%0 Generic %1 bauer2018detecting %A Bauer, Christian %A Löw, Fabian %A Klein, Doris %A Conrad, Christopher %D 2018 %K conference poster LSFE %T Detecting spatio-temporal patterns of land abandonment in the lower region of the Amu Darya and Syr Darya Rivers using earth observation data %X The collapse of the Soviet Union initiated an increased abandonment of farmland in Uzbekistan and Kazakhstan. However, the dependency on agricultural production is still high, particularly in the context of food security and the rapidly growing population. Despite vast research on land degradation, the processes and drivers for abandonment remain hardly understood. One reason for this gap may be the little attention that was paid to site-specific information about abandonment of arable land in irrigation agriculture. In this study, analysis of time series from Landsat satellite data is recognized as highly suitable to establish retrospective and current land use changes. A methodological framework was developed, which utilizes Random Forest (RF) machine learning to classify arable land and to discriminate between used and unused fields. It further subdivides fields classified as “unused” in sparse, open and dense shrubland by analyzing the intensity of their vegetation signal in the satellite data. The framework also encloses a ruleset for accurate decisions for extremely heterogeneously vegetated fields. These steps, together with the application of the methodological framework to the study region in the observation period 1998-2016 permit to receive area-wide information on the timing when a land parcel was abandoned. The RF regression was applied to assess the most important spatial parameters (drivers) leading to land abandonment. The complexity of the crop rotations, long fallow cycles, and the data scarcity challenged the detection of abandoned land. The annual maps show that abandonment mainly occurred on fields at the edges of the irrigation area and on fields close to main drainage systems. The methodological framework is now ready for the identification of abandoned land hotspots, time steps and drivers and can hence support regional land use planners to improve land management, e.g. to take site specific decisions on the potential reuse of fields, irrigation maintenance, or alternative land use options.

@conference{bauer2018detecting, abstract = {The collapse of the Soviet Union initiated an increased abandonment of farmland in Uzbekistan and Kazakhstan. However, the dependency on agricultural production is still high, particularly in the context of food security and the rapidly growing population. Despite vast research on land degradation, the processes and drivers for abandonment remain hardly understood. One reason for this gap may be the little attention that was paid to site-specific information about abandonment of arable land in irrigation agriculture. In this study, analysis of time series from Landsat satellite data is recognized as highly suitable to establish retrospective and current land use changes. A methodological framework was developed, which utilizes Random Forest (RF) machine learning to classify arable land and to discriminate between used and unused fields. It further subdivides fields classified as “unused” in sparse, open and dense shrubland by analyzing the intensity of their vegetation signal in the satellite data. The framework also encloses a ruleset for accurate decisions for extremely heterogeneously vegetated fields. These steps, together with the application of the methodological framework to the study region in the observation period 1998-2016 permit to receive area-wide information on the timing when a land parcel was abandoned. The RF regression was applied to assess the most important spatial parameters (drivers) leading to land abandonment. The complexity of the crop rotations, long fallow cycles, and the data scarcity challenged the detection of abandoned land. The annual maps show that abandonment mainly occurred on fields at the edges of the irrigation area and on fields close to main drainage systems. The methodological framework is now ready for the identification of abandoned land hotspots, time steps and drivers and can hence support regional land use planners to improve land management, e.g. to take site specific decisions on the potential reuse of fields, irrigation maintenance, or alternative land use options.}, added-at = {2020-09-11T12:04:50.000+0200}, author = {Bauer, Christian and Löw, Fabian and Klein, Doris and Conrad, Christopher}, biburl = {https://www.bibsonomy.org/bibtex/248ff3aa43cf9fdc947fae303069ac39f/earthobs_uniwue}, eventdate = {16.-19.04.2018}, eventtitle = {Between Europe and the Orient, Status symposium}, interhash = {e559acbb4fb27ddb2e4652562d901e8c}, intrahash = {48ff3aa43cf9fdc947fae303069ac39f}, keywords = {conference poster LSFE}, month = {April }, timestamp = {2020-11-18T22:08:33.000+0100}, title = {Detecting spatio-temporal patterns of land abandonment in the lower region of the Amu Darya and Syr Darya Rivers using earth observation data }, venue = {Almaty/Kasachstan}, year = 2018 }