This section presents a summary of questionnaire responses.
3.1. Applications of Soil Maps
shows a summary of the results obtained from closed-ended questions relating to a fixed list of soil map applications. Map producers were asked to rank these applications according to their relative frequency of production, whereas end-users were asked to rank them according to their needs. These rankings were then combined to obtain a final score ranging from 1 to 6. The colors are used to classify these applications into large groups.
The answers from end-users showed that, in addition to agriculture, they were also interested in other topics. According to this graph, the topics aligned along the diagonal indicate quite a good correspondence between the needs of producers and end-users. At the extremes of the diagonal, it is possible to distinguish some topics that were almost never cited (such as hunting/fishing and leisure/tourism) on the bottom-left, whereas those on the top-right indicate both highly requested and developed topics (agricultural production, water quality management, soil sealing, erosion, forest production and contamination). These six topics can thus be considered to be the customer base of soil mapping applications, as the frequency and importance of dealing with these topics are ranked equally by both users and producers.
The results show that some user needs are not covered by map producers. Along the x-axis and circled in red, some topics appear to be clearly underproduced compared to the extent that they could be generated in accordance with demand. Other topics (circled in yellow on the right side, a little below the diagonal) that are deemed important by users are also relatively underproduced. This is particularly the case for biodiversity assessment and protection, which are considered by end-users to be very important but remain poorly covered by production.
One apparent area of ‘over-production’ (wetlands) corresponds to a demand from policy makers. This case is undoubtedly linked to the legislative effect of a decree concerning the zoning of wetlands, and to the fact that most of the requests for this zoning were directly sent to data producers. Indeed, the decree concerning the zoning of wetlands was related to a very large number of local authorities who then commissioned these maps. In doing so, these local authorities exerted an influence on mapping activities that is larger than their degree of representativeness among the end-users.
The red dots represent themes identified as being the most important by the research community, whether producers or users. Their position towards the positive extremes of the diagonal suggests a rather good match between research priorities and the overall ranking, although the research partly bears this ranking. Note the particular positioning of biodiversity issues, which are identified as a priority by users but have not yet been fully integrated into the mapping activities of producers.
3.2. Land Cover and Related Issues That Are Not Sufficiently Addressed
represents an attempt to identify the main thematic gaps and to classify them according to land cover. This is conducted using information provided in answers to several open-ended questions. The corners of the triangle in Figure 3
represent the main land cover types, which are grouped into three broad categories (i.e., urban and industrial, agricultural, forest/undisturbed areas). Colors are used over this triangle to represent the following: main global issues identified as being insufficiently covered by maps (blue), missing information about soil states (black), actions for which more precise soil data are needed (red), and certain non-strictly soil information (yellow) identified by end-users as missing but required for decision making and in order to deal with certain issues. The relative positions of these blocks indicate a preferential link to a given land cover.
At the top of the triangle, and at the borders of urban/industrial and peri-urban areas, the themes of decontamination and urbanization emerge. It is considered that soil in these environments has a very high heterogeneity and that there is a subsequent lack of precise soil data. Human health is a major issue in such environments, and although important everywhere, it is a higher priority in areas that are most populated, as these are also often the most contaminated. This explains its position close to the upper part of the triangle.
Soil sealing and land use planning are located along the side of the triangle, joining the urban and agricultural corners. When dealing with these topics, the ‘market value’ of the land is a major decision criterion. This criterion is not only due to intrinsic soil properties, but also to their location (e.g., closer to or further from urban areas, or to main roads and transportation lines) and to whether there is a human-related origin, such as the existence of protected areas or the influence of markets and trade regulation.
Along the side that joins the agricultural and forest or semi-natural corners are mainly data relating to past or current human actions that have been identified as lacking. End-users strongly expressed the need for information on land management and agricultural practices, both for current practices and also for information about the past history of land cover, land use and land management. This kind of information is of particular importance when dealing with issues relating to climate change, such as CO2 or N2O fluxes between the soil and atmosphere, or the potential of soils for adaptation, which partly explains the position of this issue in the triangle.
The side joining the urban and forest/semi-natural corners appears quite empty, which suggests that no specific soil-related issue has been identified at this interface.
Moving upwards in the triangle, biodiversity remains mainly the stake of non-strictly urban environments. For this issue, the difficulty lies in the acquisition of relevant data and their translation into operational indicators.
The center of the triangle represents the generic issue of valuating ecosystem services, which is considered equally as important for all environments.
Gradually moving closer to more anthropic environments, the theme of natural risks owes its position to impacts on buildings and infrastructure and on risks incurred by populations. Finally, the lack of soil data relating to health issues appears to be associated with urban and peri-urban environments.
3.3. Correspondence between Produced and Requested Variables
indicates the correspondence between selected variables recorded and used by producers and those requested by end-users. Map producers were asked to rank these variables according to their relative frequency of production, whereas end-users were asked to rank them according to their needs. The rankings were then combined to obtain a final score ranging from 1 to 10, and colors were employed to classify these applications into large groups. Some of the variables are both requested and produced, but others need the use of pedo-transfer functions (PTFs) or are rather difficult to access and process to derive a map. Furthermore, some remain essentially qualitative or less present in databases, and/or may exhibit quite fast changes with time. Finally, some of the variables are rarely produced.
According to this graph, different groups of variables can be identified according to whether they are relatively simple or costly to measure and map and whether they are stable over time. Interestingly, quite a large number of attributes relating to soil structure and soil water behavior are often considered important by end-users but are difficult for producers to map. The same occurs for biodiversity indicators.
A more complex situation is illustrated in the following with respect to the responses of end-users concerning the statement that they would like the thickness of soil to be described and quantified for different types of soil attributes.
If we refer to internationally agreed specifications (i.e., GlobalSoilMap
]), the soil thickness, also called the ‘soil depth’, corresponds to the definition provided in [9
], which is ‘the depth of a lithic or para-lithic material’. However, the soil thickness information required by end-users covers a wide range of values from 0.2 to more than 2 m. Approximately 17% of end-users only require information about topsoil properties (i.e., depths of 0–10 to 0–30 cm), and their needs are mainly focused on indicators of soil chemical fertility (such as major nutrients (N, P, K), some oligo-elements, pH and cation exchange capacity) for plant growth, or on soil contaminants (such as certain trace elements and pesticides). Nearly 35% and 31% of end-users require information about soil thickness to a depth of 1 m and 2 m, respectively. Overall, therefore, the range of thicknesses for which end-users required information is consistent with GlobalSoilMap
] for 82% of responses. In addition, the remaining requirements of 18% of respondents for information at deeper levels (i.e., deeper than 2 m) mostly related to applications involving deep-rooted crops such as vines and trees.
This result raises questions about both the definition of ‘soil depth’ and the information about soil thickness collected and stored in databases. One question could be formulated as, ‘should a rock that has fractures enabling roots to go deeper than 2 m still be considered part of the soil?’ and/or, ‘should a friable regolith or C horizon be considered part of the soil?’ The latter question relates to the lack of information about very deep layers contained in soil databases and the need to develop methods for dealing with right-censored values that have been generated by the use of augering at limited digging depths to collect information [11
]. In fact, the responses from end-users about soil thickness information requirements appear to relate to their need to know both the thickness of the soil for which they require soil property values and the value of the soil thickness itself. This allows a better interpretation of the somewhat surprising position of soil thickness in Figure 4
The types of soil attributes that are most frequently employed by end-users to derive thematic maps are shown in Figure 5
, and these attributes are grouped into five broad categories as follows,
Soil type or soil class, according to national or international classifications;
Physical properties (including particle-size, texture, bulk density, stoniness and thickness);
Hydric properties (including available water capacity, porosity and infiltration capacity);
Chemical properties (including pH, cation exchange capacity (CEC), exchangeable cations (Ca++, Mg++, K+, Na+), major nutrients and oligo-elements);
Biological properties (including carbon, biodiversity measurements and populations of nematodes and earthworms).
Interestingly, soil type is the most commonly used attribute, and this is partly attributed to the fact that many legacy soil maps are still only scanned and thus provide only a soil type/class in their legend. Therefore, if there is no further semantic information extracted from soil profiles or from soil databases (DBs) relating to maps, the only information available to end-users is the soil type/class. Another possible reason is that information about the soil type and class (for instance, in the World Soil Reference Base [14
]) is often used as a ‘proxy’ for a large number of thematic maps (e.g., Arenosols indicate a sandy soil texture, Chernozems imply a high soil organic content in the upper and deep layers, Rendzic Leptosols provide information about the CaCO3
content, ranges of pH and soil depth and Gleysols provide useful information about water-logging). However, some high-level groups, such as Cambisols, cannot be used to derive soil properties if they are not described using enough qualifiers.
The second group of soil properties is ‘physical properties’, particularly with respect to particle-size distribution and soil texture. This is consistent with their position in Figure 4
and the fact that soil texture may be used to derive certain important soil properties by using PTFs (e.g., [15
The third group is that of ‘hydric and chemical properties’, which received similar scores. The position of hydric properties may be attributed to the fact that they do not have a large enough presence in soil DBs (see Figure 4
). However, the position of chemical properties is a little more surprising and is in contradiction with the positions of pH, CEC and cations in Figure 4
. However, it must be remembered that very few oligo-elements are included in soil DBs. Another possible explanation is that most of these properties (N, P, K, pH, CEC and cations) have been published in a report on the status of French soil resources [19
], and that all these national maps are available online [20
The final group is that of ‘biological properties’, which re-enforces the need to include biodiversity measurements and indicators in soil DBs [22
]. Indeed, although very significant progress has been made on mapping microbial abundance and biodiversity on a national scale [23
], measurements are still very sparse on more local scales and are not usually integrated into soil DBs. In addition, there are still only sparse measurements of other organisms that are very important for soil biodiversity and functions, although on-going programs aim to fill these gaps, e.g., in [28
]. Finally, many biological soil properties vary highly with time, and it is thus more difficult to map them than it is to map stable variables.
3.4. Certain Methodological Considerations about Mapping
Although quite clear to map producers, end-users still tend to confuse scale, spatial resolution and precision. In particular, end-users’ responses reflected that they are confused (or unaware) about the effects of the type and size of mapping supports employed (e.g., points or blocks, polygons or grid-cells). Figure 6
shows the responses from producers and end-users to the question, ‘At which scale are polygon soil maps useful depending on the area they cover?’
The responses from both categories (producers and end-users) roughly follow a diagonal (from small-scale maps for very large areas to large-scale maps for small ones), which appears quite logical. However, it is of note that the responses from end-users show a larger spread around this diagonal, which indicates that end-users are more confused about the notion of scale. In particular, it is astounding that some end-users still believe they can employ a 1:1,000,000 scale map for field-scale decision making.
However, when asked about their need for resolution, most end-users said they require rather fine resolution products (i.e., from 25 to 250 m), which may suggest that most applications require a fine resolution and/or that end-users are aware that it is easier to aggregate fine grids into coarser ones than the reverse.
End-users express a strong interest in obtaining uncertainty indicators. However, a large proportion of them do not know which indicators are best suited for their purpose or how to consider them and communicate about them. In addition, there are strong differences between the answers of different categories of end-users (Figure 7
Respondents from research and higher education fields (mainly research institutes, universities and agronomy high schools) were mostly very confident that the uncertainty indicators (or indicators of the prediction performance of the maps) would be easy to use and that it would be easy to communicate this to decision makers. Respondents from the development and application domains were much less confident in their ability to consider uncertainty indicators, and were not very confident that these uncertainties would be considered by decision makers. This general tendency shows that a large number of end-users still feel uncomfortable about producing and using uncertainty indicators, and this tendency was much more apparent with people from agricultural organizations.
Both producers and end-users expressed a strong interest in moving from conventional to digital soil mapping (DSM). However, most indicated that they would need strong technical support to acquire new concepts and methods, and a significant proportion of end-users felt uncomfortable with uncertainty. The responses to many open questions showed that end-users would like to not only have digital maps of soil properties but would be more interested in maps that represented changes in these properties over time, or even maps forecasting possible future changes.