Data on the Land Cover Transition, Subsequent Landscape Degradation, and Improvement in Semi-Arid Rainfed Agricultural Land in North–West Tunisia

: Understanding past landscape changes is crucial to promote agroecological landscape transitions. This study analyzes past land cover changes (LCCs) alongside subsequent degradation and improvements in the study area. The input land cover (LC) data were taken from ESRI’s ArcGIS Living Atlas of the World and then assessed for accuracy using ground truth data points randomly selected from high-resolution images on the Google Earth Engine. The LCC analyses were performed on QGIS 3.28.15 using the Semi-Automatic Classification Plugin (SCP) to generate LCC data. The degradation or improvement derived from the analyzed data was subsequently assessed using the UNCCD Good Practice Guidance to generate land cover degradation data. Using the Landscape Ecology Statistics (LecoS) plugin in QGIS, the input LC data were processed to provide landscape metrics. The data presented in this article show that the studied landscape is not static, even over a short-term time horizon (2017–2022). The transition from one LC class to another had an impact on the ecosystem and induced different states of degradation. For the three main LC classes (forest, crops, and rangeland) representing 98.9% of the total area in 2022, the landscape metrics, especially the number of patches, reflected a 105% increase in landscape fragmentation between 2017 and 2022.


Summary
There is a growing focus on promoting agroecological landscape transitions, i.e., the transition from current agricultural landscapes to new stages of development, with improvements in the principles of agroecology.In Tunisia, an ongoing initiative led by OneCGIAR focuses on agroecological transitions (more information about this initiative can be found at https://www.cgiar.org/initiative/agroecology/,accessed on 8 July 2024).The focal zone of the intervention for the implementation and demonstration of landscape transitions is the region composed of the Kef and Siliana governorates (Figure 1).This zone covers six different sites (i.e., the focal community areas), which together form the Agroecology Living Lab Landscape (ALL) (Kesra, Chouarnia, Elles, Sers, Rhahla, and Hammam Biadha), a concept adopted by the project's research team [1].Under this longterm research strategy, which remains in development, the effective transition of the current Data 2024, 9, 96 2 of 13 agricultural landscape system towards improved agroecology should be based on a correct understanding of the key landscape change processes that occurred in the past.
Data 2024, 9, x FOR PEER REVIEW 2 of 13 Hammam Biadha), a concept adopted by the project's research team [1].Under this longterm research strategy, which remains in development, the effective transition of the current agricultural landscape system towards improved agroecology should be based on a correct understanding of the key landscape change processes that occurred in the past.An LCC is usually a primary change in an agricultural landscape that leads to other changes, thereby forming a landscape transition [2][3][4][5].Firstly, a land conversion type can be ecologically negative (e.g., from forest to scrubland or cropland) or positive (e.g., monocropland to mixed cropland with improved biological management practices or agroforestry).Secondly, the particular spatial distribution of an LCC represents the degree of changes in terms of ecological connectivity in the landscape (e.g., the formation or breaking of green corridors).This factor is important in determining environmental integrity and health.Measuring landscape metrics is one of the key tools used to evaluate agroecosystems, habitat functionality, and regulatory functions [6].These measurements provide scientific evidence on landscape changes and the impact of different agricultural activities on biodiversity, soil quality, and ecosystem resilience [7].LCCs and Land Use Changeinduced landscape alterations and the identification of degradation hotspots can be assessed to aid decision making processes, support effective natural resource management, preserve biodiversity, help ensure food security, and benefit climate change mitigation research [8].Both of these assessment types facilitate the identification of areas of concern, offer guidance towards achieving Sustainable Land Management (SLM), and help to strengthen agricultural system resilience [9].Monitoring the data from these landscape processes is essential for policymakers, researchers, and practitioners involved in sustainable intensification efforts, specifically the promotion of agroecological practices [10].
The data in this study include three interrelated sets: (1) the LCCs in the study region over the period of 2017-2022, (2) the ecological degradation and/or improvements induced by past LCCs, and (3) changes in landscape fragmentation indicating current trends related to particular ecosystem services.The latter two datasets concern subsequent landscape degradation and improvements derived from LCCs.While the first two datasets are pixel based (with a 10 m pixel size), the last dataset includes landscape metrics calculated An LCC is usually a primary change in an agricultural landscape that leads to other changes, thereby forming a landscape transition [2][3][4][5].Firstly, a land conversion type can be ecologically negative (e.g., from forest to scrubland or cropland) or positive (e.g., mono-cropland to mixed cropland with improved biological management practices or agroforestry).Secondly, the particular spatial distribution of an LCC represents the degree of changes in terms of ecological connectivity in the landscape (e.g., the formation or breaking of green corridors).This factor is important in determining environmental integrity and health.Measuring landscape metrics is one of the key tools used to evaluate agroecosystems, habitat functionality, and regulatory functions [6].These measurements provide scientific evidence on landscape changes and the impact of different agricultural activities on biodiversity, soil quality, and ecosystem resilience [7].LCCs and Land Use Change-induced landscape alterations and the identification of degradation hotspots can be assessed to aid decision making processes, support effective natural resource management, preserve biodiversity, help ensure food security, and benefit climate change mitigation research [8].Both of these assessment types facilitate the identification of areas of concern, offer guidance towards achieving Sustainable Land Management (SLM), and help to strengthen agricultural system resilience [9].Monitoring the data from these landscape processes is essential for policymakers, researchers, and practitioners involved in sustainable intensification efforts, specifically the promotion of agroecological practices [10].
The data in this study include three interrelated sets: (1) the LCCs in the study region over the period of 2017-2022, (2) the ecological degradation and/or improvements induced by past LCCs, and (3) changes in landscape fragmentation indicating current trends related to particular ecosystem services.The latter two datasets concern subsequent landscape degradation and improvements derived from LCCs.While the first two datasets are pixel based (with a 10 m pixel size), the last dataset includes landscape metrics calculated for whole landscape boundaries at two different levels: the whole study region and the focal community area.We also provide the aspects of agroecological services that the landscape metrics are intended to indicate.
The input LC data were taken from ESRI's ArcGIS Living Atlas of the World and then assessed for accuracy using ground truth data points randomly selected from highresolution images on the Google Earth Engine.The LCC analyses were performed on QGIS 3.28.15using the SCP to generate LCC data for seven LC classes relevant to the national LC classification.The degradation and improvements derived from the analyzed data were assessed using the UNCCD Good Practice Guidance (GPG) [11] to generate land cover degradation data.The GPG is a reference document providing a theoretical framework to evaluate land degradation for reporting on the United Nations' Sustainable Development Goals.This GPG describes the development of analytical methods for measuring Indicator 15.3.1, as well as Land Degradation Neutrality (LDN) and its three subindicators: (1) the level of negative changes in land cover, (2) the level of land productivity decline, and (3) the level of soil organic carbon decline.The pursuit of "good practices" is due to the incorporation of recent advances in related research, along with national stakeholder engagement in the review of the GPG.This review focused on the scientific relationship between the indicator/sub-indicator metrics and actual land degradation and improvements, presenting a clear and reproducible analytical procedure for calculating and interpreting the indicator/sub-indicators and discussing how to improve data quality and availability [11].
Using the LecoS plugin in QGIS, the input LC data for 2017 and 2022 were processed to provide landscape metrics across various geographic boundaries.For several reasons, quantifying landscape metrics plays an important role in ecological landscape research.The procedure aims to analyze the ways in which different ecological functions are impacted by changes in the land cover and to understand the relationships between ecosystem processes and land cover and changes.Quantifying landscape metrics is also essential to guide the efficient management of landscapes and successful conservation initiatives.Various landscape metrics already exist.However, the implications of these metrics for landscape management always depend upon the features of the specific study landscape, e.g., the scale of the study, the development and management goals, and the specific objectives of users.

Data Description
Three types of data are included in the present study: (1) An LCC map for the Kef-Siliana region covering 2017-2022, (2) land cover degradation data at different scales, including the Kef-Siliana region and sub-spatial units of the ALL in Tunisia, and (3) landscape metrics data at different scales: the Kef-Siliana region and sub-spatial units of the ALL in Tunisia (2017-2022).These three datasets, produced by processing and analyzing input data with the QGIS 3.28.15software, are referenced in the ICARDA Dataverse and freely available for download.The quality and accuracy of the input data were verified with Google Earth satellite imagery and a comparison with a set of GPS points in different fields within the study area.
The global ESRI LC data (with a pixel size of 100 m 2 ) from 2017 and 2022 were clipped to cover the Kef-Siliana region.The accuracy of these data was assessed using the following steps: 1.
For each LC class, one hundred points (i.e., sampling points) were randomly selected, for a total of seven hundred sampling points across all seven LC classes.

2.
These sampling points were located on Google Earth Pro.The historical time function was then used to locate the target year (i.e., 2017 then 2022).Then, we zoomed in to visualize land regions with true color, high resolution satellite images.

3.
For each sampling point, we observed and identified the LC type.

4.
For each target year, we constructed an accuracy assessment table, which counted sample points with ESRI LC classes matching the visual identification using Google Earth Pro (Figure 2). 4. For each target year, we constructed an accuracy assessment table, which counted sample points with ESRI LC classes matching the visual identification using Google Earth Pro (Figure 2).The results of the accuracy assessment are reflected in Tables 1 and 2.  The results of the accuracy assessment are reflected in Tables 1 and 2. Tables 1 and 2 show that the LC data in 2017 and 2022 had accuracy values of 95.3% and 96.1%, respectively, which are sufficient for LCC analyses in research and development.

Land Cover Change Map for the Kef-Siliana Region over 2017-2022
The map in Figure 3 illustrates the LCC between 2017 and 2022 (https://hdl.handle.net/20.500.11766.1/FK2/YUXPQY,accessed on 8 July 2024) (Figure 3) with a legend including 49 types of LCCs, as presented in Table 3.Additionally, a Sankey diagram (Figure 4) is used to visualize the LC transitions, illustrating the different changes in LC classes within the Kef-Siliana region from 2017 to 2022.Tables 1 and 2 show that the LC data in 2017 and 2022 had accuracy values of 95.3% and 96.1%, respectively, which are sufficient for LCC analyses in research and development.

Land Cover Change Map for the Kef-Siliana Region over 2017-2022
The map in Figure 3 illustrates the LCC between 2017 and 2022 (https://hdl.handle.net/20.500.11766.1/FK2/YUXPQY,accessed on 8 July 2024) (Figure 3) with a legend including 49 types of LCCs, as presented in Table 3.Additionally, a Sankey diagram (Figure 4) is used to visualize the LC transitions, illustrating the different changes in LC classes within the Kef-Siliana region from 2017 to 2022.  3.  Tables 4 and 5 present the area (in ha) of gain or loss for each type of LC, as well as the percentage of change in the Kef-Siliana region and across the six focal community areas of the ALL.
Rangeland, crops, and forest are the main LC classes in the Kef-Siliana region showing constant improvement, with increases of 1.5%, 22%, and 10.6%, respectively, between 2017 and 2022.The highest gain was recorded for crops (5777 ha), and the highest loss was recorded for bare land (−74,377 ha).At the ALL level, the forest improved in Hammam Biadha and Kesra by 19.3% and 18.3%, respectively, with an area of 394 ha and 233 ha.Crop improvements of 44.7%, 14.4%, 11.2%, 8.7%, and 4.2% took place in Rhahla,  Note: To calculate the area of LCC between periods, we employed the following equation: A = AT2 − AT1, (T2 > T1), where A refers to the area of LCC in ha between periods, AT2 is the area of the LC at year T2, and AT1 is the area of the LC at year T1.The (+) and (−) symbols indicate the area's gain and loss, respectively.Note: To visualize the LCC criteria developed for relating LC transitions to degradation and nondegradation processes, red is used for LC transitions with the potential to degrade ecosystems during 2017-2022, green is used for improvements, and grey is used to depict stable areas that remained unchanged [4].All values are rounded in ha.Most land in the ALL is classified as stable, with percentages of change ranging from 81.7% to 92.5% in different areas.Areas of degradation represent a smaller proportion of the total land, with percentages ranging from 2.7% to 9.9%.The areas showing improvement range from 3% to 11.2%.The highest degradation percentages were observed in Rhahla and Chouarnia, with values of 9.7% and 9.9%, respectively.The highest improvement was observed in Elles, with a percentage of 11.2%.Finally, Kesra presented the highest level of stability with a percentage of 92.5%.

Data on Landscape Metrics at Different Scales: the Kef-Siliana Region and Sub-Spatial Units of the Agroecological Living Lab Landscape in Tunisia
A portion of the results for landscape metrics (with definitions and significance to ecological integrity/health as described in Table 8 and presented in Table 9) and all corresponding calculations are available in an Excel file (https://hdl.handle.net/20.500.11766.1 /FK2/UN7DKQ, accessed on 8 July 2024).These results indicate that the three major classes of LC in the Kef-Siliana region, forest, crops, and rangeland, occupied a total landscape proportion of 98.9% in 2022, with changes of +105.6% and +108.6%,respectively, in the number of patches (NP) and Largest Patch Index (LPI) for the three LC classes together.For the Median Patch Area, forest presented a change of +105.2%, while rangeland yielded a decrease of −25%, and crops maintained the same value of 300 m 2 between 2017 and 2022.Similar trends were observed for these metrics across different units of the ALL.
Table 8.List of computed landscape metrics and their significance for ecological integrity and health.

Landscape Proportion LP
Proportion of the landscape occupied by an LC class (index between 0 and 1) The high forest and woodland LP supports greater species diversity among plants and animals [12].

Number of Patches NP
Number of patches with the same LC class in the landscape (index ≥ 1) The high NP for health vegetative covers indicates unwanted fragmentation, which reduces species diversity and forest recovery [13].However, more patches of woody (e.g., forest and woodland) and non-woody (e.g., crops and grassland) cover may yield specific types of expected fragmentation that provide greater friction to reduce the risk of crop pests and diseases in agriculture [14].Firstly, we obtained LC maps covering 2017 and 2022 for the Kef-Siliana region with a resolution of 10 m from the Esri Sentinel-2 Land Cover Explorer ArcGIS Living Atlas of the World.Esri Sentinel-2 data are characterized by their consistency and reliability and are considered a dependable resource for LC studies.Moreover, the free open-source policy governing these data encourages widespread use and facilitates access without financial barriers.This resource is commonly used to monitor and assess land degradation in arid regions [5,15] with independent accuracy assessments using high resolution truecolor images in Google Earth Pro [16].
Next, the data were treated with QGIS 3.28.15using the SCP Python plugin 8.2.2 (1a).This algorithm automates the LC classification phases, including downloading raw data; preprocessing, processing, and postprocessing the acquired data; and generating LCC data.The output of this process was then used as an input associated with a predefined land cover degradation matrix (2a) to generate land cover degradation data.This matrix labels and quantifies each land cover change as either degraded, improved, or stable.When a transition has the potential to deteriorate ecosystems, it is considered to represent degradation; otherwise, a transition to other LC classes is considered an improvement, such as when a degraded site is restored.
The same LC maps (2017-2022) were subsequently treated using the Landscape Ecology Statistics (Lecos) Python plugin in QGIS (1b) to generate landscape metrics and provide several functions to conduct landscape analyses from raw raster data.
The presented method for generating data on land cover transition, subsequent landscape degradation, and improvements works well for several reasons.First, the data analysis results are reproducible, which is important for scientific research.The tool use procedure is also designed to be simple, easy, and accessible for users without advanced technical skills, requiring minimal training.Additionally, free access is crucial for projects with limited financial resources.Overall, the use of open-source plugins with publicly available data significantly reduces costs.This method also reduces the time required, making it advantageous for quick evaluations and fast data processing.
The validity of the presented output data is mainly relevant for studying the spatial and temporal patterns of landscape changes captured by remote sensing corresponding to land surface attributes (e.g., land cover).Although these pattern changes are evidenced by physical reflectance, the classification of land cover types and the translation of land cover changes into levels of land cover degradation always retain certain levels of human Next, the data were treated with QGIS 3.28.15using the SCP Python plugin 8.2.2 (1a).This algorithm automates the LC classification phases, including downloading raw data; preprocessing, processing, and postprocessing the acquired data; and generating LCC data.The output of this process was then used as an input associated with a predefined land cover degradation matrix (2a) to generate land cover degradation data.This matrix labels and quantifies each land cover change as either degraded, improved, or stable.
When a transition has the potential to deteriorate ecosystems, it is considered to represent degradation; otherwise, a transition to other LC classes is considered an improvement, such as when a degraded site is restored.
The same LC maps (2017-2022) were subsequently treated using the Landscape Ecology Statistics (Lecos) Python plugin in QGIS (1b) to generate landscape metrics and provide several functions to conduct landscape analyses from raw raster data.
The presented method for generating data on land cover transition, subsequent landscape degradation, and improvements works well for several reasons.First, the data analysis results are reproducible, which is important for scientific research.The tool use procedure is also designed to be simple, easy, and accessible for users without advanced technical skills, requiring minimal training.Additionally, free access is crucial for projects with limited financial resources.Overall, the use of open-source plugins with publicly available data significantly reduces costs.This method also reduces the time required, making it advantageous for quick evaluations and fast data processing.
The validity of the presented output data is mainly relevant for studying the spatial and temporal patterns of landscape changes captured by remote sensing corresponding to land surface attributes (e.g., land cover).Although these pattern changes are evidenced by physical reflectance, the classification of land cover types and the translation of land cover changes into levels of land cover degradation always retain certain levels of human subjectivity.Therefore, further evaluations of the results with independent reference data and information are recommended (e.g., land cover change mapping based on other source data and agreement with local expert opinions).Field observations at the change hotspots presented in this study are also encouraged.However, the relationship between fieldbased observations and mapped degradation should be interpreted through two types of observations: (1) observations that validate the land cover degradation (e.g., observed discrete conversions among land cover types) and ( 2) observations that reflect other aspects

Figure 1 .
Figure 1.Maps of the study area.Left map: Kef-Siliana region in Tunisia; right map: boundaries of the Kef and Siliana governorates and the six focal community areas forming the Agroecological Living Lab Landscape (ALL) of the One CGIAR Initiative on Agroecology.

Figure 1 .
Figure 1.Maps of the study area.Left map: Kef-Siliana region in Tunisia; right map: boundaries of the Kef and Siliana governorates and the six focal community areas forming the Agroecological Living Lab Landscape (ALL) of the One CGIAR Initiative on Agroecology.
Data 2024, 9, x FOR PEER REVIEW 4 of 13

Figure 2 .
Figure 2. Map of sampling points in the Kef-Siliana region.

Figure 2 .
Figure 2. Map of sampling points in the Kef-Siliana region.

Figure 3 .
Figure 3. Map of land cover changes between 2017 and 2022.The legend codes are provided in Table3.

Figure 4 .
Figure 4. Sankey plot describing the LCC transitions from 2017 to 2022.Bands represent the actual proportion of land that changed class over time.

Figure 4 .
Figure 4. Sankey plot describing the LCC transitions from 2017 to 2022.Bands represent the actual proportion of land that changed class over time.Tables 4 and 5 present the area (in ha) of gain or loss for each type of LC, as well as the percentage of change in the Kef-Siliana region and across the six focal community areas of the ALL.

Table 1 .
Accuracy assessment of ESRI LC data in 2017 in the Kef-Siliana region.LC ClassNo.

Table 1 .
Accuracy assessment of ESRI LC data in 2017 in the Kef-Siliana region.

Table 2 .
Accuracy assessment of ESRI LC data in 2022 in the Kef-Siliana region.

Table 4 .
Area differences in land cover between 2017 and 2022 in ha.

Table 6 .
Land cover degradation matrix (numbers in ha) and related LC transition type (in text).

Table 6 .
Land [4]er degradation matrix (numbers in ha) and related LC transition type (in text).Note: To visualize the LCC criteria developed for relating LC transitions to degradation and non-degradation processes, red is used for LC transitions with the potential to degrade ecosystems during 2017-2022, green is used for improvements, and grey is used to depict stable areas that remained unchanged[4].All values are rounded in ha.

Table 7 .
Land cover degradation state.