Academia.eduAcademia.edu

Figure 4 – uploaded by J. Stoorvogel

See full PDFdownloadDownload figure
Figure 3. Legacy and auxiliary data used for S-World (sources are presented in the text) using Rwanda as an example. This figure is available in colour online at wileyonlinelibrary.com/journal/ldr

Figure 3 Legacy and auxiliary data used for S-World (sources are presented in the text) using Rwanda as an example. This figure is available in colour online at wileyonlinelibrary.com/journal/ldr

Related Figures (10)

Figure 1. The conceptual model of S-World with its 5 main steps. This fig- ure is available in colour online at wileyonlinelibrary.com/journal/Idr
Figure 1. The conceptual model of S-World with its 5 main steps. This fig- ure is available in colour online at wileyonlinelibrary.com/journal/Idr
Figure 2. The process of disaggregation of complex mapping units in S-World. This figure is available in colour online at wileyonlinelibrary. com/journal/Idr
Figure 2. The process of disaggregation of complex mapping units in S-World. This figure is available in colour online at wileyonlinelibrary. com/journal/Idr
In which cj; is a constant that indicates the relative importance of the landscape property j for soil property i. Landscape property j on location x is described by p;,. The sign of c, indicates whether there is a positive or negative relationship between the landscape property and the soil The weight w;;, for landscape property / is calculated as:
In which cj; is a constant that indicates the relative importance of the landscape property j for soil property i. Landscape property j on location x is described by p;,. The sign of c, indicates whether there is a positive or negative relationship between the landscape property and the soil The weight w;;, for landscape property / is calculated as:
Table I. Reclassification of the GLOBCOV database in terms of land use intensity the vegetation will provide organic matter to the soil. Land cover is described by the Normalized Difference Vegetation Index (NDVI). Five years (2006-2010) of monthly 30 arc second resolution SPOT NDVI images (http://free.vgt.vito. be/) were downloaded from which an average vegetation in- dex was calculated. The average NDVI was used as a proxy for the vegetative cover of the soil. Illustrations of the input data for the Rwanda example are presented in Figure 3.
Table I. Reclassification of the GLOBCOV database in terms of land use intensity the vegetation will provide organic matter to the soil. Land cover is described by the Normalized Difference Vegetation Index (NDVI). Five years (2006-2010) of monthly 30 arc second resolution SPOT NDVI images (http://free.vgt.vito. be/) were downloaded from which an average vegetation in- dex was calculated. The average NDVI was used as a proxy for the vegetative cover of the soil. Illustrations of the input data for the Rwanda example are presented in Figure 3.
Table Il. Summary statistics from the analysis of soil properties over the different soil types based on the WISE database
Table Il. Summary statistics from the analysis of soil properties over the different soil types based on the WISE database
Figure 4. Disaggregated soil map for Rwanda with the lines representing the original delineated areas of the HWSD. This figure is available in colour online at wileyonlinelibrary.com/journal/Idr
Figure 4. Disaggregated soil map for Rwanda with the lines representing the original delineated areas of the HWSD. This figure is available in colour online at wileyonlinelibrary.com/journal/Idr
Table II. Modelled variation in landscape properties for the humic and haplic Acrisols at the global level and within Rwanda “SPOT expresses NDVI as an index value between 0 and 255. The values were truncated to the 50-200 range since the low and large values represent little variation in vegetation.
Table II. Modelled variation in landscape properties for the humic and haplic Acrisols at the global level and within Rwanda “SPOT expresses NDVI as an index value between 0 and 255. The values were truncated to the 50-200 range since the low and large values represent little variation in vegetation.
Table IV. Relative weights for different landscape properties on their impact on 7 different soil properties estimated using a literature surve and expert knowledge
Table IV. Relative weights for different landscape properties on their impact on 7 different soil properties estimated using a literature surve and expert knowledge
soil databases. Typically the validation requires a soil data base with a proper global coverage, that is independent to re gional or national surveys and that provides data at the same resolution as the data provided by S-World (30 arc seconds) Data at the 30 arc second resolution are extremely difficul to find, but one could establish an independent dataset o: point observations and use that to validate the S-World data base. To evaluate the improvement of the S-World database we established a single database from independent studie: from Costa Rica (n=193), Ecuador (n=95), Egyp (n=22), Ethiopia (n=67), Ivory Coast (n=31), Kenyz (n=256), Morocco (n=67), Mozambique (n=108) Netherlands (n=127), Peru (n=93), Philippines (n= 124) Russia (n=178), Senegal (n=110), Uganda (n=54), USA (n=120), and Vietnam (n=45) (Table $2). All 169( georeferenced observations included data on topsoil organic carbon or organic matter. We compared the samples to the predictions of S-World and the HWSD by calculating the Root Mean Squared Difference. The HWSD provides average SOC contents for the soil types in the mapping units but the soil types are not geo-referenced within the mapping units. Two different approaches were followed: i) the dominant soil type for the mapping unit (HWSDgominant) and ii) the weighted average for all soil types within the mapping unit (HWSD cigntea). The RMSD for S-World wa: considerably smaller (2.1% SOC) than the RMSD fo: HWSD veigntea (2.9% SOC) and HWSDygominant (3.6% SOC) However, still large deviations were found despite the fac that there were almost no organic soils in the validatior dataset. Notwithstanding some of the limitations of the validation, it suggests that S-World improves the estimate: of SOC contents. Figure 5. S-World results for modelled clay (%) for the world and Rwanda (white areas represent areas with no data as there is no soil). This figure is available in colour online at wileyonlinelibrary.com/journal/Idr Validation of global databases like S-World is extremely difficult due to i) the global scale, ii) the support of the data, and iii) independent data. Many existing datasets have been used for the development of national soil surveys which form the basis of the HWSD. At the same time S-World pre- dicts soil properties at the support of grid cells whereas data are typically observed for individual points. Nevertheless there is an urgent call to provide quality assessment to global
soil databases. Typically the validation requires a soil data base with a proper global coverage, that is independent to re gional or national surveys and that provides data at the same resolution as the data provided by S-World (30 arc seconds) Data at the 30 arc second resolution are extremely difficul to find, but one could establish an independent dataset o: point observations and use that to validate the S-World data base. To evaluate the improvement of the S-World database we established a single database from independent studie: from Costa Rica (n=193), Ecuador (n=95), Egyp (n=22), Ethiopia (n=67), Ivory Coast (n=31), Kenyz (n=256), Morocco (n=67), Mozambique (n=108) Netherlands (n=127), Peru (n=93), Philippines (n= 124) Russia (n=178), Senegal (n=110), Uganda (n=54), USA (n=120), and Vietnam (n=45) (Table $2). All 169( georeferenced observations included data on topsoil organic carbon or organic matter. We compared the samples to the predictions of S-World and the HWSD by calculating the Root Mean Squared Difference. The HWSD provides average SOC contents for the soil types in the mapping units but the soil types are not geo-referenced within the mapping units. Two different approaches were followed: i) the dominant soil type for the mapping unit (HWSDgominant) and ii) the weighted average for all soil types within the mapping unit (HWSD cigntea). The RMSD for S-World wa: considerably smaller (2.1% SOC) than the RMSD fo: HWSD veigntea (2.9% SOC) and HWSDygominant (3.6% SOC) However, still large deviations were found despite the fac that there were almost no organic soils in the validatior dataset. Notwithstanding some of the limitations of the validation, it suggests that S-World improves the estimate: of SOC contents. Figure 5. S-World results for modelled clay (%) for the world and Rwanda (white areas represent areas with no data as there is no soil). This figure is available in colour online at wileyonlinelibrary.com/journal/Idr Validation of global databases like S-World is extremely difficult due to i) the global scale, ii) the support of the data, and iii) independent data. Many existing datasets have been used for the development of national soil surveys which form the basis of the HWSD. At the same time S-World pre- dicts soil properties at the support of grid cells whereas data are typically observed for individual points. Nevertheless there is an urgent call to provide quality assessment to global
Figure 6. S-World results for modelled SOC (%) for the world and Rwanda (topsoil) (white areas represent areas with no data as there is no soil). This figure is available in colour online at wileyonlinelibrary.com/journal/ldr
Figure 6. S-World results for modelled SOC (%) for the world and Rwanda (topsoil) (white areas represent areas with no data as there is no soil). This figure is available in colour online at wileyonlinelibrary.com/journal/ldr
Related topics:
Soil ScienceSoil Survey and Land EvaluationGlobal Soil MapSoil Survey
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved