The following article is a continuation of the story of Turtle (Indorante, 9), a fictional field soil scientist in the later stage of an enjoyable career (Fig. 1). New technologies like GIS, multi-spectral imagery, high-resolution elevation data, etc. have emerged, while new colleagues have entered into Turtle's life. The new techniques and data fit right in with the tasks required of field soil scientists. However, new ways of thinking are required to utilize the new tools effectively. The latest chapters of this pedological tale explore these new technologies with a few new colleagues Turtle has gained. Turtle and Bear reviewing their latest GIS/RS soil mapping efforts. In Part III of the series, Turtle was advised by Eagle and Bear to carry on the study of the spatial component of soil mapping using the current tools of the geographic disciplines. Some of the more important topics discussed by the curious earth scientists follows. Eagle (E): What is a soil map? Turtle & Bear (T&B): At the most basic level, a “good” soil map groups soils geographically that are similar and separates soils geographically that are different. All: What could be improved? All: Soil map accuracy and precision could be improved by applying digital tools to a traditionally analog process, within the context of understanding the soil–landscape relationships. E: Expand on that. B: As Hudson (8) stated, it is the predictable pattern of soil occurrence that has allowed us to map large acreages with minimal observations of the soil itself. The process of consistently identifying similar landform segments is what has and remains to be the most difficult step in producing a good soil map using analog techniques. For example, the ability to identify areas that have concave profile curvatures in lower side slope positions with slopes less than 5% is a straightforward proposition using GIS/RS software. In addition, the results would be repeatable and easily expressed to end users in the form of rules, e.g., this soil occurs where profile curvatures are less than A, with a relative position between B and C on slopes less than D. We were confident in providing qualitative descriptions in the pre-GIS/RS days but were never confident that we could say that a particular setting as described was mapped with accuracy and precision within a soil survey project, let alone across soil survey projects. T: There is no question that if the present day data, software, and skills were available during my early days of mapping, I would have made better soil maps. I would also have been sure there was consistency among my crew. That is not a knock on the work we all did. We produced the best product possible given the tools available at the time. We had hand probes, sharp shooters, clinometers, USGS topos, and stereo paired aerial photos. We used the clinometers, topo maps, stereo pairs, and our walking feet for topographic control. We thought we had it made! The present day tools are more like a surgeon's scalpel, while the older tools were more like a stone ax. E: If Dukuchaev, Glinka, Hilgard, Marbut, or any of the early pedologists had GIS/RS tools in their day, I absolutely think they would have used them. I think the early soil surveyors that mapped a county over the course of one field season would have used these tools too. E: Why is slope class a criterion for mapping soils? T&B: Well, it is important to remember that the maps in a soil survey report are not soil type maps but are soil management maps. Map units are intended to delineate uniform management areas, not uniform soil areas. Slope gradient is an important potential limiting factor, and in the past, there wasn't much information related to things like slope gradient. It was not that long ago that USGS completed the 1:24,000 scale topographic quadrangle map series for the continental states. We remember when the 1-inch-to-the-mile topographic maps were common. Having a slope class phase was important as a matter of characterizing the landscape for the USDA as part of the soil survey program. At the same time, soil scientists were gaining a better understanding of the relationships between soil distribution and slope characteristics (e.g., soil landscape). In the 1950s, 1960s, and 1970s, regional soil landscape projects were carried out in various parts of the United States (Nettleton and Lynn, 18, 19; Fenton 6), with the goal of understanding soil landscape relationships to improve the speed, utility, and accuracy of soil survey. These regional studies were critical in the birth and acceleration of modern soil survey (Indorante et al., 11). The history and impact of these studies is recorded in the legacy soil maps, soil classification, and soil surveys in the areas of the respective studies, as well as in refereed scientific literature. T: Historically, a soil map with slope classes provided a huge amount of information that was desired for characterizing soil and land resources for the nation. This information was not available anywhere else as a synthesized product. In addition, slope class became very important as USDA programs began to put conservation requirements on some agricultural producers starting in the mid-1980s. Over time, it was discovered that slope was a very strong predictor of soil distribution. The soil-forming glasses that Dr. Owl prescribed helped me focus on soil–landscape relationships that helped segment the landscape, grouping similar soils and separating dissimilar soils geographically. Dr. Owl emphasized the importance put on the proper use of the genetic glasses with flip-down taxonomic glasses. Use the pedogenic glasses first, and then use the flip-down taxonomic glasses—flipping back and forth between the glasses helped to make a good soil map. E: That explains why it was done in the past, but what about present day, considering the increased availability of high-resolution elevation data derived from LiDAR? Wearing glasses to map soils seems so yesterday. T: All of the new GIS/RS tools are just modern technology versions of the original pedogenic glasses prescribed by Owl, optometrist in the town of Even. The “earth” glasses that let us see GIS/RS data are a much more powerful prescription, but like any prescription, they can be used properly or abused. E: What do you mean by used or abused? T: Take the slope (topographic) factor in the pedogenic glasses prescription. My glasses were able to identify landscape, landform, and landform component level units. The current GIS/RS data layers (e.g., LiDAR) are very detailed, and it is possible to focus too closely on the micro-scale topographic variability at the expense of viewing the connectivity and continuity and cross-scale relationships of natural landscapes (Roecker and Thompson, 20). There is a distinct possibility that the “hole mappers” of traditional soil survey could now morph into “pixel mappers” of the current era of digital soil survey. E: So it is possible to have too strong a prescription? T: For now, yes. I think that Bear has something to say on this subject. B: It is a given that slope gradient is a factor in soil genesis. It helps us group similar landscapes and landforms, which in turn, helps us separate and group soils accordingly. All soil series have slope parameters as a range in characteristics, with some series confined to the low or high end of the slope spectrum and others spanning a wide range of slopes. After our many talks with you Eagle, we think it would be preferable to dispense with slope phase and just map the soil series or class. When we develop slope maps from high-resolution DEM and compare them to our legacy polygon slope classes, we are often discouraged. We can see many areas defined by the slope class maps that are outside of the slope classes defined by soil polygons. It has been common to see discrepancies occupying 40 to 50% of the area when we have compared SSURGO slope classes to slope classes derived from high-resolution DEM. At first, we thought the data were suspect, but we can't recall many cases where the LiDAR was wrong and we were correct. We think the improved data will allow us to shift to a higher level of precision for slope estimation, improving mapping consistency. A soil scientist is no match for accurately determining and mapping slope gradient when compared with LiDAR. Once upon a time, we soil scientists could feel pretty good about our abilities to define and delineate slope classes with our stereo photo pairs and clinometers when compared with a 30-m DEM. We can no longer make that claim when dealing with LiDAR, and the more we think about it, the more convinced we are that slope phases just complicate the soil map. E: Are the slope class phases the problem or is it the map unit polygon data model? T&B: A little of both, since they are related. The map unit polygon data model was the only option available in the past and has proven to be quite useful. We have discussed the imprecise nature of delineating polygons. Cartographic generalization is inherent with the map unit polygon model. There is also the issue of what is conveyed to soil survey users by the hard boundaries of a polygon, which soil scientists have accommodated with descriptions and lists of components present within delineations. Users would like to see more precisely where these components occur across the landscape, and the polygon model is not the best data model for providing these potential details. However, it bears repeating that these polygon soil maps were created to convey information on management interpretations to the users, not the precise locations of specific soil types or the precise values of specific soil properties. As for the slope classes, soil scientists were often asked to generalize slope classes, thereby compromising the task of mapping a natural landform segment. Given the increased availability of high-resolution DEM, it would be possible to dispense with slope class phase altogether. An end user could group slopes any way they see fit using DEM and combine them with a raster soil map. In fact, a new raster-based system could be developed that uses raster versions of all variables related to particular interpretation and output cell-based ratings or interpretations. E: That is a big change, but I think I can see this…. T&B: Yes it is, but data and software available today provides a tremendous opportunity that was not available in the past. Legacy or current soil surveys were developed as a data synthesizer, since there were no other sources available. Now that new tools are available that perform the job better, the need to synthesize things like slope class equally for all map unit polygons is less important and may not be appropriate to include in legacy soil geographic databases. E: There would need to be an educational effort in how to use a soil map that may be more detailed compared with a map unit polygon map. There are different scales of precision and levels of confidence for each variable used in an interpretation or rating. The system of fuzzy logic currently used for soil interpretations provides a nice framework for continuing something like a completely raster-based system. New terms will also need to be developed and defined to enhance communication among soil scientists and soil survey users. T&B: That could work well. As you said, this would be a big change from the current system, but it should be a change that provides great benefits to the user. Furthermore, there is already some evidence that users are ready for raster soil maps (Grunwald et al., 7). E: We have discussed the raster data model for some time. Would it work? T&B: The raster data model seems to be the next logical step. Large amounts data in standard formats are becoming available in raster form like imagery, elevation, LiDAR vegetation derivatives (canopy height, biomass, and vegetative density), electromagnetic induction, gamma spectrometry, and climate, to name a few. The use of data like these for mapping soil classes and properties is well documented (McBratney, et.al, 15). Keeping the output data in the same format as the input makes sense. Raster soil survey products have been reviewed favorably (Grunwald, et.al. 7). We have a 100-plus year history using the polygon model and a vast infrastructure designed to support it, so it will take a period of time to transition from our current polygon model to a raster model, especially when considering the associated database. E: Speaking of the attribute database, how valuable is a perfect database if the geographic representation of soils on the ground is flawed? T&B: Both are required to have the best product possible. With regard to data management, there is a tendency to treat the spatial and tabular data as separate entities. A large emphasis has been placed on the tabular, and rightly so, but the spatial has often been ignored or neglected. E: I always chuckle at the old soil mapping analogy, “would you rather map soils on the back side or the front side of an aerial photo?” Does that apply in the present day? T&B: That analogy is timeless. Aerial photography and stereo-photography, in particular, revolutionized soil mapping. Much can be inferred from an aerial photo, and it was an indispensable tool that helped us inventory much of the USA. Given the tools available today, spatial data processed with GIS/RS can be considered the “front side” of a symbolic aerial photo while a plain orthophoto occupies the “back side.” That statement is a little extreme, but the capability to synthesize, categorize, and classify these various raster datasets for the purpose of mapping soils should not be underestimated. T&B: These GIS/RS-based tools could help us convey soil–landform relationships to soil scientists more effectively than oral tradition and on-job training. E: You mean arm waving only goes so far? T&B: Right, when you can describe something qualitatively in the office and show it in the field, it is effective. When the qualitative is defined quantitatively, mapped in the office, downloaded to a device, and taken to the field, it becomes even more effective. E: Don't forget the inverse operation: collecting georeferenced data in the field and coupling it with data in the office to help quantify the setting of soils. T: Good point. The ability to collect, define, and “project” one's understanding of the soil-forming environment is something we could not conveniently do in the past. The software today is more suited to that task than blackboards or whiteboards. E: Will these powerful tools and data without geopolitical limits eliminate field work? T: No. In the past, we would sit over a stereoscope to map our lines. That time can now be spent developing relationships with our data and documenting soil distribution using GIS/RS. B: The stereoscope was only effective when the user above the scope had an understanding of soil–landform relationships. GIS/RS is the same way. When GIS/RS is coupled with a knowledgeable soil scientist, it can be thought of as a smart stereoscope. An understanding of soil–landform relationships is required to make effective use of GIS/RS. E: You make GIS/RS sound too good to be true. T&B: We don't mean to, and we are aware that there is no silver bullet. We have both mapped in areas that are unpredictable, and using these techniques may be more time consuming than doing things the traditional way. B: There have also been situations where the soil–landscape relationships may be able to be determined and mapped via GIS/RS, but their extent is so small that by the time you figure out the soil–landscape relationships, it is effectively mapped. T: Some of the GIS/RS work I see just looks like a pretty map. Is it really useful? B&E: There is no question we can produce eye-pleasing maps with GIS/RS. But, which is the “pretty map”—one produced under a stereoscope that has been redrafted two or three times to produce “cartographically” pretty polygons on its way to becoming legacy SSURGO, or a map with quantitatively defined extents and settings produced using GIS/RS? Neither procedure advocates more or less field work. Both procedures require the same understanding of soil–landform relationships. The difference is the GIS/RS procedure can be explained, defined, and reproduced by others. The polygon map can be explained but falls short when it comes to definition and reproduction. If we assume a common understanding of soil landform relationships, the tasks becomes determining the most effective means to “project” that knowledge spatially. GIS/RS is the tool designed to facilitate this task. Not using GIS/RS is akin to playing darts while one's eyes are blindfolded. T&B: What about measures of accuracy and confidence with these new techniques? E: What metrics are available for the present SSURGO? T&B: Nothing stated in the tradition of accuracy assessments or confusion matrices exists for SSURGO. The reason for that, in our experience, is the practice did not exist for much of the time period when the soil survey was taking place. In addition, there is an assumption in most soil surveys that every delineation was visited and the named soil was observed within the confines of the polygon. Some soil survey reports presented information summarizing map unit composition based on transect data (Doolittle, et al, 4; Linsemier, 12). There are many examples in the literature related to assessing the variability and accuracy of conventional soil maps (Amos and Whiteside, 1; Brown, 3; Wilding and Drees, 22; Edmonds and Lentner, 5; Mausbach and Wilding, 14). The component table provides data related to map unit composition, and there are ranges provided for the physical and chemical properties. E: I suppose that gets at the issue a bit, but these new techniques typically have the development of an accuracy assessment assumed as part of the process (Malone et al., 13; Nauman and Thompson, 16). It is a more straightforward way of reporting the information than SSURGO. T&B: Our procedures have been in place for a long time, but there is no reason accuracy assessments could not be incorporated in the future products. E: Regarding the ranges of physical and chemical properties, I've always wondered what a representative value (RV) is, how it is determined, and if everyone across the USA uses the same criteria to make that determination. T&B: Very good question. The RV is not defined statistically. As a result, it may not be consistently defined or applied across the country. E: What if these were defined statistically so that the data could be populated and interpreted consistently? T&B: Indeed, if we have enough data, we use something like the mean and two standard deviations for the RV, low and high values. When we don't have enough data, we have to rely on our best professional judgment and expert knowledge. We are open to suggestions. E: I like your option when you have enough data but only if the data are normally distributed. Another option could be as simple as selecting the median as the RV and the low and high ranges as a quantile, like 20 and 80%. It would sure make it easier to interpret for users like me. T&B: It would definitely make it easier for us to explain to users too. E: Get back to me when that gets resolved. T&B: Well, we do edit our tabular data quite a bit, so maybe we can get that implemented. T: It is too bad all of these neat tools are coming at a time when the inventory is nearly complete. B&E: Well, do you throw your knife away when it gets dull, or do you sharpen it again? T: That is true; refining the product we have created sounds good. B&E: It is always a challenge to improve an existing product. The existing product will be a huge knowledge base for new maps. Given the choice, a violinist would rather fix a Stradivarius than buy a brand new violin. When you think about it, there is no data set in the world that can match the NRCS soil survey for the extent of coverage, level of detail, or intensity of observation. There is an excellent foundation to build upon. This quote by R.S. Smith, Director of the Illinois Soil Survey, on 27 Sept. 1928 is as appropriate now as it was back then: “I hope the answer to your question is clearly indicated in what I have written. It is that the soil survey will never be completed because I cannot conceive of the time when knowledge of soils will be complete. Our expectation is that our successors will build on what has been done, as we are building on the work of our predecessors” (Smith and Wascher, History of Illinois Soil Survey, Department of Agronomy, University of Illinois at Urbana-Champaign, unpublished, 1967). T: How would it work? B&E: Like any new endeavor, the first step is to think a bit, plan a bit, do a few trials, evaluate the results, and then repeat. It will take a few years to work out the details, but there are examples of methods being developed that incorporate legacy maps as inputs for the DSM process (Nauman and Thompson, 16, Nauman et al., 17). Once that is done, a reasonable protocol could be implemented for others to follow. We believe harnessing these new tools for evaluating the soil–landscape will help to improve our understanding of soil systems. T: Who wants to map and refine from the office? B&E: We all want to get out more often. There will always need to be time devoted to field work to build and verify the work done on the computer. These tools offer the potential to make our field work more targeted and efficient. However, there is no way these tools preclude field work. A soil scientist is required to make it all come together. One example of such work is what is being called disaggregation. The result of a disaggregated soil map is a more refined representation of the soil–landform relationships (Nauman and Thompson, 16; Nauman et al., 17). T: Maybe we will be able to spatially represent what we currently just list in the component table. B: That would be the good goal. E: Since we are talking about refining, I will be a little provocative and ask a somewhat loaded question. When I look at the plethora of soil series and read the Official Series Description, I would ask you to explain, “What are the differences between the hypothetical James, Seamus, Giacomo, Jacob, Hagop, and Jaime soil series?” T&B: We thought this was going to be a friendly discussion! The ability to split on minutia may exceed our ability to explain why it is important to do so or if the splits are functional pedogenic breaks that make sense and are predictably identified on the landscape. Brevik and Hartemink (2) show the growth of soil series over time (Fig. 2). No doubt, mapping in new landscapes during the acceleration of the soil survey added to the numbers, but at some point, it may be helpful to see if these series are actually unique, or if some could be aggregated, deactivated, combined, etc. The growth of soil series in the USA over time. Figure courtesy of Dylan Beaudette, USDA-NRCS, California Soil Resource Lab. E: You know I like spending time with you, but it can get frustrating to not get a straight answer on something like “What series is this—delta or alpha?” and have you answer, “Well, this is kind of like a deltalpa.” Make a decision. T&B: We do that to get your goat most of the time, but you have a point. E: By the way, the soil series is a powerful concept that I hope is maintained. T&B: We agree, and we need to talk about that when we have more time. E: I don't know about you, but I've had enough yapping—let's head the field, so we can get educated, eh? T&B: You bet—make sure you bring the “earth” glasses. Adopting and using GIS/RS, GPS, and data to better understand the distribution, pattern, genesis, and behavior of soils has enhanced the endeavor for Turtle and Bear. The technology required a learning curve, but it did not get in the way of the process. The many methods available for data mining and classification will ensure the need to stay abreast of new techniques and continued exploration of their respective strengths and weaknesses for use in soil survey operations. The pedogenic glasses that Dr. Owl prescribed Turtle 28 years ago to help make soil maps are still useful, but the more modern and powerful tools from Eagle and Dr. Owl help Turtle do a better, more consistent job. Oh, and one more thing. A pair of reading glasses from Dr. Owl (Fig. 3) in the town of Even can still come in very handy when reading Chapter 5 of Soil Taxonomy–Application of Soil Taxonomy to Soil Surveys (Soil Survey Staff, 21). The traditional “hole mappers” and the more recent “pixel mappers” are likely to become “whole mappers,” after a good read. Dr. Owl, the optometrist, in his office in the town of Even Numbers (Indorante, 10).