the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 28 Apr 2026
"Fertile islands from above": Utilising UAS imagery to map argan forest patches in a UNESCO biosphere reserve in Morocco
Abstract. Vegetation in dryland ecosystems often exhibits spatial patterning that leads to the formation of fertile islands. Discrete patches accumulate nutrients, organic litter, seeds and water, slow down geomorphodynamic dispersion processes and contrast with rather barren interpatch areas. While this fragmentation shapes surface dynamics in dryland environments across the globe, their spatial arrangement remains difficult to quantify at fine scales. In this study, we utilised data collected by an uncrewed aircraft system (UAS) together with field data to study surface patterns in degraded Argania spinosa forests in South Morocco that show fertile island dynamics. Point clouds generated from UAS imagery using a Structure-from-Motion photogrammetric workflow were classified into vegetation and ground points, allowing the derivation of digital terrain and digital surface models as well as conventional orthophotos and artificial ground-only orthophotos. This high-resolution geospatial data was used to map tree-influenced soil surface areas and crown areas. Their size and spatial relationships were compared and complemented by a detailed assessment of tree morphologies and terrain characteristics based on both field observations and UAS-based geodata. Spatial and statistical analyses were conducted to study the effects of tree morphology, hillslope wash, shading and wind on emerging surface patterns beneath Argania spinosa trees.
Across the 496 evaluated tree-influenced areas, the surface influence extended on average to 1.69 times the size of the crown covered area. The extent of this influence beyond the canopy cover is strongly controlled by tree size and morphology, indicating that browsing-induced degradation influences not only tree conditions but also the spatial extent of positive surface effects in the interpatch area. The tree-influenced areas exhibit a consistent north-east displacement, a pattern that surprisingly appears largely decoupled from hillslope wash and is most reasonably explained by a combined influence of shading and wind effects.
The results of this study demonstrate the potential of UAS imagery to complement fertile island research by providing spatial insight beyond conventional field-based assessments. At the same time, the impact of browsing on surface dynamics in a UNESCO biosphere reserve is highlighted as a factor contributing to degradation in this silvopastoral land-use system.
- Preprint
(33270 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 30 Jun 2026)
-
RC1: 'Comment on egusphere-2026-1587', Mohamed Mouafik, 28 May 2026
reply
-
AC1: 'Reply on RC1', Lars Engelmann, 07 Jun 2026
reply
Dear Mohamed Mouafik thank you for your valuable and encouraging feedback. We considered your suggestions carefully and believe we can improve the manuscript by incorporating them. Please find our response to your comments below, highlighted in bolt italics:
This manuscript presents an interesting and valuable contribution to understanding fertile island dynamics in Argane ecosystems through the integration of UAS imagery, photogrammetric analysis, and spatial modelling approaches. The study is methodologically comprehensive and addresses an important ecological issue in dryland environments. The following points are suggested to improve the quality, clarity, and scientific robustness of the manuscript:
1- In figure 2 the Argane forest distribution is not clearly visible; please improve its visualization by using a contrasting color or adding a clear boundary outline to enhance spatial interpretation.
We improved the figure by using a more contrasting colour for the Argane distribution area. Please have a look at the attached proposal (abb_3_1_study_area_v2.pdf).
The study provides regional location and site names, but exact coordinates of the individual test plots are not reported in the main text, which may limit full spatial reproducibility of the sampling design.
We suggest adding an additional table to the Appendix, which lists the test plots names, coordinates and the UAS survey dates to improve the reproducibility of the sampling design. This also improves the comprehensibility of the geodata available on Zenodo (https://doi.org/10.5281/zenodo.19128639).
2- At the international conference on the Argane tree, it was decided to standardize the name as Argane (with a capital “A” and an added “e” at the end) to better reflect the correct Amazigh pronunciation.
We support action to standardize the spelling in international literature while better reflecting the cultural heritage and origin of names. Therefore, we will happily adopt the agreement reached on the international conference on the Argane tree. Could you provide us with some more details on the agreement reached, so we can refer to it in the future? Currently we cannot find an according statement in the reporting on this year’s conference.
3- In Line 140: ground-only orthophoto shows the appearance of the terrain only” → repetitive (“only” repeated).
We will improve this to:
- “Other than a standard orthophoto, this ground-only orthophoto shows the appearance of the terrain surface without the Argane trees”.
4-The manuscript does not report key flight parameters such as forward (longitudinal) and side overlap of images.
Since more information and a visualisation of the rather complex flight scheme is given in the referenced open-access source (Marzolff et al. 2020), we wanted to keep the description short in this paper. But we can add more details for improved reproducibility:
- “We took overlapping RGB images at 50 m flying height (approx. 1.5 cm image resolution) following a flight scheme designed to capture full 3D tree shape as well as the ground beneath the tree crowns and parts of the stems (see Fig. 9-19B in Aber et al., 2019; Marzolff et al., 2020). To this end, vertical (90°) and oblique (60°) images in rows with approx. 80% forward overlap and 50% sidelap were complemented with convergent oblique (50°) imagery in two concentrical circles. 14 ground control points …”
5- Correct this: optimal conditions were given for visual assessment→ better: conditions were optimal for ground visibility and image interpretation// almost complete absence of shadowing→ better: minimal shadow effects.
We agree with this reformulation and can make according edits. We will change this:
- “The weather on the autumn survey days was bright but overcast, resulting in indirect illumination and minimal shadow effects, so conditions were optimal for ground visibility and image interpretation.”
6- How can ERA5 wind data at a coarse spatial resolution (0.25° grid) reliably represent local-scale wind conditions within each study area, and how do you justify assigning a single grid point to characterize microclimatic variability relevant to ecological patch-scale analysis, given the potential for spatial simplification and mismatch with actual local conditions?
The available ERA5 data certainly cannot reliably represent microclimatic variability at the individual study sites. There is a considerable gap between actual local wind conditions and the larger scale patterns captured by ERA5 that is promoted by a multitude of factors such as local topography. We mentioned these limitations in chapter 4.3. However, the ERA5 data still provide information on the broader scale wind patterns in Taroudant, Ida-Outanane and Ait Baha in absence of long-term wind measurements on each test site. This is used as a basis for discussion on the spatial surface patterns around Argane trees. The observed displacements of tree influenced areas suggest that a more detailed investigation of this topic with higher resolution wind data is worthwhile. Based on your comments 20 and 21 it seems like we generally agree on this manner. We suggest addressing the limited availability of wind data and the limits of our ERA5 approach earlier in the methodology section of the manuscript to improve the clarity of the manuscript. Therefore, we propose to add the following section in line 191 and 196:
- “Due to the absence of local wind measurement stations at our test sites, we used ERA5 data provided by the European Centre for Medium-Range Weather Forecasts (ECMWF) and described by Hersbach et al. (2020) to investigate the general wind conditions in the three study areas.”
- “While the ERA5 data provide a good representation of the general wind patterns across the tree study areas, it should be noted that local wind conditions can significantly deviate from these regional patterns.”
7- The coordinates are presented in a mixed format (decimal degrees with N/W directional labels), which is not strictly correct for decimal degree notation; in DD format, direction letters should be removed and West longitudes expressed as negative values (like: 30.0743, -9.14904).
We gave coordinates in the format of unsigned decimal degrees with directional letters as this is more reader-friendly and less ambiguous but still easier to use in in GIS contexts than DMS format. Thank you for pointing out the unconventional order of longitude and latitude, which we will correct to lat, lon.
8- The term “offsets of these centroids” is not clearly defined; it is unclear whether this refers to Euclidean distance, directional displacement, or a vector from one centroid to another (e.g., CA → TIA or TIA → CA), which reduces the reproducibility and clarity of the spatial vector analysis.
“Offset between centroids” in section 2.5 refers to the vector length and orientation between two respective centroids. We calculated such vectors between multiple combinations of TIA, offTIA, CA, offMSA, and offiTIA centroids. In the final manuscript it is mostly referred to the CA to offTIA centroid, starting from the CA. We suggest clearing that up by making some edits in section 2.5:
- „Using a horizontal 2D vector analysis, we determined the offset between these centroids, measured as vector length and orientation between the two respective centroids under consideration. The results presented in this study were always measured from the CA centroid. “
Additionally, some edits can be made to Table 1, figure caption 11 and 12 to clarify this throughout the manuscript.
9- The interpretation of results is sometimes mixed into the Results (e.g., explanations of degradation and architecture effects), which should be reserved for the Discussion section.
Together with the remarks made in comment 13 we plan to make some small modifications to the results chapter to add additional statistical details while moving some sentences to the discussion:
- “Differences in netTIA across architecture classes coincide with differences in tree size (Figure 10): More degraded trees assigned higher architecture classes have a generally smaller height and netTIA.”
- “Median displacement length decreased with more severe tree degradation. “
10- The strong correlation (R² = 0.92) between CA and TIA is reported, but no check for mathematical dependency or circular reasoning is provided.
Since both areas were mapped individually, there is no mathematical dependency between both. However, since litter input of the CA is an important driver for TIA development, a clear correlation between both was to be expected.
11- The statement “CA are generally smaller than TIA” is obvious and not sufficiently informative, given that TIA is defined to include CA.
TIA does not necessarily included CA, as it is defined as the “terrain surface that shows ecological influence from Argania spinosa trees and visually contrasts with the surrounding bare interpatch area” (146-148). It mostly extends beyond the crown area CA, but “there may also be areas beneath the tree crowns that are not visibly tree-influenced, i.e. CA but not TIA” (lines 168-169). This is also visible in Fig. 1 and Fig. 5B right.
12- The term “correlated (R² = 0.92)” is incorrect; R² should be used to describe a relationship (regression), not correlation (which requires r).
Yes, you are right. By additionally considering your concerns regarding the term ‘relationship’ in comment 14, we can improve the clarity of the statistical reporting throughout the manuscript by making a few changes to our wording. In this case we suggest:
- “There is a positive relationship between CA and TIA (R² = 0.92, Figure 6), with the corresponding linear regression showing that a 1 m² increase in CA corresponds to an average increase of 1.69 m² in TIA.”
13- The reporting of statistical results is incomplete, as key details such as exact p-values, test statistics (W, H, Z values), and effect sizes are not provided, limiting the interpretability and reproducibility of the analyses.
We worked on a significance level of 5% for all conducted statistical test and performed a Benjamin-Hochberg adjustment during post-hoc testing. In the results section we simply reported whether a test under these conditions was significant or not to improve the readability. This decision indeed limits the interpretability of our results. To address this issue, we will add the p and chi-squared values in the text for the Wallraff and Kruskal-Wallis tests as well as the detailed results of the post-hoc tests in the appendix.
14- The term “relationship” is used inconsistently across results analyses (correlation, regression, displacement), which may cause statistical confusion.
We suggest reserving the word relationship for the description of regressions only. This requires the following edits in line 10, figure caption 6, figure caption 10 and line 421:
- „Their size and spatial relation were compared and complemented by a detailed assessment of tree morphologies and terrain characteristics based on both field observations and UAS-based geodata. “
- „Figure 6. Scatter plot of the crown area (CA) [m²] against the tree-influenced area (TIA) [m²] for the observed trees (n). The resulting formula of the linear regression (red line) is shown at the top left. It explained 92% of the total variance. “
- „Figure 10. Scatter plot of the mean crown height [m] against the netTIA [m²] for the observed single tree netTIAs (n). Different architecture classes are visualised by colour from green (low degradation) to red (high degradation) as well as shape. A power function was fitted to the data. It explained 55% of the total variance. “
- “The spatial and statistical analyses focused on the relation between crown area and tree-influenced area, the role of tree-crown characteristics and the potential influences of tree shade, wind direction and slope direction on the fertile island patterns. “
15- The overlap analysis results are reported, but effect sizes are not clearly quantified, and the practical ecological significance of the observed differences is not discussed.
Yes, so far, no effect sizes are reported. We can quantify effect sizes by calculating the Cohen’s d measure of effect size (rstatix-package). The according results show strong effect sizes for lower and higher architecture classes with small to negligible effects for the classes in between:
- -0.915 for group “1/2“ indicating a large effect size
- -0.413 for group “3” indicating a small effect size
- -0.338 for group “4” indicating a small effect size
- -0.183 for group “5” indicating a negligible effect size
- 449 for group “6” indicating a small effect size
- 815 for group “7” indicating a large effect size
- 56 for group “8” indicating a large effect size
We plan to add the Cohen’s d values to the results chapter and expand chapter 4.2 with some remarks on the implications of these results.
16- The ecological interpretation of results is sometimes over-generalized, especially when linking surface patterns directly to subsurface nutrient processes without direct measurements.
Since surface patterns in fertile island systems have not yet been studied with UAS imagery, there is no available literature for a direct comparison of our results. While our approach has clear benefits for the monitoring of fertile island systems such as the evaluation of large-scale surface processes, its transferability to studies focused on soil sampling is indeed severely limited. We still believe that such larger-scale evaluations with UAS imagery can improve the understanding of accumulation and dispersion processes around fertile islands by providing information that cannot be captured in most soil sampling campaigns. Thereby the currently unclear situation of transferability might be resolved. Right now, we believe, that our results should be discussed in the context of the available soil sampling studies to facilitate this development. We agree, however, that the limitations of this transfer should be clear and believe that we can improve the manuscript in this regard. For example in line 294-300:
- “However, the described surface patterns do not necessarily have to translate to the nutrient dynamics described in other fertile island studies (e.g., Eldridge et al., 2024; Fitzpatrick et al., 2024; Mudrak et al., 2014; Perez, 2019). It was recognised that the strength of the fertile island effect (Eldridge et al., 2024; Fitzpatrick et al., 2024) and the size of the fertile islands (Butterfield and Briggs, 2009) are dependent on plant size. Mudrak et al. (2014) studied shrubs in Sonoran and Mojave Deserts of the United States, where shrub 300 size was an important predictor of the extent of nutrient distribution. This implies that the larger extent of visible surface changes beneath bigger trees observed in this study, could translate to an increase in nutrient accumulation extend and intensity. While such considerations are currently speculative, a closer link between nutrient dynamics and surface patterns observed by UAS could advance fertile island monitoring in the future. “
17- Some claims about ecological processes (fertile island strength, erosion risk, system stability) are not fully supported by the presented quantitative results.
We agree in regard to the fertile island strength and suggest according changes as proposed under comment 16. When it comes to the erosion risk and system stability, we believe that the evaluation of the results does benefit from insights gained in previous studies made on the same test sites such as Kirchhoff et al., 2022 and Kirchhoff et al., 2020.
18- The role of slope is inconsistently interpreted (dismissed as non-significant but still suggested to have indirect influence), which weakens logical consistency.
The slope did not contribute directly to the displacement of the tree influenced area in this study. In line 378-380 we wanted to point out that it nonetheless can affect the overall erosion rates and therefore individual site degradation in both tree influenced and interpatch areas. However, this evaluation is not directly derivable from the results presented in this study and thus weakens the overall consistency of the manuscript. We suggest removing the according statements.
19- The statement that wind and shading are the “most plausible explanation” is not statistically demonstrated, but presented as a near-conclusion.
The statement referred to the “most plausible explanation” for the observed displacements within the context of this study, especially compared to the slope. To make this intention clearer we suggest these changes in lines 374/375 and 430/ 431:
- „Overall, a combined influence of wind redistribution and directional shading seems to be the more plausible explanation for the observed TIA offsets than effects of hillslope wash with the slope orientation. “
- „Instead, a combination of wind-driven redistribution and directional shading, with their relative importance varying along the tree degradation gradient, better matches with the observed redistribution dynamics. “
20- The limitation of ERA5 wind data is correctly identified, but the magnitude of the resulting error or bias is not assessed.
Together with the additions suggested under comment 6 in the method section we propose the following additions in section 4.3 line 391 to address this point:
- “Consequently, the true wind conditions at the individual test sites could strongly deviate from the considered large scale patterns. This might explain the lack of statistical significance for some of our tests. To address this issue long term wind measurements on the individual test site would be required.”
21- Future studies should integrate higher-resolution, site-specific wind measurements and true 3D canopy structure models, combined with sensitivity analyses, to better quantify and separate the effects of wind and shading on tree-influenced area displacement.
We definitely agree with you in this regard, propose an inclusion of soil sampling to investigate existing links to the nutrient dynamics and hope that our study can help to facilitate this development by acting as a reference to future efforts.
-
AC1: 'Reply on RC1', Lars Engelmann, 07 Jun 2026
reply
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 163 | 49 | 19 | 231 | 22 | 24 |
- HTML: 163
- PDF: 49
- XML: 19
- Total: 231
- BibTeX: 22
- EndNote: 24
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
This manuscript presents an interesting and valuable contribution to understanding fertile island dynamics in Argane ecosystems through the integration of UAS imagery, photogrammetric analysis, and spatial modelling approaches. The study is methodologically comprehensive and addresses an important ecological issue in dryland environments. The following points are suggested to improve the quality, clarity, and scientific robustness of the manuscript:
1- In figure 2 the Argane forest distribution is not clearly visible; please improve its visualization by using a contrasting color or adding a clear boundary outline to enhance spatial interpretation.
The study provides regional location and site names, but exact coordinates of the individual test plots are not reported in the main text, which may limit full spatial reproducibility of the sampling design.
2- At the international conference on the Argane tree, it was decided to standardize the name as Argane (with a capital “A” and an added “e” at the end) to better reflect the correct Amazigh pronunciation.
3- In Line 140: ground-only orthophoto shows the appearance of the terrain only” → repetitive (“only” repeated).
4-The manuscript does not report key flight parameters such as forward (longitudinal) and side overlap of images.
5- Correct this: optimal conditions were given for visual assessment→ better: conditions were optimal for ground visibility and image interpretation// almost complete absence of shadowing→ better: minimal shadow effects.
6- How can ERA5 wind data at a coarse spatial resolution (0.25° grid) reliably represent local-scale wind conditions within each study area, and how do you justify assigning a single grid point to characterize microclimatic variability relevant to ecological patch-scale analysis, given the potential for spatial simplification and mismatch with actual local conditions?
7- The coordinates are presented in a mixed format (decimal degrees with N/W directional labels), which is not strictly correct for decimal degree notation; in DD format, direction letters should be removed and West longitudes expressed as negative values (like: 30.0743, -9.14904).
8- The term “offsets of these centroids” is not clearly defined; it is unclear whether this refers to Euclidean distance, directional displacement, or a vector from one centroid to another (e.g., CA → TIA or TIA → CA), which reduces the reproducibility and clarity of the spatial vector analysis.
9- The interpretation of results is sometimes mixed into the Results (e.g., explanations of degradation and architecture effects), which should be reserved for the Discussion section.
10- The strong correlation (R² = 0.92) between CA and TIA is reported, but no check for mathematical dependency or circular reasoning is provided.
11- The statement “CA are generally smaller than TIA” is obvious and not sufficiently informative, given that TIA is defined to include CA.
12- The term “correlated (R² = 0.92)” is incorrect; R² should be used to describe a relationship (regression), not correlation (which requires r).
13- The reporting of statistical results is incomplete, as key details such as exact p-values, test statistics (W, H, Z values), and effect sizes are not provided, limiting the interpretability and reproducibility of the analyses.
14- The term “relationship” is used inconsistently across results analyses (correlation, regression, displacement), which may cause statistical confusion.
15- The overlap analysis results are reported, but effect sizes are not clearly quantified, and the practical ecological significance of the observed differences is not discussed.
16- The ecological interpretation of results is sometimes over-generalized, especially when linking surface patterns directly to subsurface nutrient processes without direct measurements.
17- Some claims about ecological processes (fertile island strength, erosion risk, system stability) are not fully supported by the presented quantitative results.
18- The role of slope is inconsistently interpreted (dismissed as non-significant but still suggested to have indirect influence), which weakens logical consistency.
19- The statement that wind and shading are the “most plausible explanation” is not statistically demonstrated, but presented as a near-conclusion.
20- The limitation of ERA5 wind data is correctly identified, but the magnitude of the resulting error or bias is not assessed.
21- Future studies should integrate higher-resolution, site-specific wind measurements and true 3D canopy structure models, combined with sensitivity analyses, to better quantify and separate the effects of wind and shading on tree-influenced area displacement.