Desertification, Resilience and Re-greening in the African Sahel – A matter of the observation period?

Since the turn of the millennium various scientific publications have been discussing a re-greening of the Sahel after the 1980’s drought mainly based on coarse-resolution satellite data. However, own field studies suggest that the situation is far more complex and both paradigms, the “Encroaching Sahara” and the “Re10 greening Sahel”, need to be questioned. The article discusses the concepts of desertification, resilience and re-greening by addressing four main aspects: (i) the relevance of edaphic factors for a vegetation re-greening, (ii) importance of the selected observation period in the debate of Sahel greening or browning, and (iii) modifications in vegetation pattern as possible indicators for ecosystem changes (shift from originally diffuse to contracted vegetation patterns). 15 The data referred to in this article cover a time period of more than 150 years and include the author’s own research results from the early 1980s until today. A special emphasis, apart from field work data and remote sensing data, is laid on the historical documents. The key findings summarised at the end show the following: i) vegetation recovery predominantly depends on soil types; ii) when discussing Sahel greening vs. Sahel browning, the majority of research articles only focuses 20 on post-drought conditions. Taking pre-drought conditions (before 1980s) into account, however, is essential to fully understand the situation. Then botanical investigations and remote sensing based time series clearly show a substantial decline in woody species diversity and cover density compared to pre-drought conditions; iii) selforganised patchiness of vegetation is considered to be an important indicator for ecosystem changes. 25

More than 90% of the entire Sahelian population which now amounts for approx. 12017 million people, (DSW, 20176) depend on wood and charcoal (Krings, 2006;IEA, 2014;. Monitoring of its spatial distribution dynamics can provide useful information for decision makers and early warning systems. A selected observation period of thirty years (post-drought) may lead to an evaluation of the Sahelian woody cover that is different from an observation period of 100 years which includes pre-drought conditions. When 20 evaluating Sahel greening vs. browning, the use of Earth observation (EO) tools is restricted to approx. 35 years (NOAA-AVHRR) or less (12 years, MODIS). Landsat archive offers an observation period of now 44 years, aerial photographs provide a historical view of approx. 65 years. The archive of meteorological data started in Senegal in the 1880s, in Niger around 1900 and in other Sahelian countries in the early 1910s or later. Botanical data have been available since 1900. Information on the state of the ecosystem prior to1900 can be extracted 25 from reports written by European travellers.
The focus of this paper is on three main aspects: (i) How can natural resources, specifically wood resources, be assessed by using documents of different sources of more than 150 years? (ii) What conclusions can be drawn with respect to the still ongoing "greening" vs. "browning" discussion? (iii) Are there indicators for an ecosystem change?

30
The eco-climatic borders of the Sahel can be defined as the 100 ±50 mm isohyets in the North and 600 mm in the South. The Sahel stretches across Africa over 400 to 600 km wide and nearly 6000 km long covering an area of approximately 3 million km² (Le Houérou, 1989). There are two major mechanisms defining the Sahel zone. The amount of rainfall is one criterion and this comprises a North/South shift according to a surplus or deficit of precipitation. Sahelian rainfall is notoriously unreliable and is characterised by strong inter-annual variability okra, mango). The southern Sahel has potential for crops and livestock husbandry and are characterised by a mixture of Sahelian and Sudanian species (e.g. Breman and Kessler, 1995;Von Maydell, 1986;Le Houérou, 1989;Schulz and Pommel, 1992).
The 100 mm isohyet as the approximate line between the Sahara and the Sahel roughly corresponds with the borderline between contracted and scattered vegetation, defined by Monod (1954). Contracted vegetation ("mode contracté") indicates that vegetation is concentrated in depressions and water courses. This pattern is characteristic for arid ecosystems. Scattered vegetation or "mode diffuse" refers to a more continuous vegetation cover on different soil types and is representative for savanna systems (Monod, 1954). Nicholson et al. (2012) provide a semi quantitative precipitation dataset for the nineteenth century adding these data to the more modern gauge data. According to their findings a severe and long lasting drought could be documented for the beginning of 19ths century, followed by a moderate recovery in rainfall mixed with some dry years. The 20ths century is well documented. Four drought periods (1908-1914; in the 1940s, beginning of 1970s and 1980s) and one humid period (1950s) can be distinguished (e.g. Nicholson, 1981Nicholson, , 1989Reichelt, 1987;Druyan, 1989;Mainguet, 1991;Nicholson et al., 1998).
Following the drought period of the early 1980s, a slight recovery in rainfall has been observed (Nicholson, 2005). This recovery is limited and still lower than the 1950-1989 average (Kusserow and Oestreich, 1998). Changes in the characteristics of the rainfall regime have additionally been observed. There is less spatial coherence and less temporal persistence. The contrast between a dryer western Sahel and a wetter eastern Sahel 15 is becoming more significant (Lebel and Ali, 2009;Nicholson, 2013). Sanogo et al. (2015) found a statistically significant positive rainfall trend between 1980 and 2010, however, not reaching the degree of wetness of the 1950s.
During late Quaternary the Sahara advanced to the South several times showing the largest expansion during late Pleistocene some 16,000/17,000 years ago (Ogolian desert) and retreated again (Reichelt et al., 1992). For 20 the last millennium the authors found a southward shift of isohyets by 25-30 km per century.
The human impact in the Sahel region started about 7000 years ago (Schulz and Pommel, 1992). The authors discuss an anthropogenic formation of the Sahel from 4000 BP on as a result of cattle keeping, and small holdings with traditional agrarian systems. These small-scale farming consisted of exploitations of fruit trees and field crops in park systems as well as energy supply and metal production using wood resources. Principal 25 instrument for clearing was and still is fire. The transformation of the landscape resulted in a creation of a savanna system like the present Sahel, evolved from the Holocene transition of Sudanian to Saharian vegetation. Large parts of the western Sahel countries have formed part of big empires since around 800 B.C. (Krings, 1982(Krings, , 2006Ki Zerbo, 1992;Devisse and Vernet, 1993;Kusserow, 1994Kusserow, , 1995Hofbauer, 2013). European travellers like Mungo Park and Oskar Lenz (Hoffmann, 1799;Lenz, 1892) reported cultivation of maize in the area of 30 today's Canal du Sahel where nowadays even the cultivation of millet is problematic (Kusserow, 1995). The references of a more humid period had changed in the second half of the last century. Since the 1970s drought period the Sahel stands for desertification and increasing poverty (UNCOD, 1977).
To assess the vegetation dynamics in the Sahel, the author developed the following research methods. A satellite based woody vegetation interpretation key for semi-arid Mali was established in 1985Mali was established in /1987Mali was established in (enhanced 35 version in 1992Kusserow, 1986Kusserow, , 1994. The interpretation key consists of structural and floristic criteria of the ligneous vegetation cover combined with morpho-pedological characteristics. Applying visual interpretation techniques the method allows the mapping of Sahelian woody vegetation as well as a distinction of fields and fallow land. Dry season Landsat data were used to better discriminate woody from herbaceous vegetation cover. GIS techniques and on-screen digitizing (available since the end of the 1990s) were applied for research projects 40 in Niger, Mauretania, Chad and Darfur/Sudan (Kusserow, 2001(Kusserow, , 2002(Kusserow, , a, b, 2005(Kusserow, , 2014. Based on these techniques, changes in vegetation pattern and density were quantified and mapped in form of change detection maps, showing "winners" (predominantly agriculture) and "losers" (vegetation) as basis for planners and decision makers. The method was established in 1999 and has been further developed from 2001 on (Kusserow, 2001;Kirsch-Jung and Kusserow, 2002;Kusserow, 2014).


Correction of six line effect for the MSS data.  Master scene from 1991 was relatively corrected (haze correction).  Geometric correction was conducted on the basis of the topographic map (UTM 31/WGS84) by using 17 way points (scene subset: 70 km x 50 km).  MSS scenes was geometrically corrected on the basis of the master scene (TM).
 Relative calibration of the three datasets (radiometric correction) was performed by look up table modification (calibration on the basis of two test sites, showing no temporal variation). Note: vegetation classification was done using visual interpretation techniques based on detailed in-situ knowledge.  (i) EO tools: To assess the vegetation dynamics in the Sahel, the author developed the following research methods. A satellite based woody vegetation interpretation key for semi-arid Mali was established in 1985Mali was established in /1987Mali was established in (enhanced version in 1992Kusserow, 1986Kusserow, , 1994. The interpretation key consists of structural and floristic criteria of the ligneous vegetation cover combined with morpho-pedological characteristics. Applying visual 25 interpretation techniques the method allows the mapping of Sahelian woody vegetation as well as a distinction of fields and fallow land. Dry season Landsat data were used to better discriminate woody from herbaceous vegetation cover. GIS techniques and on-screen digitizing (available since the end of the 1990s) were applied for research projects in Niger, Mauretania, Chad and Darfur/Sudan (Kusserow, 2001(Kusserow, , 2002(Kusserow, , a, b, 2005(Kusserow, , 2014. Based on these techniques, changes in vegetation pattern and density were quantified and mapped in form of change 30 detection maps, showing "winners" (predominantly agriculture) and "losers" (vegetation) as basis for planners and decision makers. The method was established in 1999 and has been further developed from 2001 on (Kusserow, 2001;Kirsch-Jung and Kusserow, 2002;Kusserow, 2010;Kusserow, 2014).
Two different Landsat satellite data sets were referred to in this article. The first data set includes Landsat MSS, TM, ETM+ and OLI data and was ordered for free from the United States Geological Survey through 35 Earth Explorer (http://earthexplorer.usgs.gov/). These data are already pre-processed and systematically corrected.  (Example Mali, dates: 1976, 1985. The raw data were bought from US Geological survey in 1985 and 1991 and were processed in 1991 using ERDAS 7.4.1 and 7.5 (Kusserow, 1990(Kusserow, , 1994(Kusserow, , 1995. RGB = 4-2-1 In addition to Landsat data, change detection assessments were performed, using aerial photographs from the 1950s and 1970s (Kusserow, 1994) and aerial photographs and kite photographs from the 1970s and 1990s (Kusserow and Haenisch, 1999).
(ii) Botanical in situ measurements and further investigations: The author carried out botanical inventories in 1985, 1991 and 1992 in the Canal du Sahel area in Mali. The inventory included measurements of all woody 5 individuals in a 0,1 ha test plots (height, diameter at ground and breast height and crown diameter) as such providing detailed information about vegetation composition and density in the area (Kusserow, 1986(Kusserow, , 1994(Kusserow, , 1995. Further investigations included multitemporal analyses of soil algae crusts (Hahn and Kusserow, 1998) and molecular genetic studies of wild Sahelian forages  as well as analyses of rainfall data (Kusserow and Oestreich, 1998).

10
(iii) Extended ground truth since 1985: The term "ground truth" includes detailed field checks of preliminary satellite image interpretation, in particular documentation of landscape elements, monitoring of woody species and species constitution as well as different types of land use. All observations were documented by photos and GPS coordinates (before 1993: classical approach with topographic maps and notation of driven kilometres). Transect observations included detailed notes, photos and GPS coordinates during the field surveys.

15
Attention will be drawn especially to the so-called "Darfur Project", started in 2010. This still ongoing research project is aimed to prepare a multi-layered dynamic state of the art natural resources and land use database (NRDB) for Darfur/Sudan. Within this international project, funded by the Sudanese government and carried out by the Munich-based Gesellschaft fuer Angewandte Fernerkundung (GAF) AG, six major topics or layers, including geology and hydrology, geomorphology/soil, socio-economics, eco-biology and land cover/land use 20 forms part of the investigation. The author is responsible for the last two layers. Targeted to elaborate a comprehensive report on the ecology and land use in Darfur/Sudan and to prepare a database and maps (1:250,000), both layers shall provide information for planners and decision makers. The projects tasks were:  Satellite based interpretation of vegetation and land use classes, scale1:250 000 (2010, 2000, 1970s) and analyses of ecological changes.

25
 Satellite based interpretation of vegetation and land use classes for selected areas, scale1 :50 000 (2009-2011).  Review of documents (1950s until 2015).  Field survey (measurements, observations, questionnaires). ESRI's ARC GIS 9.2 software was used for change detection analyses and map production. Satellite data 30 interpretation (on-screen digitizing) was performed using a tailor-made software (Georover) designed by GAF company. For the plant specimen inventory a systematic sampling scheme was developed. Based on soilvegetation units derived from satellite imageries and the GPS coordinates, sample sites (size of 20 x 20 m for woody vegetation and 1 x 1 m for herbaceous vegetation) were established to quantify the distribution and relative abundance of plant species. Within the sample sites the following parameters were 35 measured/documented: trees/shrubs maximum height, stem diameter, crown diameter and species status (threatened, rare, common etc.). Despite insecurity in some areas, a total of 665 sample plots which accounts to 85% of the originally selected plots could be successfully measured.
Additional information on natural resources conditions was gained through interviews as part of the socioeconomic investigations. Four communities in four states were targeted (urban, rural, nomads and the internally 40 displaced persons [IDPs]). In total 102 communities and 2,547 households in Darfur had been interviewed during May to October 2014. Group discussion (12-20 people) was used as method for data collection. The interviewed persons were between 30 and 50 years old. Moreover, wild life investigations along transects and land use surveys were carried out.   in the Sahel (e.g. Tucker et al. 1991Tucker et al. , 1999Herrmann et al., 2005;Dardel et al., 2014b;Knauer et al., 2014). Global Inventory Modeling and Mapping Studies (GIMMS) with a very coarse spatial resolution (5-8 km, Mbow et al., 2015 are applied (Sect. 4.1 and 4.3).
New satellite derived imageries such as the new generation GIMMS-3g with a coarse resolution factor of 8 km; MODIS (Moderate Resolution Imaging Spectroradiometer) available since 2000 with a spectral resolution of 250 m and "SPOT-Vegetation (VGT) with a resolution of 5 km and 1km (since 1999) are used more recently 5 (Anyamba et al., 2014;Brandt et al., 2014a;Mbow et al., 2014;Rasmussen et al., 2014). Fensholt et al. (2004Fensholt et al. ( , 2015 and Brandt et al. (2014a, d) used datasets from MODIS, Geoland GEOV1 (5 km resolution) and GIMMS3g (8 km resolution) FAPAR (Fraction of absorbed photosynthetically active radiation) to assess local vegetation trends in Senegal and Mali. Horion et al. (2014) explored how dry season NDVI min can be used as proxy indicator for assessing changes in tree cover density.
Recent studies include biophysical variables like FAPAR and LAI (leaf area index), seasonal vegetation dynamics and land surface phenology (Ivits et al., 2013;Brandt et al., 2014a, d;Fensholt et al., 2015;Gessner et al., 2015;Diouf et al., 2015Diouf et al., , 2016. Brandt et al. (2016a, b) apply a phenology-driven model for estimating woody canopy cover in the Sahel at 1/0.5 km resolution scale on the basis of MODIS and SPOT-Vegetation FAPAR data.

30
The desertification debate started in the 1970s, caused by attracting attention in the scientific community as a result of the severe drought period in the early 1970s. The first international conference on environmental issues was held in Nairobi in 1977 (UNCOD, 1977). The most frequently-used definition of desertification is defined by the UN Convention to combat Desertification (UNCCD, 1994): "Desertification is land degradation in arid, semi-arid and dry sub-humid areas resulting from various factors, including climatic variations and human activities". Mainguet (1999) stressed, that any concern to define the term "desertification" will end up in ambiguity. The main question -how to discern an irreversible state of land degradation and degraded levels which are partly reversible -remains still open. Also Rasmussen (1999) pointed to a weakness of the definition and use of concepts. Prince (2016) compared global maps of land degradation and desertification and concluded the absence of reliable maps or means of desertification monitoring ("what is degraded?", "where does it occur?", "how severe is the degradation?").
The concept of an "Encroaching Sahara" came up during the early 1920s when first European scientists visited the region (Bovill, 1921;Stebbing, 1935Stebbing, , 1938. They discussed a growing aridification and established the concept of a southward shift of the Sahara towards the savannas in the south. In the late 1930s, a British-French expedition assessed the "Encroaching Sahara" concept in Niger and found that the vegetation cover has recovered and the tree cover in particular was in a very good status (Jones, 1938). Stebbing (1935Stebbing ( , 1938 and others interpreted the post-drought situation as a desert encroachment thus misleading future scientists. The first drought period in the past century lasted from 1909-1915(Nicholson, 2012 but the early researchers neglected, 50 that the data for such a short time span would rather indicate only climatic fluctuations instead of real climatic crisis (Mainguet, 1991).
Actually, a set of indicators and various sources of information are necessary to assess such complex phenomenon as ecosystem fluctuation. Besides remote sensing data (satellite data, aerial photographs) other documents like botanical surveys from the first part of 20ths century, rainfall data, maps, historical documents, 20 reports and questionnaires should be used for assessing land dynamics. Higginbottom and Symeonakis (2014) who reviewed more than 150 article regarding assessments of degradation, called for a "multi-faceted methodology". The longer the observation period is, the more sound will be the information for identification of long-term degradation processes (Miehe et al., 2010).
The question of decreasing or increasing woody cover is fundamental for people's livelihood. Particularly the 25 resource "wood" as main energy supply plays a key role in ensuring the survival of the local people and curbing emigration. Declining wood resources aggravate the already critical situation in the Sahel states (Ouedraogo et al., 2010). The discussion of a system's ability to recover after drought is a key focus in the desertification/re-greening debate. As learned from scientific literature of the early 1920s and 1930s, the vegetation cover in the Sahel-Sudan ecozone has been recovered from the severe drought period in the beginning of the 20 th century. Own contributions: During numerous field studies in Mali, Burkina Faso and Niger in the 1990s, the author documented and investigated the phenomenon of crusted soils in situ and on the basis of satellite-based vegetation analyses (Kusserow, 1995, 2014, Hahn and Kusserow, 1998, Kusserow and Haenisch, 1999. Based

45
The situation in the field is shown in Fig. 3 (a and b), demonstrating different abilities of vegetation recovery as a function of soil type and protection.  The left photo (a) taken from the author at the site in Toukounous, Central Niger, show the ability of the system to recover from drought periods on sandy soils and -this should be emphasised -under protection and Formatiert: Englisch (Großbritannien) controlled grazing. This site forms part of a national cattle breeding station in the 1990s. Outside of the protected area highly grazed dunes are visible. The grass height is less than 5 cm. Brandt et al. (2014c) also refer to the importance of soil properties in the context of ecosystem resilience.
The right picture (b) shows a largely degraded landscape in Mali (Canal du Sahel area). The fertile upper soil layer had been removed by wind and water activity, resulting in soil crust formations. Dead branches still fix a 5 sandy layer. If branches are collected by local population the small sandy residuals will be also blown out. Resource protecting measures and management are the only alternative for recovery. Brandt et al. (2014a, b) and Spiekermann et al. (2015) present comparable situations in Senegal and other parts of Mali (Bandiagara).
Dune systems of late Quaternary age are one of the major land types in the African Sahel and have high importance as one of the main agricultural regions in the Sahel zone. Showing a predominant ENE-WSW orientation, they are extensively cultivated and referred to as "Erg Ogolièn" in the western Sahel and "Qoz" in the eastern Sahel (Le Houérou, 1989;. Satellite images from Mauritania, Mali, Niger, Burkina Faso, Chad and Darfur document the key significance of Quaternary dune systems for rural livelihood (Kusserow, 2014).
The Mare d'Oursi site in Burkina Faso, well known as the "Oursi-dune", and first analysed by Toutain and de

15
Wispelaere (1978), is a good example to discuss resilience, re-greening as well as desertification. De Wispelaere (1990) documented an increase of un-vegetated areas on the Palaeolithic dune systems north of the Mare on the basis of aerial photographs from 1955, 1974-1976 and 1981 and interpreted this development as desertification.
Landsat system's archive with high resolution satellite imageries (<1982: 80 m, 1982-2012/13: 30 m, >2013: 30 m/15 m) offers the possibility of change detection analysis. Images from the early 1970s provide information 20 on pre-drought conditions. Woody cover reflects the situation of the Sahelian ecosystem at the beginning of the drought and the end of the 1950/60s more humid phase (Kusserow, 1986(Kusserow, , 1995(Kusserow, , 2014. Own   (Toutain and de Wispelaere, 1978), is now completely denuded. This banding pattern of actual bare areas stretches over approx. 600-700 km until western Niger (example from western Niger in Sect. 4.46, Fig. 15).
The Landsat imagery recorded on 4 July 1973 (Landsat MSS, RGB=421) shows the impacts of the 1970s drought period. The Palaeolithic dune system has a very limited vegetation cover due to low rainfall. The vegetation distribution on the Precambrian basement (see yellow boxellipsoid and yellow circle) is still in a 40 homogeneous pattern state. The light purple colour is typical for the predominant woody vegetation cover (see also example Niger) during the drought year 1973.
The comparison of both images indicates an apparent inversion. The dune systems in 1973 appear only sparsely covered by vegetation whereas the Precambrian basement shows a more homogeneous vegetation cover. A contrary situation is visible in 2013. The former tiger bush covered basement now appears with a clearly fragmented vegetation pattern (see yellow box Sect. 6) the eastern and northern parts are already bare. Due to higher rainfall the dunes show a good vegetation cover and extensive fields. These sandy deposits are extensively cultivated with millet. It is worth noting that the fragmented vegetation patches are is located in the valleys (see reference image from 1973). Due to higher water availability, caused by increased run off from already crusted higher areas and due to an accumulation of fine soil particles, vegetation is increasing in these parts. Own change detection assessments in the Canal du Sahel area in Mali (aerial photos, dated 1953 and high resolution SPOT satellite data of 1992), documented an increase in run off due to losses in woody cover, triggered by clearing for agriculture and fuel wood (Kusserow, 1994).
The results of the satellite image analysis can be backed up with the extensive work of Toutain and De Wispelaere (1978). Maps of the region document a dense to open savanna vegetation cover. Additional information regarding vegetation cover is given by the topographic map Hombori, (Feuille ND-30-NE, IGN Paris 1961; see Fig. 5). There, the dune systems close to Oursi village is still covered by either savanna or grass savannas (prairie). Major parts of Precambrian basement show pattern vegetation (tiger bush). All maps of West 5 Africa printed by IGN in the early 1960s are based on analyses of aerial photographs (flight periods were the early 1950s). It is worth mentioning that the topographic maps document a great extent of tiger bush pattern as well as savanna and grass savanna vegetation distribution even until 16° N. In contrast to the situation in the 1950s-1970s, the current image shows bare ground with marked vegetation patchesterns.  Fig. 6). According to a photo taken January 1977, a dense bush formation with a single large Adansonia digitata can still be found there at that time. Krings (1980) mentioned also extended 20 vegetation losses in the tree savanna region in northern Oudalan.  Gourma, Mali, status 1976-1977, modified after Krings (1980. The box shows the position of the satellite image section (see Fig. 4).

25
The preceding investigations clearly show a re-greening on sandy soils (Quaternary dune systems) but a browning on poorly developed soils of the underlying Precambrian basement. This indicates the ecosystem's ability to recover almost exclusively on Quaternary dunes and not on shallow soils. Mensching (1990) presented a map indicating the occurrence of late Quaternary dune systems throughout the Sahel (Fig. 7). The picture is incomplete since dune systems in Eastern Chad and western Sudan (Darfur) seem to be underrepresented.
However, this map gives a first overview of areas with potential for vegetation recovery and areas with other soil types possibly indicating less resilience capability.  Mensching (1990).

35
Contributions of other authors: Many authors documented a re-greening trend on sandy soils and discussed the ability of post drought regeneration as a function of soil type and topographic position (Hiernaux et al., 2009aVincke et al., 2010;Brandt et al., 2014aBrandt et al., , b, c, 2015Dardel et al., 2014b;Rasmussen et al., 2014). All study areas cited in this article and reporting a re-greening trend, as Senegal (e.g. Herrmann and Tappan, 2013;Brandt et al. 2014aBrandt et al. , b, 2015Herrmann and Sop, 2016), Mali/Burkina Faso (Hiernaux et al., 2009a, b;Dardel et 40 al., 2014a, b;Rasmussen et al., 2014;Brandt et al., 2016b) and Niger (Hiernaux et al., 2009a, b;Boubacar, 2016) are situated on the sandy soils of the Quaternary dune systems. For the "Oursi-dune" example, Rasmussen (1999) found an increase in vegetation (bushes and herbs) by comparing aerial photos from 1955, 1974, 1981 and 1996. Regarding the aspect of topographic position, Kaptué et al. (2015) confirmed an increasing woody cover at the watershed scale in the majority of their samples. The drier more northerly Sahelian watersheds in Senegal and eastern Mali appear to show stronger reforestation trends than the more mesic region of western Mali and high human population area near Niamey, Niger. Mainguet (1991) has already pointed out that due to increasing run off towards the valleys vegetation re-growth is improved. The importance of different landscape elements for re-greening processes is also mentioned by Vincke et al. (2010) and Rassmussen et al. (2014). The latter two authors found negative NDVI-pixels on the plateaus and slopes and positive pixels in the valleys. Vincke et al. (2010) reported a distinctive regression of woody vegetation in the high relief areas. Own research studies for western Niger (Kusserow and Haenisch, 1999;Kusserow, 2010Kusserow, , 2014 documented severe woody vegetation losses on lateritic plateaus in southwest Niger. To conclude: recovery of vegetation widely depends on morpho-pedological factors. A recovery on sandy soils have been often documented (e.g. Hiernaux et al., 2009a), whereas on poorly developed soils and crusted soils less or no regeneration can be found (Hiernaux et al., 2009a;Brandt et al., 2014aBrandt et al., , b, 2015Dardel et al., 2014b;Rasmussen et al., 2014). Hiernaux et al. (2016) concluded for the Gourma region in Mali a strong resilience on sandy soils but a collapse and profound mutation of the vegetation on shallow soils.

Plant species change
A different understanding of re-greening becomes evident when the viewing angle is extended to pre-drought condition. Comparing the period 1980-2015 with earlier years, the so called "re-greening" can be quoted as "dramatic decline" in vegetation due to a much higher floristic composition and vegetation density prior to the 1980s. For West Africa, botanical investigations had been conducted by e.g. Chevalier, 1900;Furon, 1929;Trochain, 1940;Roberty, 1946;Aubrèville, 1949;Monod, 1954;Toutain and De Wispelaere, 1978 (see Fig. 8). For East Africa (Sudan) botanical studies from e.g. Andrew, 1950;Harrison and Jackson, 1958;Ramsay, 1958;Lebon, 1965;Wickens, 1976;Ibrahim, 1980;Miehe, 1988 and various Hunting Technical Reports from the 1950s till the mid of 1990s are available. Fig. 8 presents a compilation of selected authors.  area in Mali. The inventory included measurements of all woody individuals in a 0,1 ha test plots (height, diameter at ground and breast height and crown diameter) as such providing detailed information about vegetation composition and density in the area (Kusserow, 1986(Kusserow, , 1994(Kusserow, , 1995. These data were compared with botanical inventories, realised by Roberty (1946) in the 1940s. Results showed significant changes in species composition. Roberty (1946) still documented mesic (Sudanian) woody species like Terminalia avicennioides, Bombax costatum, Pterocarpus lucens, Sclerocarya birrea, Sterculia setigera. Own investigations show a clear shift in the range of species into more robust (Combretum glutinosum, Guiera senegalensis) and arid-tolerant ones as Commiphora africana. C. glutinosum and Guiera senegalensis (family Combretaceae) are typical invaders on fallow land. Still today large areas of the Sudano-Sahelian ecozones are characterised by these two species. Savannas dominated by Combretaceae seem to develop from dense woodland that has been subject to 30 intensive clearing and wood cutting (Trochain, 1940;Aubréville, 1944Aubréville, , 1949Aubréville, , 1950Le Houérou, 1989). Figure 9 shows a shift of isohyets comparing Climatological Normals (CLINO) 1931(CLINO) -1960(CLINO) with 1961(CLINO) -1990. During CLINO 1 the average rainfall varies between 500-600 mm in the research area, whereas CLINO 2 indicates a significant shift to 300-400 mm which represents a typically Sahelian climate. The average rainfall of 500-600 mm fits into the Sudano-Sahelian ecozone, allowing more mesic species to grow. The ligneous 35 fingerprint clearly indicates that more moisture demanding species with Sudanian provenience had been part of the woody population until approx. the early 1980s (Kusserow and Oestreich, 1998). The post drought conditions favour more drought tolerant species, leading to a selective die back of species (e.g. Pterocarpus lucens; Kusserow, 1995). This southward trend of annual rainfall crossing the 600 mm annual rainfall threshold for Sudanian flora (Le Houérou, 1989) and has a tipping point like ecological significance (Maranz, 2009;Kusserow, 2014).  1931-1960) and CLINO 2 (1961, position of research areas is indicated by the red box (modified after Kusserow and Oestreich, 1998).

45
One driver of the observed decrease in biodiversity is the lower level of precipitation rates which is/was responsible for the species turnover in the ligneous population towards more xeric species. The other driver is human activity (deforestation, clearing for fields). The widespread occurrence of Combretum glutinosum and Guiera senegalensis which according to own observations formtogether with a few other speciesthe prevailing ligneous cover in the Sudano-Sahelian ecozones across Africa. Both species are first pioneers on fallow land and clear indicators for former agricultural activity. In the Sahelian ecozone they are represented by Leptadenia pyrotechnica. The region between Zinder and Goure in eastern Niger are exclusively covered with this highly drought-resistant species. Leptadenia pyrotechnica has Saharian affinities (Le Houérou, 1989) and form almost pure stands in eastern Niger, indicating former agricultural areas. These areas experienced a long lasting human settlement history as part of the big empire Kanem-Borno (Krings, 1982).
A massive abundance of Leptadenia pyrotechnica has been observed during own field surveys in 2009 in the area of Tillabéri (Niger), as could not be seen in the mid of the 1990s. The same development was found around of El Fasher, North-Darfur, Sudan during own field surveys in 2014 as part of the Darfur project, executed by GAF-Munich/Germany and financed by the Sudanese government. The accompanying 60 years old forester, grown up in this area, confirmed millet fields and a much more diverse savanna vegetation as well as a rich fauna having been present in the 1960s. Leptadenia pyrotechnica as a typical representative of the eco-climatic "Sahara" and "Saharo-Sahelian" zones (Le Houérou 1989) and classified as ecological indicator species (Miehe et al., 2010) points to changing ecosystem conditions. Also Hiernaux et al. (2009a) reported an increase of Leptadenia pyrotechnica in the Gourma region in Mali. Rasmussen (1999) as well found an increase of Leptadenia pyrotechnica on the denuded part of the Oursi dunes in Burkina Faso. He confirmed species changes and observed new invaders after the almost eradication of several woody species following the drought periods of the 1970s and 1980s.
The Eastern Sahel shows significant species changes and a turnover into more drought tolerant species. First

20
documents clearly describe a mixture of Sudanian, Guinean (South Darfur) and Sahelian species (e.g. Harrison and Jackson, 1958;Ibrahim, 1984) in Darfur before the 1970s drought. During a field survey in 2014 mesic trees had only been observed in depressions and around ponds. In West-Darfur approx. 10-12 years old fallow land did not show any regrowth of mesic species. The most dominant trees/shrubs in fallows were Guiera senegalensis and Boscia senegalensis, both pioneers and typical representatives of succession states. Figure 10 25 shows residuals of Dalberghia melanoxylon (Sudanian species) in the already highly fragmented savanna in West Darfur. Development of biological soil crusts is a typical phenomenon in desertification processes (Hahn and Kusserow, 1998).

30
As one part of the still ongoing Darfur project additional information on natural resources conditions was gained through interviews. Four communities in four states were targeted (urban, rural, nomads and the internally displaced persons [IDPs]). In total 102 communities and 2,547 households in Darfur had been interviewed during May to October 2014. Group discussion (12-20 people) was used as method for data 35 collection. The interviewed persons were between 30 and 50 years old. The findings confirmed an increase of more drought tolerant species at the expense of Sudanian species and an increase in bush fallow pioneers like Guiera senegalensis.
Contributions of other authors: When they compared post-drought with pre-drought conditions, several authors observed a species change since the beginning of the millennium. This is in line with findings from 40 questionnaire-based surveys among elder people in the Sahel (e.g. Rasmussen et al., 2001;Gonzales et al., 2004Gonzales et al., , 2012Ouedraogo et al., 2010;Brandt et al., 2014c;Sambou et al., 2016) Herrmann and Tappan, 2013;Brandt et al., 2014bBrandt et al., , 2015Brandt et al., , 2016aSpiekermann et al., 2015). Kaptué et al. (2015) refers to a post drought tree population recovery and attested a decline in populations of economically and culturally important trees and shrubs despite the increase in woody cover. For Burkina Faso (Oursi-dune), Rasmussen (1999) noted that the species currently invading the live dunes -created in the 1970s -are not the same as those dominating before. Analyses of neighbouring Gourma region in Mali confirmed these findings (Hiernaux et al., 2009a). Rasmussen et al. (2001) documented a significant increase of Balanites aegyptiaca on the basis of group interview with Peulh pastoralists. Wezel (2004) observed a decline of economically important trees, as well as a decrease in species which in drought periods are absolutely essential for survival like Boscia senegalensis for sites in Burkina Faso, Niger and Senegal. Gonzales et al. (2004Gonzales et al. ( , 2012 investigated changes in forest species on the basis of interviews and field observations in 14 villages across the Sahel (5 states). The observation period included forty years .
They found a significant decrease in forest species richness and discussed a shift of Sahel, Sudan and Guinea vegetation zones. An increase in bush fallows with prevailing Guiera senegalensis and Combretum glutinosum had been observed in Bandiagara (Brandt et al., 2014a) and in Ferlo, Senegal (Vincke et al., 2010;Brandt et al., 2014a, Brandt et al., 2017a. The increase of ecological key indicator species (Miehe et al., 2010) like Balanites aegyptiaca, Acacia raddiana accounts for a fundamental change in the former Sudano-Sahelian ecozone. Trochain (1940) already discussed two forms of substitution of the original flora in Senegal: (1) "Paratype of substitution" Balanites aegyptiaca is dominant

20
(2) "Savanegarrique anthropozoogène" Combretum glutinosum is dominant An overview of plant species changes according to authors and region is given in Sect. 7 (Fig.18). The observed changes in species composition in the Sahel clearly indicate an impoverishment of an originally much more diverse flora. The parkland system which dominates large parts of the Sahelo-Sudanian zone is most likely of human origin (Schulz and Pommel, 1992;Maranz, 2009;Schulz et al., 2009). Parklands are fundamental for 25 livelihood in the Sahel-Sudan region. Low annual crop yields are often offset by fruit yields from trees maintained in parklands. As such they play an important role for local subsistence and enable the rural population to better overcome dry periods (Krings, 1991b;Maranz, 2009). The latter author found a remarkable decline in biodiversity including losses of species of Sudanian and Guinean provenience, associated with an increase of Sahelian species. He discussed the widely observed senescence and disappearance of mesic species 30 as a collapse of an anthropogenic system that is no longer adapted to increasing arid conditions due to ecologically critical rainfall shifts.
Ibrahim and Schulz (in press2017) investigated a sediment core from the Guidimouna-lake in SE Niger. Based on sedimentology, micromorphology and high-resoultion pollen diagrams, they were able to reconstruct the lake history during the last 90 years. The author discussed the 1970s drought as the key trigger for plant species 35 composition change favouring more drought tolerant species. They concluded, that the system could not recover completely yet, even regeneration of vegetation and soil is still present.
An additional indicator for the shifting ecosystem is wildlife. In Niger Giraffes, that were described to be present in the 1970s in the region of Tillabéri close to the border to Mali (Bernus and Hamidou, 1980), are now found in the area of Dosso southeast of the capital Niamey (own observations), thus indicating a southward shift 40 of approx. 150 km. Reichelt (pers. communication) reported the occurrence of giraffes and elephants even in the Gourma area in Mali at the end of the 1950s during his numerous geological field surveys (Reichelt, 1972).
It would be essential to continue field work based studies as mentioned by several authors (e.g. Miehe, 2010;Brandt et al, 2017a). However, the actual security situation in the Sahel especially in Mali, Burkina Faso and Nigerregions like Gourma in Mali, Oudalan in Burkina Faso and almost the entire Niger are insecure (Weiss, 2016)makes a reliable field survey planning impossible.
Positive developments are farmer-managed natural regeneration of selected trees on fields, reported for southeastern Niger (Larwanou and Saadou, 2011;Sendzimir et al., 2011;Boubacar, 2016;Herrmann and Sop, 2016). In a recent study Brandt et al. (2017 a) observed an increase of Pterocarpus lucens (sudanian element) in those parts of eastern Senegal that are characterised by less human pressure. The authors discussed low human impact 50 as a key factor for regeneration.

Re-greening Sahel based on NOAA-AVHRR, GIMMS3g, MODIS, SPOT VGT studies
One key argument in recent scientific papers dealing with the question of Sahelian re-greening is based on NDVI analysis of coarse resolution satellite images. Given a concrete example, the author discusses limitations, when using post-drought satellite data.
Own contributions based on a multi-line approach in the Canal du Sahel area (southern Mali, close to the Mauritanian border) show a re-greening trend for the research site when comparing Landsat data, recorded in 1985 and 1991. However a comparison of data, received in 1976 with 1991 show the opposite trend -a significant decline (Figs. 11 and 12, Kusserow, 1994). The MSS false colour images (vegetation is red) of 1985 displays nearly no vegetation signal apart from the rice plantations of the Canal du Sahel recognisable in the eastern part of the image (Fig. 11). However, the attached photograph of 1985 indicates dense woody vegetation 10 cover. Due to the drought period (1982)(1983)(1984) trees and shrubs were completely leafless. Many species in the area show different strategies to overcome the annually dry season without a lot of rain. The trees/shrubs either shed their leaves completely/partially (deciduous or semi-deciduous) or they are evergreen (Le Houérou, 1989).Typical aspects outside of drought periods are semi-deciduous coverage of ligneous vegetation, whereas the foliage degree depends on tree/bush species and conditions of the previous rainy season. Is a drought period severe (as it was during the early1980s), woody vegetation shed their leaves completely. It appears that woody vegetation cover have been died off. The 1991 TM imagery depicts a recovery in woody vegetation cover. However ground survey indicated that some tree species, particularly Sudanian ecozone species as Pterocarpus lucens have not survived (Kusserow, 1994). The NOAA-AVHRR sensors from that period may have also recorded dense woody vegetation cover without 20 identifying them as woody vegetation because of its leafless state. Due to the strong precipitation deficit the herbaceous cover was hardly present. One could therefore argue that the observed re-greening since the early 1980s was predominantly based on the increase in agricultural crops and herbaceous cover.

30
In fig. 12 "re-greening" (period 1985-1991, top row) is compared with "browning" (period 1976-1991, bottom row). A legend provides information to identify individual land units. The MSS imagery recorded 1976 (bottom left) shows a dense and uniform woody vegetation cover (recognisable in the western part of the image by its brownish-reddish colour). The black patches are actual burnings. Between both dominating landscape elements (uniform savanna and rice plantations) a small region with less or no vegetation can be discriminated. These areas are alluvial soils (light blue) and sandy areas/dunes (white-yellow, Kusserow 1994). Although the region had been experienced the drought period of the early 1970s, a uniform woody vegetation pattern still exists. The ecosystem resilience still seemed to be quite strong during the 1970s. The turning point showed up with the renewed and more severe drought period from1982-1984. This decade can be classified as the starting point of a 40 new ecosystem state.
Comparing the TM 1991 imagery with a MSS Landsat image from 1976 (bottom left see Fig. 12) clearly indicates a significant decrease in woody vegetation cover which points to is indicating a browning trend. Contributions of other authors: to monitor vegetation trends in the Sahel, researchers used vVarious versions of the NOAA-AVHRR (National Oceanic and Atmospheric Administration -Advanced Very High Resolution Radiometer) data have been used to monitor vegetation trends in the Sahel (e.g. Tucker et al. 1991Tucker et al. , 1999 greening trend in the Sahel since the early 1980s (e.g. Herrmann and Hutchinson, 2005;Herrmann et al., 2005;Olsson et al., 2005;Helldén and Tottrup, 2008;Hiernaux et al., 2009b;Nicholson et al., 2012;Anyamba et al., 2014;Fensholt et al., 2013, Brandt et al., 2016a, b, 2017a. New satellite derived imageries such as the new generation GIMMS-3g with a coarse resolution factor of 8 km; MODIS (Moderate Resolution Imaging Spectroradiometer) available since 2000 with a spectral resolution 5 of 250 m and "SPOT-Vegetation (VGT) with a resolution of 5 km and 1km (since 1999) are used more recently (Anyamba et al., 2014;Brandt et al., 2014a;Mbow et al., 2014;Rasmussen et al., 2014). Fensholt et al. (2004Fensholt et al. ( , 2015 and Brandt et al. (2014a, d)  are not sufficient enough or the degradation is not strong enough to be discernible by remote sensing methods (Miehe et al., 2010;Dardel et al., 2014b).  According to Brandt et al., (2014b) crusting surfaces are not detectable at a scale of 5 km and 25 sometimes not even at 250 m (MODIS).  Areas of farmer-managed natural regeneration in southern Niger, where field tree cover is said to have improve, do not stand out in satellite-derived greenness trends (Hermann and Sop, 2016). In their review article on greening trends in the Sahel/Sudan, Knauer et al. (2014) noted that the observed trend is obviously due to various causes and can be interpreted as improvement but also as degradation. Herrmann and Sop (2016) concluded that long time series of NDVI has proven insufficient for detecting the woody fraction in semi-arid environments. Bachmann et al. (2015) emphasized the importance of a consistent pre-processing and harmonisation of the generated AVHRR time series.
Increase in greenness could also be caused by agricultural crops and grasses (Gonzales et al., 2012;Brandt et al., 2014a;Dardel et al., 2014a. NDVI variability and trends are predominantly linked to 35 herbaceous cover dynamics. Due to the influence of peak rainy season (August and September) the radiometric response of a woody plant cover is hardly distinguishable from an herbaceous cover (Dardel et al., 2014a). The authors concluded that an increase/decrease of woody vegetation cover could not be detected when using NDVI data from peak season at 1 km scale or larger. Even more important is that potential changes in the woody vegetation cover are not easily linked to the overall re-greening trend, because the re-greening is mainly linked 40 with herbaceous and agricultural productivity . According to Bégué et al. (2011) who analysed NOAA-AVHRR for a period of 25 years, an increased cropping intensity is responsible for an increase in the annual NDVI for the Sahelian part of the Bani catchment area in Mali.

45
natural non-forest vegetation, agriculture and barren land in Sub-Sahara Africa (SSA) using high spatial resolution Earth observation satellites (Landsat). For the whole SSA the Sudanian eco-region shows the highest increase in agriculture (26%). The losses in non forest vegetation are also dominated by the Sudanian region (36%), followed by the Sahel region (29%). Compared to 1975, the Sudanian zone has also significant increases in barren lands (26%, Brink and Eva, 2009). Knauer et al. (2017) found that 91% of the agricultural area in Burkina Faso has expanded between 2001 and 2014. The expansion of agricultural land is also a result of the increasing population -Sahel countries have the worldwide highest population growth ratesand aggravate not only the strain on natural resources but also the already existing conflicts between agriculturalists and pastoralists (Müller et al., 2011;Brücher et al., 2015). With the expected increase of the Sahelian population from about 1342 million today to 197200 million in the year 2030 and 32645 million in 2050 (DWD, 20176), food demand, expansion of agricultural areas and demand of wood resources (fire wood) will increase dramatically.
Recent studies include biophysical variables like FAPAR and LAI (leaf area index), seasonal vegetation dynamics and land surface phenology (Ivits et al., 2013;Brandt et al., 2014a, d;Fensholt et al., 2015;Gessner et al., 2015;Diouf et al., 2015Diouf et al., , 2016. Brandt et al. (2016a, b) apply a phenology driven model to estimate woody canopy cover in the Sahel at 1/0.5 km resolution scale on the basis of MODIS and SPOT-Vegetation FAPAR data. Based on several assumptions, their extrapolated woody cover map for the Sahel shows a site specific trend with areas documenting a positive development and areas with vegetation losses. A phenology-driven model for estimating woody canopy cover in the Sahel at 1/0.5 km resolution scale on the basis of MODIS and SPOT-Vegetation FAPAR data were applied by Brandt et al. (2016a, b). Based on several assumptions, their extrapolated woody cover map for the Sahel shows a site specific trend with areas documenting a positive development and areas with vegetation losses. The authors concluded an overall positive trend in woody cover, emphasising the resilience of the ecosystem. In a further study carried out by Brandt et al. (2017a)

20
2017, who investigated the coupling between NDVI trends and cropland changes for a test site in western Niger (Fakara). The authors found positive NDVI trends based on more frequent fallow years and a negative NDVI trend associated with an increase in cropped fields. Mbow et al. (2013) stressed that major changes in plant species dominance should be taken into account when analysing NDVI time series.

Changes in vegetation pattern, self-organised patchiness
Own contributions: Significant changes in the spatial distribution of woody vegetation were first observed when using multitemporal remote sensing data from the early 1950s to 1992 (aerial photographs, Landsat, SPOT) for 30 vegetation monitoring in Mali (Kusserow, 1990(Kusserow, , 1994. Within 40 years the originally homogeneous vegetation pattern, still recognisable in the 1950s (example Mali, Canal du Sahel region) has turned into a highly fragmented pattern with spots and isolated bands. The main trigger for this development was found to be human impact (clearing and wood cutting). Derived from these remote sensing based observations, a principal model (Fig.13) for the development of woody vegetation patterns towards desertification was developed (Kusserow, 35 1994).

Figure 13.
Left: Changes in the spatial vegetation distribution from North to South: Sahara -South Sahara/North-Sahelian transition zone -Sahel (modified according to Kusserow 1994). Right: Chronology of changes in spatial vegetation distribution in the south, i.e. Sahel (modified according to Kusserow 1994).

40
The changes in the spatial woody vegetation distribution from North to South (Fig. 13, left) correspond with the rainfall gradient from the arid Sahara to the more humid areas in the south, i.e. from the desert ecosystem to the savanna ecosystem. The same pattern changes but in reverse order were observed in savanna areas (Sahel) when analysing aerial photographs and satellite data from the 1950s/1970s until today (Fig.13, right). Referring

45
to Monod (1954) who defined the approximate borderline between Sahara and Sahel as one between contracted (Fig.13 upper left picture) and scattered-diffuse vegetation types (Fig. 13 bottom left), the author discussed these observed changes as a principal indicator for desertification.
Gaps, stripes and spots as detectable in the 1975 aerial photograph had been later postulated by several model studies (e.g. Lejeune et al., 2002;Dekker et al., 2007;Gilad et al., 2007;Meron, 2012). On the basis of freely accessible Landsat series of the Canal du Sahel region, the development from 1976 until 2010 could be continued. The comparison of two Landsat imageries (size of subset is approx. 80 km x 55 km) recorded on 26 February and in different years (1976 and 2010), demonstrates a transition from an originally uniform woody vegetation pattern into a banded and spotted distribution within a time span of 34 years (Fig.14). A legend helps to identify individual land units and principal land cover changes. Main emphasis is laid on the process of vegetation pattern development. The false colour images of 1976 still shows a dense and uniform savanna vegetation in the western part of the imagery. The corresponding image of 2010 documents significant changes: the originally dense and uniform savanna vegetation has disappeared and a fragmented, discontinuous woody  The actual burnings are restricted to the dune systems in 2010 (visible in the lower half and the uppermost part) whereas the upper half indicate no burning but significant vegetation pattern formation (western part of the imagery). As discussed further above, regeneration can be documented on sandy soils. The region between both 15 dune systems is characterised by shallow soil cover (Ferric luvisols, Di Bernardo et al., 1986) over Precambrian basement. These areas show none or only little agricultural activity. The shallow soil layers are extremely vulnerable to degradation and desertification processes (see Sect. 4.13). The development of woody vegetation pattern (contraction), clearly visible in the 2010 imagery, started exactly from these areas. Analyses of additional Landsat time series for the test site in Mali (not presented here) document the starting of fragmentation processes 20 (pattern formation) after the 1980s drought. The imagery recorded in 1991 does not yet show any significant structures, whereas at the image dated 1999 pattern formation has already started to form. This pattern is a lot more clearly identifiable from 2001 on and much more significant from 2010 on. Based on interpretations of images from the Landsat satellite image archive, the process of pattern development was estimated to have been completed within 10 to 15 years (example Mali).

25
The second example is located in western Niger, north of Ouallam and close to the border of Mali (Fig. 15). The size of the area is around 60 km x 45 km. Two main landscapes are recognisable: lateritic plateaus (dark green) and dune systems (yellow-white) partly covered by vegetation. The focus is on the process of woody vegetation pattern development (fragmentation/contraction). Identification of individual land units is supported by a legend. The image recorded on 30 September 1973 shows the impact of the 1970s drought period,

35
Uniform woody vegetation (violet-purple) is still dominant in 1973 (30 September). For the early 1960s the occurrence of bush and tree savannas was confirmed by IGN topographic maps (sheet Ouallam, République du Niger Feuille ND-31-XV, IGN Paris, March 1961). The second image was recorded 40 years later on 27 September 2013 and presents a higher rainfall situation indicated by a dense herbaceous vegetation cover and agricultural activity on the dunes. Although the precipitation dept is higher (4117 mm) compared to 1973, the vegetation cover appears in a reversed order, i.e. vegetated dunes are recognisable but bare areas and fragmented woody vegetation patterns had been developed on the argillaceous sandstones of the Continental Terminal (Greigert and Pougnet, 1965). Parts with bluish colours (former vegetated areas) are now representing degraded, mainly crusted soils. The woody vegetation patches observable in the satellite image series (Fig.15) emerged in the valley bottoms observations). The following Landsat satellite based time series depict show the forming of woody vegetation pattern (Fig. 16). Subsets of six satellite images show principal changes in vegetation distribution (please note that Landsat imageries, recorded 30 Sep 1973 and19 Sep 2016, are only shown as subset).
The imagery recorded 13 October 1984 already presents a significant pattern of stripes and spotted areas with some residuals relics of a more uniform distribution (rainfall amount in 1984: 16075 mm). Due to the drought 5 period of the early 1980s the dunes show less vegetation cover and hardly agricultural activity. The image recorded on 3 October 2002 (high rainfall with 4890 mm) shows that dunes are re-vegetated and agricultural activity (fields) is clearly recognisable. In the upper northern part no regeneration is detectable but pattern formation (stripes and spots) of woody vegetation is well visible. During a field trip in April 2001, farmers in the small village Tuizégourou complained about harsh environmental conditions and low agricultural productivity with increasing risk of crop failure. According to local people these areas were starting points for emigration. The region has a longstanding settlement history which dates back until the late Neolithic period. Devisse and Vernet (1993)

25
For the research side in West-Niger the pattern formation process was observed to have been occurred between 1973 and 1984 (two drought periods) which would be a very short formation period of 11 years. Pattern changes are triggered by increasing aridity and exacerbated by human impact. This development is mainly observed on shallow soils (see Fig. 14-16).
Based on time series of Landsat MSS/TM and SPOT satellite images, aerial photographs and kite photographs, Kusserow and Haenisch (1999) analysed the dynamics of a tiger bush site southeast of Niamey, Niger. They found that the banded patterns were formed out of an originally uniform state in the 1950s and interpreted vegetation stripes as a relic habitat or kind of biodiversity pool. Vegetation bands, soil sealing and crusting between bands form a surface layer protecting the seed bank (Hahn and Kusserow, 1998), thus constituting a crucial part of a natural in situ conservation strategy (Kusserow and Haenisch, 1999).

35
First results of an ongoing project in Darfur/Sudan show distinct vegetation pattern changes when comparing Landsat MSS data (spatial resolution 80 m) from the early 1970s and data from an Indian Microsatellite system (spatial resolution 37 m) recorded in 2010. Unlike the situation in the western Sahel, tiger bush areas could not be found, which is mainly due to a different geomorphology. Vegetation distribution changes can also be identified. The origin is similar: woody vegetation pattern formation has formed out of an originally uniform 40 vegetation cover still recognisable in the 1970s satellite imageries. For all own research areas presented above, a strong human impact had been identified as main driver of the pattern development. As already mentioned, the author discusses vegetation pattern formation as observed in satellite time series as a key indicator for desertification processes. The driving force is a feedback between drought and increasing human intervention, i.e. wood cutting and clearing for cropping.

45
Contributions of other authors: Special vegetation mosaics known as "tiger bush" (brousse tigrée) are common vegetation pattern in dry regions. These patterns consist of bushy stripes and arcs alternating with open, non vegetated areas that are often crusted, and situated on very gentle and uniform slopes. The type of pattern can vary between "spotted", "broadly" and "horizontally banded" (following the contours) (White, 1970;Hiernaux and Gérard, 1999;Valentin et al., 1999). Due to the striped appearance on aerial photographs Several authors described the main formation mechanism: due to a better water balance in the upper soil (generated by sheet run-off on the bare inter-bands), the self-modifying system of vegetation stripes offer more demanding species the possibility to survive in habitats with less rainfall (White, 1970;Cornet et al., 1992;. D'Herbès and  and Valentin et al. (1999) discussed the Niger tiger bush as a natural water harvesting system. According to their findings, the mean annual water infiltration into the thicket 5 cores of vegetation bands enables wood production similar to that of woodland and forest in the wet savanna zones and even exceeded forestry industrial plantations. Kusserow and Haenisch (1999) analysed the dynamics of a tiger bush site southeast of Niamey, Niger, on the basis of time series of Landsat MSS and TM and SPOT satellite images, aerial photographs and kite photographs. They found that the banded patterns were formed out of an originally uniform state in the 1950s and interpreted vegetation stripes as a relic habitat or kind of biodiversity pool. Vegetation bands, soil sealing and crusting between bands form a surface layer protecting the seed bank (Hahn and Kusserow, 1998), thus constituting a crucial part of a natural in situ conservation strategy (Kusserow and Haenisch, 1999). Thiéry et al. (1995) discussed that the two common hypothesesdegradation of an initially uniform pattern or colonisation of previously bare zonesare two aspects of the same phenomenon.
These field and satellite based results were later confirmed by mathematical models of vegetation growth (von Hardenberg et al., 2001;Rietkerk et al., 2004). Recent studies using more advanced modelling techniques discuss this phenomenon as characteristics of landscapes with water-limited systems. Mosaics of patches differ in resource concentration, biomass production and species richness (Gilad et al., 2007). They are a key factor in driving ecological processes at different spatial-temporal scales and modifying vegetation distribution and 20 species diversity, and may contain information on desertification processes (Von Hardenberg, 2010). Two types of vegetation patchiness in water-limited systems are discussed: a periodic pattern and an irregular scale free pattern; the latter one is more common in nature (Von Hardenberg et al. 2010;Kletter et al., 2012). Rietkerk et al. (2004) highlighted the importance of two processes that attracted considerable attention in scientific community during the past decade: "ecosystem engineering" and "self-organized patchiness".

25
Ecosystem engineers (Jones et al., 1994(Jones et al., , 1997Gilad et al., 2007;Meron, 2012) are organisms that modify, maintain and create habitats by causing physical state changes in biotic or abiotic materials and as such provide habitats for other species. Self-organised patchiness means a mechanism of positive feedback between plant growth and availability of water (Valentin et al., 1999;Rietkerk et al., 2004). As indicated by mathematical models, vegetation patternss are clearly related to an instability of spatially uniform vegetation (Thiéry et al., 1995;von Hardenberg et al., 2001;Rietkerk et al., 2004). The models predicted three pattern states of vegetation according to the rainfall amount: (1) an uniform vegetated state (dry-subhumid), (2) an arid and semi-arid state and (3) an uniform bare (hyperarid) state. The models also predicted a possible coexistence of different stable states under the same rainfall conditions. The range of coexisting patterns and bare states determines the extent of irreversibility of associated desertification process (von Hardenberg et al., 2001). Gilad et al. (2007) 35 summarised five basic vegetation stages along the rainfall gradient ( Fig. 17):  Uniform stages at high rainfall  Periodic gap, stripe and spot patterns at decreasing rainfall and  Bare soil at low rainfall 40 Figure 17. Model results show vegetation pattern development from an originally bare state into an uniformly distributed vegetation, reflecting system's ability of optimal self-organisation with respect to water resources (Gilad et al., 2007). Rietkerk et al. (2004) reviewed studies that linked self-organised patchiness to catastrophic shifts in 45 ecosystems. Such catastrophes are commonly attributed to the existence of two alternative stable states in ecosystems. The authors defined this as bistability (see also Thiéry et al., 1995). Increased resource scarcity leads to a spatial reorganisation (see also Cornet et al., 1992 andValentin et al., 1999). According to the model, certain spatial structures may develop in real ecosystems that only arise when resource availability has decreased. Simulations showed that under growing aridity conditions bare spots merge into 'labyrinthine' stripes which subsequently become a bare matrix interspersed with vegetation spots. If these vegetated spots disappear a complete desert may occur (Barbier et al., 2006). Rietkerk et al., (2011) proposed the hypothesis that imminent catastrophic shifts in ecosystems can be predicted by self-organized patchiness (ecosystem engineering). In these models the vegetation shifts catastrophically from a spotted state to a bare homogeneous state if rainfall is decreased beyond a threshold. It is worth mentioning that according to Rietkerk et al. (2011), increased rainfall may not recover the spotted state because the resource concentration mechanism (concentration of soil water under vegetation patches) fails. If this development is associated with a massive loss of ecological and economic resources, it wills effects human societies dramatically. The previously mentioned numerical modelling studies 5 mainly discuss pattern formation triggered by natural phenomenon (rainfall, geomorphology). Human impact is mentioned but not investigated in detail (Gilad et al., 2007;von Hardenberg et al., 2010). Vincke et al. (2010) also observed a contraction phenomenon in the Ferlo region in Senegal where they documented an increasing shift of two robust species (Boscia senegalensis and Guiera senegalensis) from tops to depressions. The authors suggested that these changes may have contributed to the shift from homogeneous vegetation pattern into a patchy distribution of vegetation. Barbier et al. (2006) applied Fourier analysis to highresolution remote sensing data in south-west Niger. The analysed aerial photographs covered a period of 40 years (1996 to 1956). According to their results the formerly homogeneous savanna has been dramatically changed into a spotted pattern. Protected areas showed less-spotted pattern than areas characterised by strong human impact. The authors discussed the observed spatial vegetation changes as potential indicators for climatic 15 and anthropogenic constraints and underlined that the intensity of the patterning process during the observation period of forty years was exacerbated by human activities. This is in line with investigations of Barbier et al. (2006) in south-west Niger, who had analysed aerial photographs for the period 1956 to 2006. The forming of contracted vegetation was confirmed by Couteron et al. (1997) on the basis of aerial photograph data from 1955 and 1984 for a site in Burkina Faso.

57 Key messages and new aspects
The following key messages and new aspects can be summarised from the above discussions:   Analyses solely based on NDVI data argue for a Sahel re-greening while a broader approach with Landsat-based analyses does not (see Scet. 4.35). Investigation in the Canal du Sahel area shows a regreening trend for the research site if Landsat data recorded in 1985 and 1991 are used. If earlier Landsat data from 1976 are compared with those from 1991, however, the opposite trend is visiblea significant decline of ligneous vegetation (Kusserow, 1994(Kusserow, , 1995. This example also points towards 5 the key question: which observation period is best used for trend analyses? Due to the severe drought period in the early 1980s, trees and shrubs may have shed their leaves completely. Thus, the NOAA-AVHRR sensors from that period may have also recorded dense woody vegetation cover without identifying them as woody vegetation because of its leafless state. The NDVI based Sahel greening is therefore to be questioned with three main arguments: (1) The observed re-greening since the early 1980s seems to be predominantly based on an increase in agricultural crops and herbaceous cover (currently under review, see Sect.4.35); (2) statements on the development of post-drought ligneous cover bear little significance due to the temporarily leafless trees and shrubs and (3) a comparison with satellite imageries recorded in the 1970s indicate much more dense woody vegetation cover.  Changes in woody vegetation distribution in relation to self-organised patchiness can be used as key

15
indicator for desertification processes as discussed in Sect 4.46. Own investigations conducted in Mali in the early 1990s indicated that the spatial cover of woody vegetation has changed within a time span of 40 years. The woody cover, originally characterised by an uniform (scattered-diffuse) distribution pattern, has turned into a highly fragmented pattern with spots and isolated bands. This specific pattern formation was considered as a principal indicator for desertification (Kusserow, 1994). This first 20 scheme could be confirmed by analysing sets of Landsat data for test areas in Mali, Burkina Faso and Niger (see Sect. 4.13,4.2,4.46). This is a new aspect, brought into the debate of Sahel greening/browning. It could be also shown, that the postulated (based on numerical modelling) pattern formation time span of 37 years (Gilad et al., 2007)   The discussion above shows the importance of the selected observation period when debating Sahel greening 45 vs. Sahel browning. In addition, the findings presented in this article argue for a new understanding of the process of desertification in the Sahel region rather than further accentuating the two contrary positions of a greening Sahel vs. a browning Sahel. The author suggests considering the Sahel as an ecosystem that changed from an originally "greener" state into a new more desert-like system. Main indicators are species turn over and vegetation pattern formation. The tipping point was the renewed drought period in the early 1980s. Finally, two key questions shall be raised for further research and debate:  Which observation period should be taken as default period to assess ecosystem changes? Do we want to be dependent on temporal limitations of methods?  How should the stability of an ecosystem be evaluated? Should we discuss any plant spreads (even it is an indicator for degradation and losses in biodiversity) as a positive sign? What does that mean with regard to the worldwide highest population growth rates in the Sahel?
Acknowledgements. I am grateful to Brigitte John, University of Bayreuth, for supporting me with the acquisition of satellite images, and Andrea Oestreich, Freie Universitaet of Berlin, for providing me with meteorological data and rainfall analyses for almost 30 years. I would also like to thank my husband Christian for his fruitful comments and critical remarks. I am especially grateful to all both reviewers who helped        Gourma, Mali, status 1976-1977, modified after Krings (1980. The box shows the position of the satellite image section (see Fig. 4).

Figure 7.
Position of late Quaternary dune systems in the Sahel according to Mensching (1990).   (1931( -1960( ) and CLINO 2 (1961, position of research areas is indicated by the red box (modified after Kusserow and Oestreich, 1998).    . Left: Changes in the spatial vegetation distribution from North to South: Sahara -South Sahara/North-Sahelian transition zone -Sahel (modified according to Kusserow 1994). Right: Chronology of changes in spatial vegetation distribution in the south, i.e. Sahel (modified according to Kusserow 1994).    Model results show vegetation pattern development from an originally bare state into an uniformly distributed vegetation, reflecting system's ability of optimal self-organisation with respect to water resources (Gilad et al., 2007).