1 Background

Land use land cover (LULC) change is affected worldwide ecological processes, which results in major environmental challenges of global importance [14, 19]. It were estimated that the human footprint has affected 83% of the global terrestrial land surface. About 60% of the ecosystem services in the past 50 years alone were degraded by LULC change [15]. LULC change is arguably the most pervasive socioeconomic force driving change and degradation of ecosystem. Deforestation, urban development, agriculture and other human activities have substantially altered the Earth’s landscape [23]. Nowadays, agriculture occupies 38% of the globe’s ice-free terrestrial surface and is the largest land cover type by area [10]. The rate of agricultural expansion are decreasing globally. The impact is expected to be high in developing countries like Ethiopia [1].

1.1 Statement of the problems

While globally important, LULC change occurs locally [22]. LULC change is one of environmental challenges in Ethiopia [7, 18, 21].

Land use practices and topography determine the land cover change extent of the area [11]. For example, agricultural practices in the western and northern part of Ethiopia are different. Historical evidence shows that northern Ethiopia agriculture depends on slash and burning, whereas western part is depended on mixed farming system [9]. Such difference in farming practices significantly changes the land use land cover rates from place to place. Agricultural expansion is happening in the study area due to rapid population growth and usually it is encroaching forest, grass lands and other land use types. Because of this soil erosion, land degradation, loss of productivity and climate variability are exacerbated in the upper watershed areas. However, in the lower watershed areas drinking and irrigation water scarcity, floods, drought and loss of productivity is encountered. LULC change analysis, nowadays is the basic method to understand the dynamics of the land use/ land cover status [5, 12]. The main goal of this study was to evaluate LULC change dynamics of the land use classes and vegetation cover changes starting from 1984 to 2015 using multi-temporal satellite data, in order to monitor the changes, so as to understand the extent of the problem that the study area is facing. The study is expected to pave the way for informed decision-making to implement land restoration and watershed conservation interventions.

2 Materials and methods

2.1 Methodological framework

The methodological framework included in this study was preprocessing and post-processing activities. The preprocessing activities included the land sat data preparation and watershed delineation. The cloud-free land sat images were downloaded from the United States Geological Survey (USGS). Atmospheric correction was performed in this process using Earth Resources Data Analysis System (ERDAS) software 2015 version. The post-processing activities (such as classification and accuracy assessment) were done to identify land use class. The ground truth verification points were automatically selected by using the random sampling method for accuracy assessment of the classification result. Based on this 256 points were verified by using Google Earth to investigate historical land sat images. The intensity analysis was performed by using analyzing error matrix at interval, category and transition levels.

2.2 Description of the study area

The Jewha watershed is characterized by diverse topographic conditions. It is located in the headwaters of the Awash River, which is Ethiopia’s major inland drainage system. The watershed elevation of the study area ranges from 1200 to 3200 m above mean sea level. The topography is mountainous with dissected terrain and steep slopes characterized in most parts by scarce resource in the sub-watershed due to high population pressure and land degradation.

The study area is located in northeastern part of Ethiopia. It has an average annual rainfall of 1007 mm, with short rain, March–April and long rain June–September. The mean annual minimum and maximum temperatures are 16.5 and 31 °C. The community relies on subsistence agricultural system. The region has gone through substantial changes in settlement and land use over the years. However, research has not been conducted in the area, in regard to LULC.

The watershed is in the rift valley escarpments of Eastern Amhara region, which encompasses the lowlands of North Shewa zone. Based on the 2007 national census conducted by the Central Statistical Agency of Ethiopia (CSA), Kewot District has a total population of 118,381, an increase of 9.97% over the 1994 census. About 29,058 households were counted in this district, resulting in an average of 4.07 persons to a household, and 28,104 housing units (Fig. 1).

Fig. 1
figure 1

Map of the study area

Full size image

2.3 Data source

The input data for land sat image were obtained from USGS. Cloud-free images of years 1984, 1995, 2005 and 2015 were used. The digital elevation model (DEM) with the resolution of 30 m used from the same sources. Observational data for ground truth verification were collected from field observation in addition to Google Earth map.

2.4 Methodology

2.4.1 Land use/land cover (LULC) classification

Landsat images with path (168) and row (53) were obtained from the USGS (see Table 1). Cloud-free images in the years 1984, 1995, 2005 and 2015 were taken for LULC classification.

Table 1 Types of Landsat used in the study
Full size table

The LULC map for ‘Jewha’ watershed was delineated using Soil and Water Assessment Tool (SWAT) software. An intensive preprocessing such as geo-referencing, layer stacking was carried out in order to Orto-rectify the satellite images [17]. The classification then was processed using ERDAS Imagine 2015 software after the satellite image bands correction was carried out using radiometric algorithm. Then, from the stacked satellite image the study area subset was extracted. Further, the gap in Landsat 7 image was filled by using the interpolation method.

2.4.2 Accuracy assessment of classification

To check the classification accuracy, assessment of error matrix was done for all images of the study periods. According to [12], maximum likelihood algorithm produced acceptable LULC classification. Each of the land use/cover class of the study area was compared to the reference points to assess the accuracy of the classification. For this purpose, 256 reference points were collected (Fig. 2). The reference data were prepared by using distribution parameters of equalized random points from the topographic map and verified by using historical data of Google Earth [3].

Fig. 2
figure 2

Reference points used for accuracy assessment

Full size image

2.4.3 Intensity analysis

Intensity analysis is a mathematical framework that compares a uniform intensity to observed intensities of temporal changes among categories [16]. In this study intensity analysis used at three levels: interval, category and transition.

Interval level shows the results of change area between two periods. The study periods have three intervals: these are 1984–1995, 1995–2005 and 2005–2015. The duration of the intervals was 11, 10 and 10 years respectively. The temporal intensity change for each time interval was calculated as stated by [16].

At the category level, each category to measure how the size and intensity of both gross losses and gross gains varies across space were examined. After the intensity analysis of gross losses and gross gains, calculated for each category, the observed intensities and uniform intensity of annual change were compared.

At the transition level, the size and intensity of the transition variation among categories were examined. When a category losses or gains, this level of analysis can identify which other categories are intensively avoided versus targeted for transition by comparing the observed intensity of each transition with a uniform intensity that would exist if the transition were distributed uniformly among the categories available for the transition [4].

The land cover class defined as follow (Table 2) in the study area. The researcher verified the land cover classes on ground by using observation and historical images of Google Earth.

Table 2 Land cover classification scheme
Full size table

3 Result and discussion

3.1 Land use/land cover change analysis and intensity analysis

The result shows that there is significant land cover change from the forest to cultivated land for the years 1984–1995 but the change is not significant from the year 2005–2015 (Table 6). The forest cover status of ‘Yelen’ forest, which found in the study watershed, was analyzed by I-tree canopy and Google Earth software. This natural forest is decreased from 268 ha in 1987 to 168 ha in 2015 (Fig. 4). From field observation it was understood that this was due to human interference and livestock population growth (Figs. 3, 4).

Fig. 3
figure 3

Classified images of Jewha watershed for the years 1984, 1995, 2005 and 2015

Full size image
Fig. 4
figure 4

‘Yelen’ forest cover trend from 1987 to 2016 in the watershed

Full size image

Workflow of multi-temporal land use/land cover change detection proved to be efficient. The accuracy assessment done from intensity analysis at interval, category and transition levels presented as follow:

3.1.1 Interval level

From the figure below, the transition matrix 1984–1995 in percent of landscape change occurs on 4.36%. The transition matrix in the intervals 1995–2005 and 2005–2015 are 4.60 and 5.21%, respectively. The annual change is fast in the interval of 2005–2015. The change was slow between 1984–1995 and 1995–2005 (Fig. 5).

Fig. 5
figure 5

Calculation of hypothetical errors of interval level

Full size image

3.1.2 Category level

Categorical change in the interval of 1995–2005 (Fig. 6) is active in grass land, shrub land, natural forest plantation forest and bare land which are above uniform intensity of 4.60. Plantation forest emerged due to the intervention in the highlands of the watershed.

Fig. 6
figure 6

Categorical level annual change area and annual change intensity on 1995–2005

Full size image

Changes of 2005–2015 interval active (Fig. 7) in categories of grass land, shrub land, cultivated land, natural forest and bare land which are above uniform intensity of 5.21. Natural forest and bare land are highly active in this case. From observation the bare land is increased due to loss of productivity in the sloppy area and gully formation in the lowland transition from cultivated land. Plantation forest is dormant during this time and it may be due to the protection it receives.

Fig. 7
figure 7

Annual change area and annual change intensity on 2005–2015

Full size image

3.1.3 Land use/cover change detection at transition level

Cross-tabulation is a means to determine amounts of conversions from a particular land cover to the other land cover categories at later date. The LULC change detection of each consecutive year was calculated, that is, the years between 1984–1995, 1995–2005 and 2005–2015. In the table below, the column indicates initial year (1984) and the row indicate the final year (1995).

The transition matrix (Table 3) shows that the highest probability of change (7049 ha), grass land were changed to cultivated land and the second were cultivated land that were changed to shrub land, this may be due to loss of productivity in the sloppy area, whereas the lowest probability of change were that natural forest to bare land. In this period, there was no plantation forest detected, this is because the plantation activities were started only around 1977 E.C. Finally, agriculture has led to rapid land use change contributed to increase of cultivated land from 9046 in 1984 to 12,042 ha in 1995, whereas forest land, which has high environmental protection value, decreased from 1519 ha in 1984 to 1185 ha in 1995.

Table 3 LULC change detection between 1984 and 1995 in hectare (ha)
Full size table

In the years between 1995 and 2005 (Table 4) the change of probability were high from shrub land to cultivated land (6680 ha) the lowest were from bare land to plantation forest. In this case, cultivated land has decreased from 25,769 in 1995 to 25,224 ha in 2005. This may be because of loss of productivity that 4789 and 3745 ha of cultivated land changed to grass and shrub lands, respectively.

Table 4 LULC change detection between 1995 and 2005 in hectare (ha)
Full size table

Table 5 indicates transition matrix which represented the replacement of one type of land use to another. In the years between 2005 and 2015, the highest probabilities of change, at (4581 ha), are that shrub land was changed to cultivated land, while the lowest, zero probabilities were presented from bare land to plantation forest, this is because of most of bare land are rivers bed. Shrub land has been replaced by other land uses such as grass land (4496 ha) and natural forests (620 ha). Natural forest was replaced by shrub and cultivated lands. This may be due to forest degradation and agricultural expansion. The results of the study are in conformity with researches undertaken in highland watershed as seen in [8, 20].

Table 5 LULC change detection between 2005 and 2015 in hectare (ha)
Full size table

The table below (Table 6) shows the cumulative change of each land use in the specified year; those were natural forest and shrub land which decreased from 1518 and 6638 ha in 1984 to 975 and 4219 ha in 2015, respectively. Cultivated land and bare land show increasing manner. This led to high land degradation in the watershed and gulley formation in the riverbeds.

Table 6 Summary of area statistics for the maps of 1984, 1995, 2005 and 2015
Full size table

In the transition between low and mid-land, grass land is changed to shrub land in 2015 (Fig. 3). This may be due to recovering process [6, 13], since the area is not suitable due to steepness of the slope for grazing vegetation to grow easily. However, the total percentage of grassland is increased. In the upper watershed, the cultivated land is changed to grass land, this may be due to loss of productivity in the sloppy area. According to Agidew and Singh [2], a mountainous and highly dissected terrain with steep slopes will be susceptible for land degradation due to high population pressure. Bare land in the second and third study periods decreased; this is may be due to restoration to grass and shrub land. However, it increased in 2015 because of gully formation in the rivers bed.

4 Conclusions and recommendations

The study clearly demonstrates that there is visible land use/land cover change over the last twenty years in the study area. This change has adverse environmental implications, which would be exacerbated if the same trend continues. The extent of the cultivated land had increased by 12% between 1984 and 2015. The results have clearly shown that this change has come at the expense of the vegetation cover which has decreased by 11.3% during the same period. The cultivated land had increased from 38.7% in 1984 to 51.4% in 2015. Even if the natural forest is increased in 2015 as it compared with 2005, natural forest decreased when it compared with 1984. And shrub land had decreased throughout the study periods.

The natural forest of ‘Yelen’ decreased not only by area coverage but also by the canopy cover as observed in the field as well as from i-tree canopy. This shows that the quality of forest being degraded and total change to other land use class. Furthermore, the depletion of the vegetation cover as well as the increment of bare land from 1.8% in 1984 to 2.7% in 2015 has immense impact on the area resulting in gully formation in the riverbeds as observed on the ground from field visits. The trend of the whole study periods shows there is degradation of forest and soil. The increment of bare land was from 1.8% in 1984 to 2.7% in 2015, from observation this is due to gully formation.

Based on the conclusions drawn from the results of the study, it is imperative that a robust local level natural resource management program should be put in place to mitigate the deleterious impact resulting from the degradation of natural resources. Special emphasis should be given to enhance the quality and quantity of the existing forest cover as well as concerted attempts should be made to add appropriate additional area under forest cover. The authors also recommend that a research program to understand the impact of LULC change on the water resources in tandem with the climate variability should be undertaken.