Understanding traditional agro-ecosystem dynamics in response to systematic transition processes and rainfall variability patterns at watershed-scale in Southern Ethiopia

Highlights

• This study was conducted in a traditional agroforestry-dominated land use system, Southeastern Rift Escarpment of Ethiopia.

• We have evaluated multi-temporal land cover dynamics over three decades in relation to rainfall variability patterns.

• It presents inter-category land cover transition pattern, and its correlation with rainfall amount and variability.

Abstract

Multi-temporal analysis of land cover dynamics using remote sensing can enable the determination of the spatial extent and average rate of land cover change. With the application of an appropriate change analysis method, it is also possible to distinguish whether a land cover change has occurred by the effect of a random or a systematic process. In connection with this, characterizing rainfall variability and historical meteorological drought events can allow understanding of their effects on agro-ecosystems and vegetation cover dynamics. Therefore, this study has evaluated multi-temporal land cover change in response to the possible impacts of population pressure and rainfall variability on agro-ecosystem dynamics. This was conducted on a traditional agroforestry-dominated landscape in Southern Ethiopia. Using Landsat images acquired in 1985, 2000, and 2018, a post-classification land cover change analysis approach was employed to distinguish between a systematic and random process of inter-category transitions. Assessment of drought events and rainfall variability dynamics were performed using standardized precipitation index (SPI) and rainfall coefficient of variation (CV), respectively. Mann–Kendall test was also applied for the detection of a monotonic rainfall trend. A bias-corrected Climate Hazards group Infrared Precipitation with Stations (CHIRPS) over 1981–2017 was used to calculate the SPI, CV and Mann–Kendall trend test. The analysis showed that above 41% of the landscape has experienced land cover transitions between 1985 and 2018. This has primarily resulted by a systematic and rapid expansion of agriculture, urban areas, and eucalyptus plantations, at the expense of natural vegetation ecosystems. Consequently, over the last 33 years (1985–2018), natural forest, grassland, and wetland have declined by 74.8%, 83.3%, and 78.4%, respectively. Another major land cover change identified in this study was the replacement of open-field crops by agroforestry, mainly in the western part of the catchment. Such expansion of agroforestry has appeared to be spatially correlated with a lower amount of long-term average and more variable rainfall. Perhaps, this could indicate farmers’ response to rainfall variability by diversifying agricultural production options (i.e. agroforestry system), by replacing the more risk-prone mono-cropping culture. The observed persistence and further expansion of traditional agroforestry (a combination of perennial crops and scattered trees) have implications in terms of enhancing biodiversity conservation and environmental protection. Overall, the land cover transitions that occurred over the last three decades suggest future conservation priorities for improved landscape management, with more emphasis on the most exposed natural vegetation ecosystems.

Introduction

Land cover changes are dynamic processes that occur at varying rates and spatial scales, primarily as a response to socioeconomic forces by human-environment interactions, while less frequently by natural hazards such as floods and wildfires (Campbell and Wynne, 2011, Mölders, 2011). Such human-induced changes are caused by land management decisions principally made at local-scales. Land cover changes (hereafter LCC) can range from modification of its characteristics (an alteration of the cover extent and/or shift in location) without affecting the number of previous information classes, to the extreme case of complete replacement by other cover types, referred to as land conversion (Mölders, 2011).

The terms ‘land cover’ and ‘land use’ are fundamentally different although they are usually used coining both terms together (Lillesand et al., 2015). While ‘land cover’ designates the discernable physical features on the earth’s surface, ‘land use’ defines the economic use of those land cover types (Campbell and Wynne, 2011, Lillesand et al., 2015). Since the study employs remote sensing data and methods, the term ‘land cover’ has been frequently used throughout the paper instead of coining it with the term ‘land use’.

The present study was conducted on a predominantly traditional agroforestry land cover in Southern Ethiopia. This agroforestry principally comprises perennial crops, combined with scattered native and exotic trees. As such, it supports biodiversity conservation, environmental protection and also serves as a potential for carbon sequestration (Kanshie, 2002, Tesemma, 2013). Generally, agroforestry appears to be a favorable land cover, particularly under the phenomenon of global climate change (Verchot et al., 2007). More specifically, the traditional agroforestry in the present study area offers various ecosystem services due to the combination of annual and perennial crops, as well as native and exotic tree species (Abebe and Bongers, 2012, Abebe et al., 2010, Gebrehiwot, 2013, Kanshie, 2002, Mellisse et al., 2018). However, despite its multipurpose environmental and socioeconomic benefits, a growing transition to mono-cropping culture has started to influence the sustainability of this environment-friendly agricultural production system (Abebe and Bongers, 2012, Abebe et al., 2010, Gebrehiwot, 2013, Mellisse et al., 2018). This has occurred due to the cultivation of new cash crops (mainly khat/Catha edulis), eucalyptus plantation (Abebe and Bongers, 2012, Gebrehiwot, 2013, Negash, 2002), and expansion of open-field crops (Abebe and Bongers, 2012). The densely inhabited and rapidly growing population, particularly in the highland parts of the study area has led to a declining farmland size per household, enforcing farmers to increase the proportion of open-field crops, by replacing the more sustainable agroforestry system, to meet their immediate subsistence needs (Abebe and Bongers, 2012). In contrast, while several survey-based studies have indicated a growing land cover transition in this area (Abebe and Bongers, 2012, Abebe et al., 2010, Gebrehiwot, 2013, Mellisse et al., 2018), remote sensing-based spatiotemporal information on land cover dynamics has been lacking. Therefore, the ongoing LCC coupled with the lack of up-to-date information on its spatiotemporal dynamics, calls for multi-temporal evaluation along with examining the underlying systematic or random processes triggering the change.

On the other hand, since the launch of the Landsat program in the 1970s, satellite remote sensing has played a leading role in monitoring land cover dynamics at varying spatiotemporal scales. This has been enhanced by the growing access to the increasing archives of earth observation data and image processing platforms (Keenan et al., 2015, Zhang et al., 2013). Moreover, the growing multi-sensor satellite image data has augmented the availability of satellite-derived rainfall estimates useful for monitoring rainfall dynamics and characterizing meteorological drought events (Bayissa et al., 2017, Sein et al., 2018, Tesfamariam et al., 2019b). Remote sensing has thus become suitable both for land cover change analysis and for the assessment of rainfall variability and drought conditions. Attributed to its repeated imaging of the earth’s surface, remote sensing offers a unique capability in monitoring LCC by tracking gradual processes occurred over time. In addition, analysis of long-term rainfall variability and meteorological drought events based on satellite rainfall products can help to understand its spatial correlation with the patterns of land cover change, with more emphasis on vegetation cover dynamics. Such integrated biophysical information may help to distinguish direct human impacts and climate variability influences, enhancing to gain useful insights on the development of mitigation strategies to alleviate undesired land cover change.

Therefore, based on remotely sensed tools, the main objective of this study was to evaluate multi-temporal land cover change resulted by a random or systematic process, along with assessing the impact of rainfall variability and meteorological drought events on vegetation cover dynamics, in the upper catchment of the Gidabo river basin, Southeastern escarpment of the Ethiopian Rift Valley Lakes Basin. Accordingly, the study has first evaluated inter-category land cover transitions over the last three decades (1985–2018). Then using a post-classification change analysis approach, we have tried to reasonably associate the observed land cover changes with a random or a systematic process. Finally, by correlating their spatiotemporal distribution patterns, the possible effect of rainfall variability and historical meteorological drought events on vegetation cover change was assessed.