Local soil quality assessment of north-central Namibia: integrating farmers’ and technical knowledge

Soil degradation is a major threat for farmers of semi-arid north-central Namibia. Soil conservation practices can be promoted by the development of soil quality (SQ) evaluation toolboxes that provide ways to evaluate soil degradation. However, such toolboxes must be adapted to local conditions to reach farmers. Based on qualitative (interviews and soil descriptions) and quantitative (laboratory analyses) data, we developed a set of SQ indicators relevant for our study area that integrate farmers’ field experiences (FFE) and technical knowledge. We suggest using participatory mapping to delineate 5 soil units (Oshikwanyama Soil Units, KwSUs) based on FFE, which highlight mostly soil properties that integrate long-term productivity and soil hydrological characteristics (i.e. internal SQ). The actual SQ of a location depends on the KwSU described and is thereafter assessed by field soil texture evaluation (i.e. chemical fertility potential) and by soil colour shade (i.e. SOC status). The resulting information includes internal SQ (KwSU), chemical fertility potential (sand content) and the soil organic carbon content status (colour shade). This three-level information reveals SQ improvement potential and aims to help farmers, 10 rural development planners and researchers from all fields of studies understanding SQ issues in north-central Namibia. This SQ toolbox suggestion is adapted to a restricted area of north-central Namibia but similar tools could be developed in most areas where small-scale agriculture prevails.

. Frequently used soil properties that may be used as field soil quality (SQ) indicators, possible field measurements techniques and challenges for local users (adapted from Wienhold et al., 2004).

Technical soil quality assessment
Many SQ indicators have been developed over the past decades (e.g. Mueller et al., 2010;Wienhold et al., 2004) and the need to adapt SQ indicators to local conditions was acknowledged very early (Granatstein and Bezdicek, 1992;Nicholls et al., 2004). Most of the indicators require measuring physical, chemical and/or biological soil characteristics that need laboratory measurements, specific technical material and/or experts' knowledge (Table 1). Therefore, most SQ indicators cannot be used 5 directly by farmers (Nicholls et al., 2004), which is particularly problematic in low-income regions due to limited availability of laboratory and experts' services (Musinguzi et al., 2015), like in NCN.
Many SQ indicators are based on yield data collected during two (e.g. Andrews et al., 2004) or even only one year . With such short records, it is impossible to consider how inter-annual climatic variability affects subsistence farmers, who aim to reduce the risk of harvest failure (Graef and Haigis, 2001). Therefore, most SQ indicators developed using yield 10 data collected during periods too short to fully reflect climatic constraints to production are of limited relevance in areas with high interannual rainfall variability. Considering the shortcomings of some SQ indicators, it is therefore imperative to develop "cost-effective and user-friendly tools" (Musinguzi et al., 2015) to evaluate SQ based on land users requirements.

Farmers' field experiences
Farmers' field experiences (FFE) include all farmer-based soil fertility assessment techniques (Musinguzi et al., 2015). This 15 terminology is preferred over "indigenous knowledge" or "local knowledge" because it refers to a clearly defined group of land users, all people involved in farming (farm owners, workers, children). FFE are essential as entry point for outsiders to understand local land use practices and local soil variability (Mairura et al., 2007;Ramisch, 2004). Many studies incorporate FFE to select the most appropriate properties to use as SQ indicators (Musinguzi et al., 2015;Nicholls et al., 2004). The resulting local SQ indicators cover broader agronomic properties than technical SQ indicators as they may account for economic issues (Warren, 1991), long-term productivity or risk management practices (Graef and Haigis, 2001) , for example dealing 5 with rainfall variability.
Aside from improving the relevance of SQ indicators, the use of FFE involves farmers in the evolution of agricultural practices (Ditzler and Tugel, 2002;Mairura et al., 2007;Warren, 1991). However, FFE can be inaccurate, biased by social context (Gray and Morant, 2003) and resilient against environmental and socio-economic changes (Briggs and Moyo, 2012). Technical knowledge, on the other hand, is valuable for its level of standardisation, which allows for spatial and temporal comparisons 10 and facilitates international communication (Niemeijer and Mazzucato, 2003). Scientists should therefore integrate both knowledge systems to provide tools connecting FFE and technical knowledge (Lima et al., 2011). Methodologies to select indicators for SQ based on the integration of FFE with technical knowledge have been developed and discussed, and yielded promising results (Barrios et al., 2006). Most studies concerning integrated soil knowledge showed the parallels between technical and farmers assessment, but only a few developed local SQ toolboxes to fully evaluate the SQ conditions (Table 2). 15 Farmers knowledge of environmental factors and SQ in NCN has been already collected and discussed in various studies Rigourd et al., 1999;Verlinden and Dayot, 2005), but there is still "a lack of understanding [of local land classification system] by scientists or extensionists [. . . ]" (Verlinden and Dayot, 2005). A relatively high number of "indigenous land units" were described based on vegetation, landforms and/or soils Verlinden and Dayot, 2005). These studies present an interesting collection of FFE, but none was developed into locally adapted SQ indicators. Yet, 20 such indicators are essential to allow researchers and farmers to assess SQ at a specific location and time-period relevant for agricultural cycles (Barrios et al., 2006).
Based on qualitative (semi-structured interviews, soil profile descriptions) and quantitative data (field soil profile descriptions, laboratory measurements), we suggest a set of SQ indicators relevant for our study area that integrate FFE and technical assessment. Following Barrios and Coutinho (2012) these indicators must be: a) Practical and easy to use under field conditions; 25 b) easy to interpret; c) relatively economical; d) sufficiently sensitive to highlight the changes under study; e) integrate physical, chemical and biological characteristics and processes; f) useful for estimating all relevant soil properties; g) give good correlations between plant productivity and soil health. We aim to verify the benefits of using FFE for soil quality assessment as the development of SQ estimation tools is vital for SQ management in areas where small-scale family agriculture represents large proportion of land use.

Study area
In NCN, the climate is semi-arid subtropical with a rainy season from December to April. Average annual precipitation ranges from 350 to 550 mm with large inter-and intra-annual variability (Mendelsohn et al., 2000). In Ondangwa, the annual rainfall during 1959-1973 ranged from 200 to 1039 mm with an average of 495 mm . Crop production failure 5 because of rain quantity and distribution occurs every second year (Keyler, 1995). The area lies over the Owambo sedimentary basin with the upper part constituted of aeolian sands redistributed throughout the Quaternary Period (Miller et al., 2010). The region is characterised by the endorheic Cuvelai drainage basin and the north-eastern Kalahari woodlands or Kalahari Sandveld ( Figure 1; Mendelsohn et al., 2000).
Non-commercial agricultural activities are the most important land use in NCN (Mendelsohn et al., 2000). Around 120'000 10 households are farming in the region, mostly cultivating small-scale (1-4 ha) rainfed pearl millet (Pennisetum glaucum; Mendelsohn et al., 2013). Average yields of millet are very low, (220 kg ha-1 in average in Ohangwena region), highly variable from year to year and from household to household, due to low soil fertility, low nutrient supply, irregular rainfall and pests (Central Bureau of Statistics, 2003;Mendelsohn et al., 2000;Rukandema et al., 2009).
Three groups of villages in Ohangwena region were selected (Omhedi, Ondobe, Ekolola; Figure 1) based on dialect homogene-15 ity (Oshikwanyama) and environmental heterogeneity (vegetation, soils). These villages lie on a west-east climatic, edaphic and land-use gradient with a mosaic pattern of soil and vegetation (Mendelsohn et al., 2013). The annual rainfall quantity, the proportion of deep sandy soils and forest cover increase eastwards. The westernmost area (Omhedi) is largely influenced by the active drainage system of the Cuvelai River, which creates a network of water channels (called locally iishana) that significantly influenced soil development (fluvial deposits, salinization). Ondobe is located between the drainage basin in the west, 20 and the Kalahari Sandveld in the east. Further east, Ekolola is characterised by the Kalahari Sandveld, which is dominated by deep loose sand deposits (Mendelsohn et al., 2000). All three areas were recently settled by immigrants from Angola, mostly during the 1910-1920s', but population density increased more dramatically in the westernmost areas due to water accessibility (Kreike, 2004). agricultural knowledge and open to discussion were visited several times. Mostly people above the age of 50 (75% of interview time) were surveyed because of their availability to talk and the knowledge they wished to share, typically elderly men (49% of total interview time). Most interviews were held in the house providing conceptual references, but some were held in the 30 fields or in front of soil pits, providing locational references (Oudwater and Martin, 2003). Questions aimed to generate information on the types of soil that are cultivated and the characteristics that differentiate them. With "Oshikwanyama Soil Units"

Assessment of farmers' field experiences
(KwSU) we refer to the soil units that are distinguished by the farmers by sight, touch, experienced yields or others (following Vegetation appears in green, bare soil appears in orange, water bodies in blue (Digital Atlas of Namibia).
the definition of Indigenous Land Units suggested by Verlinden and Dayot, 2005).
All the interviews were held in Oshikwanyama and audio-recorded. Direct interpretation was performed by, mostly, Ms Martha Shekupe Fillemon (20). The English interpretation was afterwards completely transcribed. Parts of the interviews were transcribed in Oshikwanyama and translated into English by non-professional local translators. The interviews were annotated using MaxQDA 11 (VERBI GmbH, 2014) to facilitate the qualitative data analysis. The annotation system included KwSU 5 names (omutunda, omufitu, elondo, ehenene, ehenge) and "soil quality". The latter annotation was used to select quotes in which a certain location or a specific KwSU was characterised with regards to the suitability for pearl millet cultivation.
Over the total number of informants (46), we calculated the proportion of them who mentioned each KwSU. Afterwards we associated these interviews to specific soil properties, which are finally grouped into five frequently mentioned properties: hardness, soil hydrology, productivity potential, soil colour shade and soil colour hue.

Field soil profile description and sampling
The Guidelines for soil description (FAO, Land and Water Division, 2006) were used for standardised soil profile description.
In the context of this study, we only discuss the horizon limits, clods consistence, bulk density and moist colour down to 40 5 cm, as they are best suited to the objective of developing an SQ tool that could be used by various land users, who have not the resources and expertise to go through a full soil description. Soil colour was estimated in the field using Munsell soil colour chart on a moist sample for each horizon. Soil colour provides information about soil formation processes (e.g. leaching, clay alteration) and soil organic carbon content (SOC) (Viscarra Rossel et al., 2006). The dry consistence was evaluated by crushing a clod of soil between the fingers. This property informs on the amount and type of clay, SOC and soil particles organisation Two 100 cm 3 -sampling rings were collected from each described horizon and homogenised to create a single mixed sample per horizon. Dried-samples were weighted to calculate bulk density, sieved (2 mm) and used for further analysis.

Laboratory analyses
Soil texture is the most important soil characteristic having a direct influence on most soil processes and properties (Vos et al.,15 2016). It was calculated using laser diffraction (Malvern Mastersizer 2000) that measures volumetric particle-size distribution.
Prior to measurement, samples were shaken overnight in water and dispersed during 9 J ml −1 ultrasonic energy. The particle size class <20 µm was considered as the active mineral fraction (Feng et al., 2013).
SOC plays an important function as adsorbing material and is often used to evaluate SQ (Musinguzi et al., 2015). SOC saturation (C-saturation) is defined as "the ratio of the present topsoil total [SOC] level relative to the same soil in its undisturbed 20 [. . . ] state" (Sanchez et al., 2003). Various models have been developed to evaluate the SOC of a C-saturated soil Zinn et al., 2007), for example based on the proportion of <20 µm fraction (Feng et al., 2013). We choose the model from Feng et al. (2013) because it is based on a large review of studies, and developed for soils with predominantly 1:1 clay minerals, common in the tropics.
SOC and inorganic carbon contents were determined with a LECO ® analyser (RC-612). Soil electrical conductivity was mea-25 sured in 1:5 (soil-water) suspension and pH CaCl2 in a 1: Cation exchange capacity and base saturation values indicate the cation reservoir of a soil and are important characteristics to evaluate the ability of a soil to sustain plant growth. Both properties were not measured in this study because the presence of calcium carbonates (secondary precipitations observed in various soil profiles) and soluble salt (high EC in ehenene, mostly NaCl) strongly influences the measurements (Sparks et al., 1996), which makes results very difficult to use for comparison, es-30 pecially considering the low expected values due to low cation exchanging materials (mostly clay and organic matter). Instead, we used robust and sufficiently accurate methods as proxy for cation exchange capacity (soil organic carbon and the <20 µm fraction content) and for base saturation (soil pH) (Blume et al., 2011).
Known to be limiting nutrients in most agricultural land and in particular in sub-saharan Africa, nitrogen and phosphor availability are most likely significant for plant growth. However, didn't include these analyses in the current study given that this study aims at understanding and following longer-term soil fertility discussion while these nutrients aer more related to soil short-term fertilisation.
3 Results and Discussion to pearl millet cultivation) using several properties. In cultivated areas, five Oshikwanyama soil units (KwSUs) were frequently described: omutunda, ehenge, ehenene, omufitu, and elondo (Table 3). Knowledge and descriptions of these local soils were largely shared among the interviewed population, and we did not observe differences based on gender, generations or studied 10 eco-regions. Some criteria used in the FFE were general (e.g. productivity potential), while others were more specific (e.g. soil colour shade and hardness, waterlogging risk; Table 3).
KwSUs' names define specific objects in the landscape. For example, the suffix -tunda in omutunda means "something on a hill" (TN, 65, Ekolola) 1 and omufitu refers to woodlands located close to villages ("a land with many bushes and trees"; KS, 60, Ondobe). These names are instilled in the everyday language, which explains the homogeneity of the soil-related vocabulary 15 among the population and suggests that labelling of places (with KwSUs) changes little over time.
We calculated the proportion of informants mentioning specific characteristics for each KwSU to highlight the most prominent characteristics, per KwSU and based on total number of informants mentioning any of the five KwSUs (Table 3). The properties that were the most frequently used to describe KwSUs were related to soil hardness (63.5%), productivity potential (57.7%), soil hydrology (43.8%) and soil colour shade (38.0%). The morphological properties (colour shade, consistence) referred 20 mostly to topsoil layers as farmers indicated characteristics that were discussed during transect walks. The consistence, or the concept of hardness, is evaluated under dry conditions, which impacts importantly the difficulty of ploughing (performed early in the rainy season). As observed by Verlinden and Dayot (2005), the predominance of each characteristic varies depending on the unit described. For example, hardness/softness is a prominent characteristic to describe omutunda and omufitu (used by resp. 72.2% and 70.8%) while soil hydrological characteristics were important to describe ehenge (68.8%). 25 The high frequency of interviews mentioning productivity (57.7%; Table 3) might have been influenced by the aim of the study and frequent questions concerning productivity by the researchers. Farmers considered unanimously omutunda as the most fertile soil and agreed that pearl millet productivity is strongly limited in ehenene (Table 4). Productivity in elondo, ehenge and omufitu did not reach consensus. The productivity of these KwSUs may largely depend on factors less dependent on soil (rainfall, fertiliser availability). Notably, ehenge is good in poor rainfall years, but poor in good rainfall years (Table 4). 30 Each KwSU is characterised by a series of indicators. A selection of these indicators is illustrated in Table 4. However, it should 1 To keep the informants anonymous, we used a code that indicates: 1) a two-letter name, 2) the farmers' age and 3) the study area of the farm.  Table 3. List of farmers' field experiences characteristics used to describe each KwSU, with the number of informants mentioning each KwSU (n) and the proportion of informants mentioning each characteristic (in relation to n). Values are only indicative as the data collection method was not adapted for statistical analyses.
be kept in mind that these descriptions are only a summary of the characteristics mentioned by the informants.
The productivity of soils depends not only on internal soil properties and processes (waterlogging risks, landscape position) or climatic conditions, but also on management strategies (e.g. fertiliser application). The effect of management was acknowledged by farmers who explained that KwSUs represent not accurately the actual SQ (e.g. "omutunda is not always fertile, it needs to be dark", SK, 60, Ondobe). Farmers estimated the actual SQ of a location also based on crop health, soil consistence 5 and soil colour shades ("[. . . ] needs to be dark", SK, Ondobe), hardness ("millet likes hard soil", HP Ondobe). We will discuss the technical significance of these properties below.
Soil hydrological properties were mentioned frequently to describe KwSUs. These properties need to be understood in relation to rainfall variability (Table 4). Productivity of omutunda drops during droughts ("pearl millet is burned", JL, Ondobe), while it increases in ehenge ("ehenge is good in year with lack of rain", LS, Ondobe). Therefore ehenge secures minimum harvest 10 during poor rainfall years, which is essential for farmers relying on yearly food production (Graef and Haigis, 2001). Conversely, ehenge undergoes waterlogging during good rainfall years ("[ehenge] used to be full of water", NJ, Ondobe), which strongly limits pearl millet growth. These soil hydrological characteristics are difficult to assess during standard field surveys and the integration of these characteristics in KwSU definitions is crucial for SQ evaluation as soil water availability is the most significant limitation in semi-arid regions (McDonagh and Hillyer, 2003). 15

Technical analysis of farmers' field experiences
Results from technical analyses are summarised in Table 5, in which the soil characteristics are calculated for the layer 5-15 cm and 25-35 cm using an arithmetic mean of the different values weighted by the depth of each horizon. All described soils have very low organic carbon (<5 mgOC g −1 ) and high sand content (>70% in the 5-15 cm layer). Omutunda has a larger proportion of <20 µm fraction (6.5 to 22.8% in the layer 5-15 cm) and more SOC (1.4 to 4.4 mgOC g −1 ) than all other studied 20 KwSUs. Furthermore, slightly alkaline conditions (Table 5) indicate a high base saturation. All these characteristics suggest the higher potential of omutunda to provide nutrients, coming from any sources, compared to the other KwSUs. This capacity is hereafter called chemical fertility. A slightly more acid soil solution, a smaller amount of <20 µm particles and SOC in elondo indicate lower chemical fertility. The proportion of <20 µm fraction in ehenene can be high (up to 16.4 %), but high pH in water (up to 10.1; results not shown) restricts plant growth. All ehenge and omufitu described have very low proportion of <20 µm fraction (<6.5 %) down to 40 cm. Our laboratory results therefore support farmers' assessment pointing to the greater 5 chemical fertility potential of omutunda.

International classification: the WRB
Only one diagnostic horizon and a limited number of diagnostic properties or materials could be described following the WRB (2016). The soil texture is the main characteristic used to describe the Reference Soil Group (RSG). Indeed, soil profiles without layers finer than Loamy sand were categorized as Arenosols (17)  Ondobe (n= 10) in comparison to the other areas (Omhedi= 2, Ekolola= 3), the statistics presented in Table 7 are skewed towards the characteristics of omutunda in Ondobe. This does not jeopardize the substance of these results given the diversity found in the area (transition from floodplain environment to Kalahari woodlands).
From FFE perspective, omutunda was mostly defined by excluding areas not suitable for pearl millet because (i) it does not experience waterlogging (hypoxic conditions); (ii) it does not have loose sand topsoil (very poor chemical fertility); and (iii) it 5 does not have very shallow hard soil layer (limits water storage capacity and restrict workability). Pearl millet can be cultivated on various soils (Baligar and Fageria, 2007), which contributes to the large variability of soils considered as suitable for its cultivation. Temporal variation of SQ was acknowledged in FFE and various degrees of degradation (e.g. organic and nutrients depletion, salinization) lead to variability in SQ of omutunda at a specific time. Management practices (amount of fertiliser, ploughing) therefore also contribute to add some variability. There were small differences depending on the area of study 10 and the surrounding environment (Table 5). Omutunda described in Ekolola (Kalahari Sandveld environment) has coarser texture compared to the omutunda described in Omhedi and Ondobe (Floodplain; Table 5). These differences were expected as FFE were constructed based on comparative observations (e.g. "harder than") and therefore influenced by the surrounding environment (Birmingham, 2003;Niemeijer and Mazzucato, 2003).
The variability described in the various studied omutunda illustrates the need of developing tools for standardisation. This 15 would help avoiding classifying soils that should not be compared directly, but need to be considered as various entities that show similar features.

Importance of a soil quality evaluation toolbox
We showed that KwSUs represent locations in the fields with specific soil characteristics and provide information about their 20 potential productivity. It notably includes soil hydraulic characteristics. Clearly, the KwSU knowledge is land use orientated (e.g. suitability for pearl millet, workability), adapted to local conditions (rainfall variability) and represents the local soil productivity potential. Farmers also include crop health, soil consistence and colour shade to evaluate SQ of a specific location (Sect. Oshikwanyama Soil Units: A homogeneous body of soil knowledge). We also showed that each KwSU includes a large variety of soils properties (especially omutunda) for which SQ for pearl millet production differs. To estimate SQ, it is therefore 25 important to standardise the assessment of the SQ at a specific location and time. This would allow a comparison based on, for example, agricultural or climatic cycles or management techniques. Technical soil characterisation (e.g. soil texture, colour) proved to be suitable to standardise SQ assessment in other locations (Niemeijer and Mazzucato, 2003).
The World Reference Base for Soil Resources (IUSS Working Group WRB, 2014) is used to draw the Namibian soil map.
This classification is mainly orientated towards representing "primary pedogenetic process[es]". Therefore, the use of this 30 classification is not relevant to highlight SQ differences at small-scale in a region with poorly developped soil profiles given the low prevalence of diagnostic properties and horizons.  Table 7. Summary of the chemical and physical characteristics of topsoil (5-15 cm) and subsoil (25-35 cm) layers of the studied omutunda soil profiles. CV= coefficient of variation.
We will first show the meanings of the soil characteristics used by the farmers to evaluate SQ and link these with soil technical analyses. Based on these links, we will suggest ways to use this knowledge and to standardise the SQ assessment.

Important characteristics for field soil quality evaluation
Soils with a high proportion of <20 µm particles are harder in dry conditions than soils with coarser texture (Welch's F (3, 55.3)= 28.46, p-value< 0.01; Table 8), results that are supported by specific studies (Harper and Gilkes, 2004;Rawls and 5 Pachepsky, 2002). These fine-textured soils have larger area of active surfaces, which play important role in fixing SOC and nutrients (Feng et al., 2013). Through talking about hardness, farmers indirectly refer to the proportion of fine soil particles (Osbahr and Allan, 2003). It therefore indicates a major property contributing to fertility. The proportion of <20 µm fraction content in soils was increased through homestead shifting (clay-bricks remains) or mining riverbeds (Kreike, 2013). Sand content (>63 µm) can be used to estimate the proportion of <20 µm fraction given the good correlation between the proportion of 10 these two classes (p-value< 0.01, R 2 = 0.98). We referred to it as the potential chemical fertility because it requires appropriate fertilisation to fully achieve maximum yields.
FFE acknowledge the importance of "soil darkness" to estimate SQ. SOC is used as an index of SQ in many studies because of sensitivity to management practices (Barrios and Trejo, 2003;Lima et al., 2011;Musinguzi et al., 2015;Osbahr and Allan, 15 2003). Sanchez et al. (2003) used the concept of C-saturation to evaluate the Soil Fertility Capability, in which a C-saturation above 80 % indicated good soil conditions. For various textural classes, SOC of undisturbed soils was calculated using Feng et al. (2013) and the colour shade value related to it was estimated using Blume et al. (2011, p.51) (Table 9).  Table 9. Calculated soil organic carbon content (SOC) of 80%-C-saturated soil for various sand content using Feng et al. (2013) and the estimated colour shade value (Blume et al., 2011, p.51).

The soil quality evaluation toolbox
Based on the link between FFE and soil technical properties, a toolbox for evaluating SQ based on indicators adapted to Western Ohangwena region was developed. With this toolbox, SQ is assessed in two steps (Table 10): 1) Field participatory mapping of KwSUs; 2) technical SQ evaluation at specific locations using soil colour shade and sand content.
With KwSUs, farmers classify soils with comparable internal properties and suitability for pearl millet production (Table 4). 5 The distribution of KwSUs in the fields is known by most household members. With participatory mapping, the farm can therefore be divided into KwSUs (omutunda¸ehenge, ehenene, elondo and omufitu), which represent internal soil properties.
Most elondo are fine sandy soils, in which coarse texture (>90%) would indicate ongoing or past degradation (e.g. overland flows, eluviation) because elondo is described as a fertile soil. Conversely, the proportions of sand are very high in ehenge and omufitu (Table 5) and <90% sand indicates that major soil improvements had been undertaken (e.g. former homestead location). Given that omufitu are defined by their high sand content, omufitu will never present <90% sand without human 15 activity. Plant growth in ehenene is limited by the high soil pH and high runoff intensity (Rigourd et al., 1999) and the soil texture is not relevant for SQ evaluation for this specific KwSU.
Theoretical colour shade value of C-saturated soils vary from 3.5 for fine soils (<80 % sand) to 4.5 for very coarse soils (>95 % sand; Table 9) (Blume et al., 2011, p.51). The three levels indicate fertilisation status. Positive meaning sufficient organic inputs and negative meaning largely missing inputs. Munsell colour charts are a standardised tool commonly used to evaluate 20 bulk soil colours. Few issues are related to the use of Munsell in this context. First, the charts are relatively expensive, but affordable for regional agricultural offices and are available for researchers from most soil science research groups. Second, the colour evaluation is somehow subjective and in the context of North-Central Namibia mostly only small differences in soil colour could be observed. Therefore, we suggest creating a collection of soil samples representing the regional soils and compare based on those standards. 25 To make the evaluation closer to FFE, we suggest adapting the colour value scale for ehenge and ehenene (optimal colour value +1) because these soils are lighter than the other KwSUs ("in ehenge the soil will look white", KS, 60, Ondobe) and cannot reach low colour values.

Outcome of toolbox application
The developed toolbox is and remains a suggestion for evaluating SQ and for prioritising SQ-improvement practices. The 30 resulting SQ assessment gives a number of values, which bring more information about improvement potential than a single value (Ditzler and Tugel, 2002  chemical fertility potential of the soil, which can be improved only with medium-term (decade) management practices (homestead relocation, erosion reduction). Colour shade indicates the SOC status and can be modified in short-term, by agricultural techniques (e.g. manuring, conservation tillage). For each characteristic, the soil can be classified into two to five categories, number that can be easily handled with for mapping purposes. The overall number of possible classes (29 classes) would be however too high to be used to create meanigful maps. The objective of the current work was to help farmers evaluating the 5 improvements potentials of their soils, which is achived by using the set of indocators.
The toolbox output provides three-value estimates that need to be interpreted based on local soil knowledge and socio-economic context. For example, a soil can be characterised, by "ehenge poor+" (Table 6), which means that: 1) The location undergoes waterlogging and is valuable during poor rainfall years (ehenge); 2) the chemical fertility potential is low (poor); and 3) it is well enriched with organic materials (+). Investment to improve SQ at this location could then focus on waterlogging risk 10 reduction or clay enrichment, because strategies concerning SOC are already adapted to the location and ameliorate SOC status would barely improve SQ and productivity. The test represents a way to estimate current soil status and it is therefore relevant to survey SQ in NCN. The soils described during this study present a large diversity of SQ based on the developed SQ toolbox (Table 6). Half the described omutunda (7/15) would need more organic inputs and five are considered degraded. These results highlight the threat that exists for each location and indicate the measures to prioritise for SQ improvements. There is 15 a lack of data to support the occurence of soil degradation or improvement. However, these processes were perceived by some farmers and explained during the interviews. Because of the lack of long term productivity data, it cannot be used to estimate the productivity potential of a location. However, it would be relevant to guide, for example, the systematic collection of yield data.

20
We developed a locally adapted method for SQ evaluation. Using the toolbox with farmers in NCN showed that it is practical, affordable, precise and relatively easy to interpret. The suggested toolbox combines participatory soil mapping with sand content and colour shades assessment. The toolbox fulfils the following conditions: (i) practical and easy to use under field conditions; (ii) relatively precise and easy to interpret; (iii) relatively economical; (iv) sufficiently sensitive to reflect the impact of soil use and management; (v) integrates physical, chemical and biological characteristics and processes, and (vi) be useful 25 for estimating soil properties or functions that are difficult to measure.
The combination of farmers' and technical assessment cumulates advantages of both systems of knowledge, specifically, the integrated long-term knowledge of the farmers (i.e. long-term productivity) and a short-(colour) and medium term (sand fraction) SQ status assessment, sensitive to land management practices. The toolbox can be used jointly by farmers and researchers from all fields of studies. 30 The toolbox represents a step towards better SQ evaluation in NCN. While it is adapted to a restricted area, similar approaches can be used to develop SQ tools for areas where small-scale family agriculture represents large proportion of land use. The results strongly support the use of FFE as entry point to SQ assessment at the regional level, especially in semi-arid regions with high climatic variability and limited resources for SQ assessment.

Acknowledgements
The research necessary for this paper was possible through the SNSF-DFG funded project: Communal land reform in Namibia -Implications of Individualisation of land tenure. The authors gratefully thank all the informants for sharing their knowledge 5 and their reception into their home and the translators for their help during the field work. The authors thank also the headmen and headwomen, constituency leaders, the Governor of Ohangwena region and the director of the Ministry of Land and Resettlement for facilitating field work. We thank the Polytechnic of Namibia for their collaboration, in particular A. Verlinden and D. Wyss for their advice and help. We thank also our colleagues for laboratory work, advice and reviews.  Table A1. Soil profiles illustrating three of the most common KwSUs found in the area. Soil descriptions following the Guidelines for soil description (FAO, Land and Water Division, 2006) and soil names the World reference base for soil resources (IUSS Working Group WRB, 2014).