GIS technology plays a useful role in all dimensions of drainage basin management, ranging from evaluating drainage basin characteristics all the way to modelling of human activities impacts [18] . Initial steps for hydrologic modelling in performing dam site selection comprises of delineation of streams and drainage basins, followed by determination of crucial watershed properties which includes slope, length of flow, density of stream network among others.

Assessing and managing water resources for dam site selection are essentially geographical activities which require handling of spatial data of different formats. Integration of various GIS and simulation models is essential to improve knowledge in water resource management for dams.

DEM has been commonly used for extraction of drainage patterns and other properties of a drainage basin required for hydrological modelling for dam selection. As described by [19] , drainage network extraction and watershed delineation are algorithms which are developed from a DEM to derive basic topographic characteristics.

For dam site selection several Hydrological models are derived that collectively describes a catchment’s drainage patterns. These include flow direction, flow accumulation, watersheds and stream networks among others. These functionalities forms the primary layers which are then integrated with other contributing factors in a GIS environment to derive suitable dam sites.

GIS alone doesn’t have capabilities to include all decision elements related to land suitability assessment, even though it has power in spatial analysis. Instead it should be integrated with additional evaluation and assessment tools, like AHP Multi-Criteria Decision Analysis methods.

AHP Multi-Criteria Decision Analysis has functionalities and characteristics which makes it a suitable approach, as described by [20] , and these characteristics includes the capability to tackle decision which involves subjective judgments, several decision makers and most importantly, the capability to offer consistency measures of preference. AHP Multi-Criteria Decision Analysis is described as a technique which is based on three principles of analysis: binary comparison; summarizing; prioritizing and selection [2] . Within GIS, it is applied to define the weights for the selected criteria, and has the ability to deal with inconsistent judgements.

The method is centred on the construction of a sequence of Pair-wise Comparison Matrices (PCMs), which all the criteria are compared to one another. For PCM elements, [21] suggested a scale of numbers 1 to 9. These numbers Indicates the number of times more or less important an element is over the other.

Calculation of weight estimate is then done, which is used for derivation of a consistency ratio (CR) of the pair-wise comparisons. As described by [22] , PCM comprises of a consistency check in which judgment errors are identified and a consistency ratio is calculated. If CR is greater than 0.10, then some pair-wise values requires some review and the procedure is repeated until the preferred value of CR of less than 0.10 is obtained.

Overlay analysis is then done in order to get the overall summation of the weight of each contributing factor. The overlay inputs includes all the contributing factor layers standardized to a common scale in a GIS environment. Each input raster is multiplied by the specified weight. It then adds all input rasters together to obtain the final suitability map.

The study area is located in Mbeere North Sub-County and lies approximately between latitudes 0˚25' and 0˚35'S and Longitudes 37˚32' and 37˚50'E (see Figure 1). It covers approximately 514 km² stretch along Ena River which intersects major segments of three Wards in Mbeere North Sub-County: Evurore, Nthawa and Muminji.

Mbeere North is mostly characterized by hot and dry semi-arid zones due to low annual rainfall of less than 600 mm. Water shortage in the region has been a constraining factor limiting both crop production and livestock rearing hence leading to food insecurity.

According to [23] , only 5% of Embu County is under irrigation and there is great potential to expand irrigated area through water harvesting and storage in

ponds pans and dams. The success of this effort is guided by finding a suitable site for the dam construction.

The study used AHP Multi-Criteria Decision Analysis integrated in a GIS environment to determine suitable sites for an earth dam. The methodology was implemented in 4 phases. Phase one involved planning and other necessary logistics prior to project implementation. In phase two all necessary data was collected based on the number of criteria determined. Appropriate attributes for each criterion were established in phase three, and influence of each criterion was also determined using AHP Multi-Criteria Decision Analysis. In the final phase, the criteria organised as layers were combined by performing a weighted overlay analysis to produce a suitability map.

This study involved use of various datasets. The data required was governed by a number of criteria that affect dam site selection which included Slope (Topography), Geology, Soil type, Catchment size, Land Cover, Proximity to River Centrelines and Proximity to roads.

The DEM data was obtained from the Advanced Space-borne Thermal Emission and Reflection Radiometer (ASTER) GDEM, with pixel size of 1 arc-second (approximately 30 m). It was processed and used to generate elevation (DTM) and slope as indicated in Figure 2.

The processed DTM was used for delineation of watershed and generation of stream network according to the process flow in Figure 3.

The geology of the study area was obtained from Kenya Department of Mines and is dominated by granitoid gneisses, undifferentiated predominantly irregularly banded magmatic gneisses, kenyte lava, agglomerate and tuff rocks [24] . The permeability of gneisses and granite is originally very low but orogeny-related deep-seated fracture or shear zone in these rocks serves as good pass way or reservoir of ground water, therefore making the geology of the area suitable for dam construction [25] (see Figure 4).

The soil in the study area was obtained from ILRI GIS services website [26] , consisting of four different classes i.e. Clayey, loamy, sandy and very clayey as illustrated in Figure 5.

The land cover data was sourced from Regional Centre for Mapping of Resources for Development (RCMRD). This data was already classified into six different types of land cover i.e. Dense Forest, Wooded Grassland, Open Grassland, Open Water, Perennial Cropland and Annual Cropland as illustrated in Figure 6.

The data for river and road centrelines is as shown in Figure 7. The dominant river across Embu County, Ena River was derived from the DTM while the road network was obtained from the topographical maps.

Each factor was assigned a weight which indicated their significance. In this study AHP pair-wise comparison method was used.

Several online programs are available for calculating AHP priority weights which includes Microsoft Excel, BPMSG AHP Online System [27] among others. The program used for this study was Microsoft Excel as described by [28] .

The analysis utilizes seven different map layers as inputs in the suitability; Slope, Geology, Soil type, Catchment size, Land Cover, Proximity to river and roads. Each of these was converted into raster format.

The problem was decomposed into a level of hierarchy consisting crucial elements of the decision problem. The first layer indicates the goal of decisions making, the second layer containing the criterion to be used and the third level showing the decisions making attributes as shown in Figure 8.

An assessment of the relative importance of each criterion with one another within a scale of 1 to 9 was used. This allows for the equitable comparison of intensities of suitability, with 1 being factors of equal preference and 9 being the factors with extreme preference over the other as shown in Table 2. In this process, the main assumption is that if first factor has one of the above non-zero numbers assigned to it when compared with the second factor, then the second factor is assigned the reciprocal of the first e.g. if slope compared to land cover has a factor of 9, then the land cover will have a factor of 1/9.

From previous studies on dam site selection, incorporated with the experts’ opinion on factors affecting dam site selection, the order of importance of each criterion is as indicated in Table 3.

To determine the weight of each factor, Pair-wise Comparison was used as illustrated in Table 4. A matrix was constructed in which relative to its importance, a criterion was compared with the other on a scale of 1 to 9.

The total was the summation of each column:

Slope = 1 + 1 2 + 1 5 + 1 3 + 1 7 + 1 9 = 2 2 5

Geology = 2 + 1 + 1 2 + 1 7 + 1 9 + 1 9 = 4

Soil type = 5 + 2 + 1 + 1 2 + 1 3 + 1 7 + 1 5 = 9 1 6

Catchment Size = 3 + 7 + 2 + 1 + 1 3 + 1 5 + 1 5 = 13 3 4

Land Cover = 7 + 7 + 3 + 3 + 1 + 1 2 + 1 2 = 22

Proximity to River = 9 + 9 + 5 + 5 + 2 + 1 + 1 2 = 31 1 2

Proximity to Roads = 9 + 9 + 7 + 5 + 2 + 2 + 1 = 35

while the intensity of importance is allocated to criterion i when compared to criterion j, the reciprocal value is assigned to criterion j as intensity of importance. For example, from the above matrix, i (slope) = 9 while j (Proximity to Road) = 1/9. After comparison between all possible criteria pairs is complete, the Weight (W) of criteria i is calculated using Equation (1). See Table 5 for the calculated weight (W).

W i = ∑ j = 1 n P i j / ( ∑ i = 1 n ∑ j = 1 n P i j ) (1)

Equation (1): Calculation of weight (source (Dai, 2016))

where:

Pij = Relative importance in pair-wise comparison of criterion i compared to criterion j

n = Number of factors

i & j = Criterion

W = Priority Weight

The values used for pair-wise comparison usually rely on subjective judgement which could lead to arbitrary results with bias. To evaluate the consistency of pair-wise comparison matrix, a numerical index called Consistent Ratio (CR) is used as indicated in Equation (2) below. CR shows the ratio of the Consistency Index (CI) to the average consistency index, known as Radom Index (RI) as shown in equation 3.

C R = C I R I (2)

Equation (2): Calculation of Consistency Ratio

The calculation of CI is given as:

C I = ( λ max – n ) / n – 1 (3)

Equation (3): Calculation of Consistency Index

where:

n is the number of factors = 7

λ_max is the Principal Eigen Value

λ_max = Σ of the products between each element of the priority vector and column totals

λ_max = (2 2/5 × 0.3629) + (4 × 0.2924) + (9 1/6 × 0.1368) + (13 3/4 × 0.1057) + (22 × 0.0460) + (31 1/2 × 0.0308) + (35 × 0.0254) = 7.6201

CI = (7.6201 − 7)/7 − 1

= 0.6201/6

= 0.1033

The RI used depends on the number of criteria. This study has seven criteria hence the RI used was 1.32 as indicated in Table 6.

CR = CI/RI

CR = 0.1033/1.32

= 0.07

According to [31] , the value of CR is compared to 0.1 which is the maximum CR value for an acceptable pair-wise comparison. The resulting CR for this analysis is 0.07 which is less than the acceptable maximum CR value recommended in AHP and therefore this consistency is acceptable.

Source: [31] .

Values must be prioritized even within a single raster. This is because values in a particular raster may be fit for your purpose while others may be undesirable [32] . For example a slope of 0% - 9% is ideal for dam site selection. This was done through a process of reclassification in ArcGIS.

To perform the reclassification, each raster dataset was reclassified into a common scale of 1 to 5. With 5 being more favourable hence has the highest influence for dam site selection and 1 with the lowest influence.

Classification of geological factors mainly depended on the permeability, thickness and strength of rocks foundations at the area of interest [33] . Stronger foundation was given the highest scale of 5 while weaker was given the lowest scale of 1.

Dam site construction requires topography that is well-drained and gently sloping. This is best as it minimizes construction costs as described by [34] . Gentle slope was given the highest scale of 5 steep slope was given the lowest scale of 1.

The choice of the dam site should provide a large area for water storage and this is determined by the size of the water catchment areas around the selected area. Catchment size determines what water is available within a catchment [30] . A large catchment was given the highest scale of 5 while a small catchment was given the lowest scale of 1.

According to [35] , the land covered by plant biomass is more resistant to both wind and water soil erosion and therefore usually experience moderately little erosion. In this regard, Wooded Grassland was given a highest scale of 5.

The ranking of soil was done based on the soil infiltration rate. As described by [29] , the smaller the soil particles, the slower the water infiltration rate. To ensure minimum loss of water through seepage soils must be impermeable, with more than 20% clay [34] . Therefore clayey soil was given the highest scale of 5 and the sandy soil was given the lowest scale of 1.

The closer the dam to the river, the better for economical purposes. The closest distance of 500 m buffer from the river centreline was given the highest scale of 5 while distances beyond 3 kilometres were given a lower scale of 1.

The near the dam to transport networks, the better the accessibility [2] . Therefore short distances were given the highest scale of 5 and longer distances were given the lowers scale of 1. Tables 7(a)-(g) show the corresponding standardized layers.