results may differ from expected results since the whole process is approximation to the end and not a definite conclusion of the outcome.

Figure 1 .
Figure 1.Distribution of surface water in Garissa Country.

Figure 7 .
Figure 7. Map showing location of existing boreholes and their yields.

(
Figure 16.Predicted potential plotted against the actual potential.

Table 2 .
[4]issa Country study validation data.For this study the factors that were found to play a substantial role in influencing the occurrence of ground water in Garissa Country were landcover (vegetation), soils and lithology (rock formation).Rainfall, slope (topography) and drainage though important, were found not to play a significant role since their spatial layers are linear as compared to the others which are polygons.The groundwater factors data was obtained from ILRI website[4].

Table 7 .
Weights (indexes) of the factors.

Table 8 .
Classified and ranked borehole yields.

Table 9 .
Ranking of groundwater potential zones.

Table 10 .
The model's predicted potential and boreholes actual yields.

Table 11 .
Predicted vs actual potential analysis.
Garissa yielded no water (47.7%), with only 19% of the boreholes drilled yielding high volumes of water.About 33% of the boreholes yielded low volumes (Table12).

Table 13 .
Factors relative contribution to groundwater.