Introduction

Increasing population and associatively growing demand for food and other agricultural commodities have caused an intensification and extensification of the agricultural sector witnessed in the last decade (Lambin and Meyfroidt 2011; Tscharntke et al. 2012; Rudel et al. 2009). As an agriculture dominated basin in Ethiopia, the Upper Blue Nile seems to be experiencing similar pleasures (Gebrehiwot et al. 2014; Bewket and Sterk 2002). However, the amount, location and degree of suitability of the basin for agriculture do not seem well studied and/or documented. Haphazard land-use has thus far resulted in continuing deforestation, exhausted soil fertility, increased soil erosion and land degradation especially in the basin’s highland catchments (Zeleke and Hurni 2001; Bewket 2002; Awulachew et al. 2010). Land suitability analysis can help establish strategies to increase agricultural productivity (Pramanik 2016) by identifying inherent and potential capabilities of land for intended objectives (Bandyopadhyay et al. 2009). It can also help identify priority areas for potential management and/or policy interventions through land and/or soil restoration programs, for instance.

The analytical hierarchy process (AHP) (Saaty 1980) technique integrated with GIS application environments has been used for agricultural land suitability analysis on various case study sites around the world (Zabihi et al. 2015; Akıncı et al. 2013; Zolekar and Bhagat 2015; Pramanik 2016; Malczewski 2004). It involves pair-wise and weighted multi-criteria analysis on a number of selected socio-economic and biophysical drivers. The technique has extensively been used for land suitability analysis at local and region levels for watershed planning (Steiner et al. 2000), vegetation (Zolekar and Bhagat 2015) and agriculture (Shalaby et al. 2006; Bandyopadhyay et al. 2009; Akıncı et al. 2013; Motuma et al. 2016). Biophysical parameters such as land cover, slope, elevation, and soil properties such as depth, moisture, texture and group are frequently used for assessment of land suitability evaluation (Rossiter 1996; Zolekar and Bhagat 2015). ‘Expert opinion’ is used for weighting such factors in influencing land suitability through pair-wise comparison in AHP.

In this study, we analyzed agricultural land-use suitability in the Abbay basin using AHP and GIS based weighted overlay analysis (WOA) techniques. Multiple criteria for agricultural land-use suitability mapping were derived based on literature reviews, field investigations and following FAO guidelines for agricultural land-use evaluation (Rossiter 1996; Zolekar and Bhagat 2015; Zabihi et al. 2015; Bandyopadhyay et al. 2009). Identification and mapping of agricultural land suitability is especially important in the basin given the following considerations: (1) the pressing need to increase agricultural productivity to meet growing food demands, (2) the growing risks of increased rainfall variability due to climate change in already water limited agricultural systems, and (3) the growing interest by local and regional policy and management bodies for evaluation of land capability for various land-use alternatives.

Located on the western half part of Ethiopia (Fig. 1), Abbay is the most important river basin in Ethiopia by most criteria. It accounts for about 20 % of the nation’s land area; 50 % of its total average annual runoff; 25 % of its population; and over 40 % of its agricultural production (EEPCo 2014). Divided into 16 major catchments (Fig. 1), the basin covers an area of 199,812 km2 with significant contribution of runoff and fertile highland soil to Sudan and Egypt in the downstream. Biophysical and anthropogenic factors combined with torrential runoffs in the rugged highlands in the basin are causing considerable land degradation and soil erosion in the upstream, whereas deforestation is a major concern in the mid/low lands of the basin (Hurni et al. 2005; Gebrehiwot et al. 2014). Torrential runoff washing off the rugged terrains, poor land-use management has been reported as a cause for gully formation, soil erosion and land degradation threaten the livelihood of subsistence farmers (Teferi et al. 2012; Yalew et al. 2012; Tekleab et al. 2014).

Fig. 1
figure 1

Location map of the Abbay basin and its catchments in Ethiopia

Full size image

Materials and methods

Using literatures and guidelines on land evaluation for agriculture (Bojórquez-Tapia et al. 2001; Prakash 2003; Bandyopadhyay et al. 2009; Wang 1994; Olaniyi et al. 2015; Zhang et al. 2015; Rossiter 1996) we identified nine important criteria that determine agricultural land suitability in the basin: soil type, soil depth, soil water content, soil stoniness, slope, elevation and proximity to towns, roads and water sources (Table 1). GIS raster datasets on each of these indicators were gathered and processed from several sources for the study areas. According to FAO guidelines (Rossiter 1996), land suitability for agriculture can be classified into five categories: (1) highly suitable, (2) moderately suitable, (3) marginally suitable, (4) currently unsuitable and (5) permanently unsuitable. In this study, we customized and reclassified each raster criteria layer into four categories with associated suitability score of 1–4 (4 = highly suitable; 3 = moderately suitable; 2 = marginally suitable; and 1 = unsuitable). The ‘unsuitable’ category represents the ‘permanently unsuitable’ category of FAO. Similar to what is defined as ‘currently unsuitable’ in the FAO method, we excluded forest and protected areas from the suitability computation altogether assuming that such land may not be used (and hence ‘unavailable’) for agriculture in favor of other ecological services (biodiversity conservation). Weights for each of the selected criterion were calculated using the AHP technique. After the weight of each raster dataset was computed, a GIS based WOA was carried out to establish a suitability map. The process diagram of the method used is shown in Fig. 2.

Table 1 Data and data sources
Full size table
Fig. 2
figure 2

Process diagram of the methods used in the stud

Full size image

Generation of criteria maps

Slope and elevation

Slope and elevation data layers (Fig. 3a, b), respectively) were generated from the Shuttle Radar Topography Mission (SRTM) digital elevation model (DEM) data of 30 m resolution available on the Google Earth Engine archive (Moore and Hansen 2011; Gorelick 2013). Based on FAO manual for agricultural watershed management (Sheng 1990), agricultural suitability of different slope classes for the study area are defined as in Table 2. However, since no specific crop suitability is assumed, elevation value lower than 3700 m a.s.l. is taken to be suitable for agriculture. Elevation above 3,700 m is classified as ‘high wurch’ (frosty-alpine) and thus unsuitable for agricultural purposes according to the agro-ecological zoning of Ethiopia (FAO 2003).

Fig. 3
figure 3

a Slope and b elevation maps of the Abbay basin

Full size image
Table 2 Slope classes for agricultural suitability
Full size table

Soil properties

Soil characteristics are one of the most important factors in agricultural land-use assessment (Dominati et al. 2016; Bonfante and Bouma 2015; Juhos et al. 2016). In this study, soil depth, soil water content, soil type and soil stoniness are taken as indicators to assess general soil suitability for agriculture. Soil depth and averaged soil water content maps are shown in Fig. 4. Soil type and soil stoniness are shown in Fig. 5. The soil properties used here were standardized for land suitability assessment as shown in Table 3. These soil characteristics were categorized based on the soil classification and characterization guide for agricultural suitability by FAO (Sheng 1990), and other guidelines for common biophysical criteria for defining natural constraints for agriculture (Van Orshoven et al. 2012).

Fig. 4
figure 4

a Soil depth and b soil water content in the Abbay basin

Full size image
Fig. 5
figure 5

a Soil types and b soil stoniness in the Abbay basin

Full size image
Table 3 Soil characteristics and suitability for agriculture
Full size table

Soil stoniness refers to percentage of gravel/stone content within the top 90 cm soil depth. Soil groups were classified based on their suitability and limitations for agriculture as outlined by FAO and the international livestock research institute (LRI) (1992). Based on the guides, the major soils in the study area are classified as:

  • Soils with very high potential: nitisols (NS), luvisols (LS), cambisols (CS), phaeozems (PS).

  • Soils with few limitations for agriculture: vertisols (VS), alisols (AS).

  • Soils with major limitations (low production potential, rocky terrain soils, poorly drained soils): histosols (HS), liptosols (LpS).

Soil water content dataset was derived from the spatially distributed soil–water balance model by Trabucco and Zomer (2010). In their model, Trabucco and Zomer simulated a soil–water balance model for the years from 1950 to 2000 as a height of water (in mm) per month (m) using Eq. (1):

$$\Delta SWC_{m} = {EPr}_{ec} - \Delta AET_{m} - R_{m}$$
(1)

where, \(\Delta SWC_{m}\) is the change in soil water content, \(EPr_{ec}\) is the effective precipitation, \(\Delta {\text{AE}}T_{m}\) is the actual evapotranspiration, and \(R_{m}\) is the runoff component which includes both surface runoff and subsurface drainage. Furthermore, \(SWC_{m}\) may not exceed \(SWC_{max}\), which is the total maximum soil water content (SWC) available in the soil for evapotranspiration. \(SWC_{max}\) was assumed by the modelers at a fixed spatial value of 350 mm, which corresponds to average soil texture for a plant rooting depth of 2 m. The soil water content was then computed as a linear percentage function of actual and potential (maximum) soil water content over the months and the years from 1950 to 2000 as shown with Eq. (2):

$${\text{Ksoil}}_{m} = \sum\nolimits_{k = 1}^{y} {\left( {\sum\nolimits_{k = 1}^{12} {\left( {{{SWC_{m} } \mathord{\left/ {\vphantom {{SWC_{m} } {SWC_{\hbox{max} } }}} \right. \kern-0pt} {SWC_{\hbox{max} } }}} \right)} } \right)}$$
(2)

Where, \({\text{Ksoil}}_{m}\) is percentage of average soil water content, \(SWC_{m}\) is actual soil water content in month (m), \(SWC_{ \hbox{max} }\) is the maximum (potential) soil water content, and y is year.

Proximity to water, road and towns

Spatial proximities to water sources, road and towns (Fig. 6a–c, respectively) were computed using spatial overlay of respective GIS layers. Influences of distance parameters on agricultural land suitability, Table 4, were estimated based on literature and field observation (Wale et al. 2013; Bizuwerk et al. 2005).

Fig. 6
figure 6

Distances to a town, b water sources and c road in the Abbay basin

Full size image
Table 4 Proximity influences on agricultural land suitability
Full size table

In addition to the criteria inputs for agricultural land suitability assessment discussed thus far, data on land cover and protected sites where collected for overlay analysis to serve as constraint layers on the final suitability map. Land cover map (Fig. 7a) was derived and reclassified from GlobCover2009. GlobCover2009 is a global land cover map based on ENVISAT’s Medium Resolution Imaging Spectrometer (MERIS) Level 1B data acquired in full resolution mode with a spatial resolution of 300 m (Bontemps et al. 2011). Map on protected sites (Fig. 7b) which includes areas such as national parks and reserve sites was derived from the ‘Protected Sites’ global dataset of UNEP’s World Conservation Monitoring Centre (UNEP-WCMC 2012).

Fig. 7
figure 7

a Reclassified land cover and b protected areas in the Abbay basin

Full size image

We assumed that forest, protected areas and water bodies as unavailable (and hence ‘currently unsuitable’) for agriculture. Changes in policy or management could easily change the suitability of these layers. A forest may, for instance, be deforested for large scale agriculture and thus changing its land suitability. These layers (Fig. 8) are therefore used as constraints that are superimposed on top of the computed suitability map.

Fig. 8
figure 8

Constraint layers. Note that constraint layers are not included in the suitability categorization but instead superimposed on the computed suitability map

Full size image

Standardization of criteria maps

The selected criteria maps are initially in different units. For executing WOA for land suitability, the criteria maps need to be converted into a similar scale through standardization techniques. Standardization techniques convert the measurements in each criteria map into uniform measurement scale so that the resulting maps lose their dimension along with their measurement unit (Reshmidevi et al. 2009; Zabihi et al. 2015). For standardization, all the criteria vector maps were converted to raster data formats. The raster maps were then reclassified using the Spatial Analyst tool in ArcMap into four comparative categories as discussed earlier: highly suitable, moderately suitable, marginally suitable and unsuitable. Once all the criteria maps are standardized, weights of each criteria map can be calculated using AHP. Then WOA method will be applied to produce the final suitability map.

Calculation of weight for criteria maps

The analytic hierarchy process (AHP) is used to calculate weights for the criteria maps. It is a structured method for analyzing complex decisions by breaking them into pair-wise alternatives of two at a time (Saaty 1988, 2008). It involves sub-dividing big and intangible decision problems into minute sub-problems amenable for pair-wise comparison (Saaty 1987). An AHP plugin tool for the ArcGIS environment (Marinoni 2009) was used to compute weights for the different criteria layers. Using the pair-wise comparison matrix, the analytic hierarchy process calculates comparative weights for individual criterion layers. It also produces consistency ratio (CR) that serves as a measure of logical inconsistency of expert/user judgments during pair-wise criteria comparisons, measured using Eq. (3).

$${\text{CR}} = \frac{CI}{RI}$$
(3)

where, \(CI\) represents consistency index, and \(RR\) represents random index.

The CR measurement facilitates identification of potential errors and thus judgment improvements depend on these values. According to Saaty (1988), if the CR value is much in excess of 0.1, the judgments during pair-wise comparison are untrustworthy because they are too close for randomness. Saaty (1988) provided a ‘fundamental scale’ for computing pair-wise comparison matrix of the criteria layers while performing an AHP (Table 5). This involves a construction of a matrix where each criterion is compared with the other criteria, relative to its importance, on a scale from 1 to 9. Scale 1 indicates equal preference between a pair of criteria layers whereas 9 indicates a particular criteria layer is extremely favored over the other during expert judgment (Saaty 1988; Malczewski 2004).

Table 5 The fundamental scale for pair-wise comparison matrix (Saaty 1980)
Full size table

After determining the relative importance of each criteria layer, through pair-wise comparison matrix, these values are entered on an ArcGIS based AHP tool to produce associated weights and CR value. Table 6 shows inputs for the pair-wise comparison inputs for the AHP to compute weights for the criteria layers. The weights produced from the AHP procedure using inputs in Table 6 range between 0 and 1, where 0 denotes the least important and 1 the most important criteria determination of land suitability.

Table 6 Pair-wise comparison matrix for multi-criteria decision layers
Full size table

Table 7 shows various indices produced from the GIS based AHP tool. The consistency ratio (CR) value is a function of CI and RI as defined in Eq. 3.

Table 7 Indices computed using the GIS based AHP tool table
Full size table

WOA

After computation of weights for each raster layer using AHP, weighted overlay analysis (WOA) is performed on an ArcGIS environment. Weighted overlay is an intersection of standardized and differently weighted layers during suitability analysis (Zolekar and Bhagat 2015). The weights quantify the relative importance of the suitability criteria considered. The suitability scores assigned for the sub-criteria within each criteria layer were multiplied with the weights assigned for each criterion to calculate the final suitability map using the WOA technique (see Eq. 4).

$$S = \sum\nolimits_{i = 1}^{n} {W_{i} X_{i} }$$
(4)

where \(S\) is the total suitability score, \(W_{i}\) is the weight of the selected suitability criteria layer, \(X_{\text{i}}\) is the assigned sub-criteria score of suitability criteria layer i, and n is the total number of suitability criteria layer (Cengiz and Akbulak 2009; Pramanik 2016).

Results

The weighted overlay analysis carried out using the criteria layers with their respective weights generated a combined suitability map (Fig. 9). Forest, protected area and water bodies were then superimposed on this suitability map to determine the final suitability map (Fig. 10). According to this map, it was determined that 28.6 % (57,050 km2) of the study area is highly suitable, 48.9 % (97,812 km2) is moderately suitable, and 6.2 % (12,378 km2) is marginally suitable. About 6 % (11,978 km2) is determined to be ‘unsuitable’ whereas the rest 10.3 % (20,594 km2) is determined unavailable (or currently unsuitable) categories (see Table 8).

Fig. 9
figure 9

Agricultural land suitability in the Abbay basin excluding constraint layers

Full size image
Fig. 10
figure 10

Agricultural land suitability in the Abbay basin with constraint layers. The constraint layers include water bodies, forest and protected areas which were assumed to be fixed in this study

Full size image
Table 8 Summary of agricultural land suitability map of the Abbay basin
Full size table

Water bodies, forest cover and protected areas are treated as ‘unavailable’ or are constraints for the suitability analysis and are instead superimposed on the final suitability map (Fig. 10).

Table 9 presents details of area and percentage coverage of the different suitability categories per each catchment in the Abbay basin.

Table 9 Summary of land suitability level for agriculture cca catchment area [km2 (%)]
Full size table

Discussion and recommendation

A closer look at the percentage coverage of the suitable lands with the catchments indicates very high variation in between the different catchments of the basin (see Table 9). About 50 % of the Tana catchment (North), for instance, is classified as ‘highly suitable’ whereas as low as only 12 % is classified in the same category in Beshilo (North-East). Similarly, there is a large variation in percentage coverage of ‘moderately suitable’ lands per catchment which ranges from 70 % in the Dinder catchment (North-West) to about 24 % in the Tana catchment. A much lower variation in percentage area between catchments is seen when considering the sum of ‘highly suitable’ and ‘moderately suitable’ categories (max = 86 %; min = 57 %; mean = 77 %; SD = 7) compared with the sum of ‘marginally suitable’ and ‘unsuitable’ lands (max = 41 %; min = 3 %; mean = 12; SD = 10).

What is generally noticeable is that the North, North–West, South and South–West catchments of the basin seem to have larger percentage area for ‘highly suitable’ and ‘moderately suitable’ land for agriculture. On the other hand, the Western, North-Western and Central highlands of the basin seem to have higher coverage of ‘marginally suitable and ‘unsuitable’ lands for agriculture. Looking at some of the main factors weighing into the AHP analysis such as slope, soil water content and soil stoniness, it is easy to see that the North-Western and central highlands are dominated by steep slope ranges (25–80°, Fig. 3a), low percentage of soil water content (22 %, Fig. 4b) and high level of soil stoniness (75 %, Fig. 5b). This part of the basin is also located on a relatively higher elevation range (3000–4239 m a. s. l.) than the South and South–West part of the basin. The combinations of steep slopes and higher elevation may imply a high susceptibility for land degradation and soil erosion, among other things, in the catchments in this part of the basin resulting in higher percentage of stony upper soil. It is advised that land and water managers and policy makers in the basin prioritize such areas during land and/or soil restoration efforts.