SSSAJ Journal of Natural Resources and Life Sciences Education
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Published online 21 June 2006
Published in Soil Sci Soc Am J 70:1377-1386 (2006)
DOI: 10.2136/sssaj2004.0165
© 2006 Soil Science Society of America
677 S. Segoe Rd., Madison, WI 53711 USA
This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via ISI Web of Science (2)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Zhao, Y.
Right arrow Articles by Wang, H.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Zhao, Y.
Right arrow Articles by Wang, H.
Agricola
Right arrow Articles by Zhao, Y.
Right arrow Articles by Wang, H.
Related Collections
Right arrow Carbon Sequestration
Right arrow Hillslope Analysis
Right arrow Pedology

Pedology

Map Scale Effects on Soil Organic Carbon Stock Estimation in North China

Yongcun Zhaoa, Xuezheng Shia, David C. Weindorfb,*, Dongsheng Yua, Weixia Suna and Hongjie Wanga

a State Key Lab. of Soil and Sustainable Agriculture, Inst. of Soil Science, Chinese Academy of Sciences, P.O. Box 821, Nanjing 210008, China
b Texas Agricultural Experiment Station, 1229 N. US Hwy 281, Stephenville, TX 76401

* Corresponding author (weindorf{at}tarleton.edu)


   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Digital soil maps of different scales have been compiled in China, but exactly how map scale affects the estimation of regional SOC (soil organic carbon) stocks remains unclear. To test the effect, median, mean, and a pedological professional knowledge based method (PKB) were used to link soil profiles to soil maps at five scales ranging from 1:500 000 to 1:10 000 000 for the Hebei Province. Excluding the 1:4000 000 soil map, SOC stocks decreased as the map scale decreased. The estimated SOC stocks obtained using the mean were always higher than those using the median or PKB method. The changes in estimation due to different map scales and linking methods affected the process of assigning SOCD (soil organic carbon density) values to digital soil surveys. The differences in SOCD values resulted from the change in the total nonurban land area of each soil type as a result of the different methods and scales of maps used in the regional SOC stock estimation process.

Abbreviations: GSCC, genetic soil classification of China • PKB, pedological professional knowledge based method • SOC, soil organic carbon • SOCD, soil organic carbon density


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
THE TERRESTRIAL ENVIRONMENT is the most important living space for human beings and soils are at the core of the terrestrial ecosystem. Soil C stocks are nearly three times as large as stocks in the vegetation of terrestrial ecosystems (Post et al., 1990) and twice as large as stocks in the atmosphere (Eswaran et al., 1993). Changes in SOC will affect the density of greenhouse gases in the atmosphere, possibly leading to global climate change (Lal et al., 1998; Uri, 2000). Accurately quantifying SOC stores in soils is considered fundamental to global climate change modeling (Tan and Shibasaki, 2003; Janzen, 2004).

Correlating the SOCD (soil organic carbon density) values of each soil to each corresponding map unit component is the first step in producing regional SOC distribution maps at county to continental scales. Map units are extrapolated spatial delineations of average soil properties. Map unit properties are estimated from the composition percentage of major components and representative soil sampling of those components. Maps created from such techniques permit the calculation of the area-weighted estimates of SOC (Kern, 1994; Homann et al., 1998); however, the choice of soil map scales used in the estimation of regional SOC may lead to uncertainty (Arnold, 1995) because map delineations and map unit composition vary with scale. Moreover, soils large in area and extensive in distribution on large-scaled soil maps are likely to be major components of smaller scale general soil map units, while the soils small in distribution area on large-scale maps are in danger of being eliminated or merged into other map units during the map generalization process (Rapalee et al., 1998; Galbraith et al., 2003). Kern (1994) used three maps of scales ranging from 1:5 000 000 to a scale with resolution of 0.5° in longitude and latitude to estimate SOCD for the contiguous USA. Kern (1994) reported that the mean of SOCD increased as map scale decreased, but those results were influenced by the use of different data sources and map unit compositions at each map scale. Davidson and Lefebvre (1993) reported that SOC increased as map scale decreased. They used the same NRCS-SIR (Soil Interpretation Record) dataset on 1:24 000, 1:250 000, and 1:5 000 000 maps. The three scales had increasing compositions of Histosols and Spodosols. Homann et al. (1998) applied pedon data and NRCS-SIR data to soil maps of different scales to estimate regional SOCD, and they obtained a decrease in SOCD as map scale decreased. The results compared different data sources and different map unit compositions at different map scales. Galbraith et al. (2003) estimated the regional SOCD for the Tughill Plateau of northern New York on the basis of 1:15840, 1:62 500, 1:250 000, and 1:750 000 soil maps using a single dataset. They also found that SOCD increased because the composition of higher C soils was higher on smaller scale maps. In each of these studies, another very important factor is the method of aggregating soil profile properties to represent mapped areas (delineations) when soil maps of multiple scales are used in the estimation. Results obtained by Galbraith et al. (2003) indicate that the variability of sampled SOC values within a specific soil series is often high and the question of whether the sampled soils truly represent the in situ SOC stocks is a great source of uncertainty, as also reported by Davidson and Lefebvre (1993) and Homann et al. (1998). Therefore, the method and source of soil SOC values and the methods of linking soil series and their associated properties to spatial information is very important in regional SOC studies and comparisons of differing maps and scales. Calculating the median or mean of multiple SOC values to represent the SOC value of each map unit component is the most frequently used method for linking soil properties to spatial information (Martial et al., 2002; Batjes, 2000).

Estimates of SOC stocks in China are commonly based on the soil profiles cited from the Soil Series of China (National Soil Survey Office, 1995) and the corresponding acreage of each soil type recorded therein (Pan, 1999; Wang et al., 2000; Li et al., 2001; Jin et al., 2001). With the development of GIS (geographic information system) technology, many researchers use the mean or median methods of linking soil properties of the 2473 profiles recorded in the Soil Series of China to estimate regional SOC stocks for the 1:4 000 000 digital soil map of China (Wang et al., 2002; Xie et al., 2004). The Institute of Soil Science, Chinese Academy of Sciences, has compiled a 1:1 000 000 digital soil map of China and 1:500 000 digital soil maps of several provinces. The soil properties associated with the digital maps include both the profiles recorded in Soil Series of China and the profiles recorded in provincial soil surveys. The linking method between soil properties and digital soil maps is based on a comprehensive pool of PKB (pedological professional knowledge base), as proposed by Shi et al. (2004a). There has not been a published systematic study on the influence of map scale on the estimation of regional SOC based on the characteristics of the soils of China. In this study, three kinds of linking methods, median, mean, and PKB, and soil profiles from Ding (1992) were applied to soil maps of five different scales to investigate the potential influence of soil map scale on the estimation of regional SOC stocks in the study area.


   MATERIALS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Study Area
Hebei Province surrounds Beijing, the capital of China (Fig. 1), and is representative of northern China. The topography of the province is characterized by high mountains and tablelands in the northwest and low plains in the southeast. The mountains and tablelands are mostly >1000 m in elevation with some peaks exceeding 2000 m, while the plains are only 3 to 5 m above sea level.


Figure 1
View larger version (37K):
[in this window]
[in a new window]
  		Fig. 1. Geographical location of the study area—Hebei Province, China.

 
The total land area of Hebei Province is 187693 km2 (Ding, 1992). The tablelands occupy 9.3% of the province, the mountains 49.5%, and the plains 41.2%. With a temperate continental monsoon climate, Hebei Province has a mean annual temperature ranging from 0 to 13°C and annual precipitation between 300 and 800 mm. Significant differences in climate, geomorphology, geohydrological conditions, and parent materials throughout the province result in a great variety of vegetation and soil types and distribution patterns.

Data Sources and Aggregation Methods
The soil attributes from a total of 363 soil profiles were taken from Ding (1992), 72 of which were also recorded in the Soil Series of China. Soil profiles in the Soil Series of China were considered the most representative soils in the different regions of China and were classified according to the GSCC (genetic soil classification of China; Shi et al., 2004b). This information was derived from the second national soil survey of China and is the most comprehensive and detailed study on soil characteristics in Hebei Province, and so these 363 profiles were selected to estimate SOC stocks in this study. The spatial soil properties are from maps of five different scales ranging from 1:500 000 to 1:10 000 000. The basic characteristics of the soil maps are presented in Table 1.


View this table:
[in this window]
[in a new window]
  		Table 1. Characteristics of the soil maps used to estimate regional soil organic C.

 
The soil map unit system of many Chinese soil maps is based on the classification units in the GSCC (Shi et al., 2004b). In general, the classification level of map units increases as map scale decreases. Soil maps used in this study were all based on the GSCC. Soil maps at scales of 1:500 000, 1:1 000 000, 1:2 500 000 and 1:10 000 000 were compiled during the second national soil survey of China. Because the 1:4,000,000 scale map was compiled in 1978 and its map unit system is based on an older version of GSCC, however, the names of several map units were correlated with those on the more recent soil maps. Several very representative sample profile locations were labeled on the 1:500 000 scale map, but were replaced by more general profiles on smaller scale maps. Detailed information about the sample profiles, including the location of profiles, was recorded in the soil survey reports in Ding (1992) and National Soil Survey Office (1995).

Estimation of Profile Soil Organic Carbon Density
For a soil profile with the depth of D (cm), the total SOC by volume, SOCDD (kg C m–2), is given as (Wang et al., 2003; Kazuhito et al., 2004)

Formula 1[1]
where n is the number of pedogenic horizons in the soil survey, {delta}I% represents the volumetric percentage of the fraction >2 mm (rock fragments), {rho}i is the bulk density (g cm–3), Ci is the organic C content (g kg–1), and Ti represents the thickness (cm) of the layer i. The organic C content is calculated by multiplying soil organic matter content by 0.58 (the Bemmelen index), which converts organic matter concentration to organic C content (Wen, 1984). The SOC was estimated to a maximum depth of 1 m (Sun et al., 2003). For profiles with no bulk density value, the mean bulk density value of the corresponding depth in all profiles belonging to the same soil family was used.

Methods of Aggregating Profile Data to Represent Map Units
The soil profiles belonging to each map unit were examined according to the soil type name based on the GSCC. The SOCD values of the profiles with the same soil type name were linked to spatial delineations within corresponding map units using one of the following methods:

Median Method
The median of SOCD values of all profiles of the same soil type name within a map unit was calculated and assigned to the proportion of that soil type within that map unit.

Mean Method
The arithmetic mean of the SOCD values of all profiles of the same soil type name within a map unit was calculated and assigned to the proportion of that soil type within that map unit.

The spatial locations of the profiles were not considered in either the Median or Mean method. Both methods aggregate data by soil type name and province in the calculation of the SOCD values of map units, so different areas of the same map unit will get the same SOCD value, the median or mean SOCD value of all profiles of the province belonging to that map unit. Figure 2 shows the linking process using the median or mean method. For example, two profiles (A1 and A2) belong to Soil Family A. The median or mean SOCD value of Profile A1 and A2 will be associated with the entire area covered by Map Unit A delineations.


Figure 2
View larger version (12K):
[in this window]
[in a new window]
  		Fig. 2. Application of sampled data by soil type to all occurrences of delineations of that same soil type in a sample study area using the mean or median methods.

 
Pedological Professional Knowledge Based Method
The PKB method considers soil data from the same soil type name and the spatial location of the sampled profiles. The PKB links the soil properties of each profile to its corresponding delineations according to the identity or similarity in soil parent materials and distribution area recorded in provincial or regional soil series.

The second national soil survey of China was implemented based on the county rather than the provincial level. Linkage of SOCD values to delineations on soil maps of five scales used in this study was conducted by county with the PKB method, e.g., for the 1:1 000 000 scale soil map, all profiles within a county were used to link SOCD to soil delineations in the same county. Figure 3 shows the linking process using the PKB method. For example, the SOCD for Profile C1 sampled in Ren County was linked to two delineations, Ca and Cb, of Map Unit C within Ren County (one profile linked to multiple delineations). For Profiles B1 and B2, the mean SOCD was linked to only one delineation of Map Unit B (multiple profiles linked to one delineation). For Profiles A0, A1, A2, and A3, there were two delineations (Aa and Ab) of Map Unit A (multiple profiles linked to multiple delineations). Profiles A0, A1, and A2 were linked to Ab, while A3 was linked to Aa on bottom right corner of the map. For Profile A0 in Longrao County near the edge of Ab, the distance from A0 to the edge of the upper part of Ab was shorter than the distance to Aa. Thus, the mean SOCD value of Profiles A0, A1, and A2 were linked to Ab in Map Unit A, even though Profile A0 occurred in a different county than Profiles A1 and A2. Using the PKB method, delineations with different spatial locations but belonging to the same map unit may obtain different SOCD values, e.g., Aa and Ab of Map Unit A.


Figure 3
View larger version (21K):
[in this window]
[in a new window]
  		Fig. 3. Application of sampled data by soil type, identity, or similarity in soil parent materials, spatial location, and distribution area to all occurrences of delineations of that same soil type in a sample study area using the pedological professional knowledge based method.

 
All three linking methods and the 363 soil profiles in Ding (1992) were applied to soil maps of the five different scales to explore the overall effect of map scale on the estimation of SOC stocks in the study area.

Estimation of Regional Soil Organic Carbon Stocks
The regional SOC stock was calculated according to the methods used by Batjes and Dijkshoorn (1999). Water and urban areas are excluded from the calculation.


   RESULTS AND DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
CONCLUSIONS
 REFERENCES
 
Map Scale Effects on Regional Soil Organic Carbon Estimates
Similar trends in estimation of SOC stocks were observed for all three methods as map scale decreased from 1:500 000 to 1:10 000 000 scale (Fig. 4). The estimated SOC stocks for the study area decreased as the map scale decreased from 1:500 000 to 1:2 500 000 scale; however, SOC stocks estimated from the 1:4 000 000 maps were much higher than those for any other scale, regardless of linking method. The SOC stocks from the 1:10 000 000 scale maps were the lowest or second lowest for all linking methods. The mean linking method produced higher values than the median or PKB methods. The estimated results using the PKB method with 1:500 000 and 1:1 000 000 scale maps were lower than those obtained by the corresponding median method, but that trend reversed on soil maps of scales 1:2 500 000 to 1:10 000 000. There are few delineations per map unit at smaller scales (Table 1), so the PKB method approaches the mean methods of linking SOCD values (see Fig. 2 and 3); however, they are not identical because using the PKB method on different areas of a map unit may result in different SOCD values since the PKB method used a different dataset based on location rather than inclusion. The median method has lower values at small scales because the dataset is smaller and influenced more by a few low values that were the most frequent.


Figure 4
View larger version (16K):
[in this window]
[in a new window]
  		Fig. 4. The effect of soil map scale on the soil organic C stock estimates in the Hebei Province of China, using the mean, median, and PKB (pedological professional knowledge based) methods.

 
Soil Organic Carbon Stock Estimation Using Soil Maps of Different Scales at the Great Group Level
According to Batjes (2000), a more pronounced difference can be observed between the results of regional SOC stock when estimated at the GSCC great group level, which may explain the discrepancies between the regional SOC stocks in Fig. 4. The area percentage and SOCD (using the median linking method as an example) of soil great groups calculated from maps of five different scales are presented in Table 2. The soils in Table 2 are ranked from the top down in order of highest to lowest SOC stock calculated from the 1:500 000 scale soil map. Each GSCC great group is presented with the corresponding soil subgroup in the U.S. Soil Taxonomy to facilitate understanding (Shi et al., 2006). Composition of soil great groups in the Hebei Province varies from map to map. The SOC stocks in the Hebei Province are mainly controlled by three soil great groups, namely cinnamon soils, brown soils, and castanozems (Table 2, Fig. 5). Composition is similar between the three largest scale maps, but there is a major composition difference between the two small-scale maps and the three larger scale maps in 19 of 22 GSCC great groups. The crossover between the median and PKB methods at the 1:2 500 000 scale is due primarily to the larger increase in SOC stocks from cinnamon soils using the PKB method combined with a larger decrease in SOC stocks from brown soils using the median method. The large increase in SOC stocks at the 1:4 000 000 scale is due to the large 12% increase in SOC stocks from cinnamon soils, a 6.5% increase from castanozems, and a 6% decrease from brown soils (Fig. 5). These changes appear to be related to the changes in composition; there was a 12% increase in cinnamon soils, a 6.5% increase in castanozems, and a 6% decrease in brown soils. At the 1:10 000 000 scale, the overall decrease in SOC stocks is due primarily to a decrease in the composition of cinnamon soils and castanozems.


View this table:
[in this window]
[in a new window]
  		Table 2. The area percentage (AP) and soil organic C density (SOCD) variation of soil great groups on soil maps of different scales.

 

Figure 5
View larger version (25K):
[in this window]
[in a new window]
  		Fig. 5. The SOC (soil organic C) stock estimates for the soil great groups in the study area using the median, mean, and PKB (pedological professional knowledge based) methods and soil maps at five different scales.

 
Map Generalization Influence as Scale Decreases from 1:500 000 to 1:2 500 000
The influence of soil map scale on estimation of SOC stocks in the Hebei Province resulted mainly from the map generalization process, as concluded by Galbraith et al. (2003). The estimated SOC stocks for all map units of all brown soils are presented in Table 3 as an example of how the composition of GSCC great groups changed with scale. Differences between Chinese soil families and U.S. Soil Taxonomy subgroups are given in Table 4 for comparison. The most significant influence of map units on the estimates of SOC stocks for brown soils was due to changes in estimated SOC stocks for "coarse-dispersed brown soils," the difference being 21.91 Tg between the 1:500 000 and 1:1 000 000 scale map when the median method was applied.


View this table:
[in this window]
[in a new window]
  		Table 3. The estimates of soil organic C stocks for map units of brown soils (soil great group) as map scale decreased from 1:500 000 to 1:1 000 000 using the mean, median, and PKB (pedological professional knowledge based) methods.

 

View this table:
[in this window]
[in a new window]
  		Table 4. Differentiating characteristics of Chinese soil families within the same U.S. Soil Taxonomy subgroup.

 
Figure 6 shows the map generalization process for the coarse-dispersed brown soils map unit. The SOCD value (23.79 kg C m–2) was linked to all occurrences of the map unit coarse-dispersed brown soils when the median method was applied. Delineations with 293.50 Tg C on the 1:500 000 scale map for coarse-dispersed brown soils were maintained on the 1:1 000 000 scale map, whereas delineations with 85.28 Tg C on the 1:500 000 scale map were merged into the other map units such as diluvial brown soils, weakly developed brown soils (subgroup), etc., on the 1:1 000 000 scale map. Moreover, delineations with 62.59 Tg C on the 1:1 000 000 scale map were from the other map units of the 1:500 000 scale map, inherited during the map generalization process for the coarse-dispersed brown soils as map scale decreased from 1:500 000 to 1:1 000 000.


Figure 6
View larger version (33K):
[in this window]
[in a new window]
  		Fig. 6. Changes in the estimates of SOC (soil organic C) stocks using the median method for coarse-dispersed brown soils (soil family) as map scale decreases from 1:500 000 to 1:1 000 000. Arrows indicate the map generalization process as map scale decreased, and the two SOC stock values in each shaded box were based on some delineations of a map unit being preserved while others were merged into other map units during the map generalization process. For example, after the overlay analysis of the 1:500 000 and 1:1 000 000 soil maps using GIS (geographic information system) software, 12 337 km2 (293.50 Tg C/23.79 kg C m–2) of coarse-dispersed brown soils on the 1:500 000 scale soil map were preserved at the same locations of the 1:1 000 000 scale soil map; 3585 km2 (85.28 Tg C/23.79 kg C m–2) of coarse-dispersed brown soils on the 1:500 000 scale soil map were merged into delineations of diluvial brown soils, weakly developed brown soils, etc., on the 1:1 000 000 scale soil map; and 2631 km2 (62.59 Tg C/23.79 kg C m–2) of mountain prairie–meadow soils, dark-compacted brown soils, etc., on the 1:500 000 scale soil map were merged into coarse-dispersed brown soils on the 1:1 000 000 soil map during the map generalization process.

 
As the map scale decreased from 1:1 000 000 to 1:2 500 000 (Table 5), the subgroup aquic brown soils on the 1:1 000 000 scale map was eliminated and emerged as "brown soils" (Fig. 7). On the 1:2 500 000 scale map, the drop in SOC stocks for brown soils and weakly developed brown soils using the median method was drastic, explaining the reversal with the PKB-method SOC stocks (Fig. 4). The gap between the mean and the PKB methods decreased at the 1:4 000 000 scale, probably also due to the larger drop in SOC stocks in the brown soils and weakly developed brown soils for the mean than for the PKB method.


View this table:
[in this window]
[in a new window]
  		Table 5. Estimates of SOC (soil organic C) stocks for subgroups of brown soils (soil great group) using the 1:1 000 000 scale map and map units of brown soils using the 1:2 500 000 scale map, using the mean, median, and PKB (pedological professional knowledge based) methods.

 

Figure 7
View larger version (22K):
[in this window]
[in a new window]
  		Fig. 7. Changes in the estimates of SOC (soil organic C) stocks using the median method for brown soils (subgroup) as map scale decreases from 1:1 000 000 to 1:2 500 000. Arrows indicate the map generalization process as map scale decreased, and the two SOC stock values in each shaded box were based on some delineations of a map unit being preserved while others were merged into other map units during the map generalization process.

 
From the 1:1 000 000 scale map, 15541.6 km2 (308.19 Tg/19.83 kg C m–2 x 103) of brown soils were maintained on the 1:2 500 000 scale map (Fig. 7). Delineations with 85.43 Tg C on 1:1 000 000 scale map were merged into the other map units of the 1:2 500 000 scale map, i.e., cinnamon soils, calcareous cinnamon soils, etc.; however, delineations with 104.50 Tg C on the 1:2 500 000 scale map came from subgroups of other than brown soils on the 1:1 000 000 scale map during the map generalization process of brown soils, i.e., aquic brown soils, weakly developed brown soils, etc. Delineations of brown soils increased as the map scale decreased from 1:1 000 000 to 1:2 500 000, whereas the SOC stocks for the map unit "brown soils" on the 1:2 500 000 scale map were still less than the estimated SOC stocks for brown soils on the 1:1 000 000 scale map. This was mainly because the brown soil delineations on the 1:1 000 000 and 1:2 500 000 scale maps had different SOCD values. For the 1:1 000 000 scale map, the map unit of brown soils (subgroup) is the soil family and the median of the SOCD values of profiles (soil series) belong to the same map unit (soil family) of brown soils. These were linked to all delineations of that map unit. The area-weighted mean of the SOCD of brown soils (subgroup) was then calculated at 19.83 kg C m–2; however, the map unit of the 1:2 500 000 scale map represents the subgroup, not the soil family. This suggests that the median of the SOCD values of profiles belongs to the same map unit (subgroup) of brown soils, which was linked to all occurrences of that map unit on the 1:2 500 000 scale map. The SOCD value was 11.07 kg C m–2.

Map Generalization Influence as Scale Decreases from 1:2 500 000 to 1:10 000 000
Estimates for cinnamon soils had the greatest change as map scale decreased from 1:2 500 000 to 1:10 000 000 (Fig. 5). The map units calcareous cinnamon soils and weakly developed cinnamon soils were eliminated as the map scale changed from 1:2 500 000 to 1:4 000 000. Next, the leached cinnamon soils were eliminated on the 1:10 000 000 scale map (Table 6). Table 6 shows that the estimation of SOC stocks for leached cinnamon soils had the most significant influence on the estimation of cinnamon soils as the map scale decreased from 1:2 500 000 to 1:10 000 000.


View this table:
[in this window]
[in a new window]
  		Table 6. Estimated soil organic C stocks for map units of cinnamon soils (soil great group) as map scale decreases from 1:2 500 000 to 1:10 000 000 using the mean, median, and PKB (pedological professional knowledge based) methods.

 
Figure 8 illustrates the map generalization process of the map unit leached cinnamon soils as the map scale decreases from 1:2 500 000 to 1:10 000 000. Delineations with 67.31 Tg C of leached cinnamon soils on the 1:2 500 000 scale map were preserved on the 1:4 000 000 scale soil map. The estimated SOC stocks for leached cinnamon soils using the 1:4 000 000 scale map were remarkably higher than those obtained with the 1:2 500 000 scale map. This can be attributed to some occurrences of other map units (i.e., brown soils, Castano-cinnamon soils, water bodies, etc.) on the 1:2 500 000 scale map being replaced by leached cinnamon soils on the 1:4 000 000 scale map. Such delineations were as large as 46318 km2, probably due to the fact that the 1:4 000 000 scale soil map is an old version, compiled before 1978. As the map scale decreased from 1:4 000 000 to 1:10 000 000, the map unit leached cinnamon soils was eliminated and the corresponding delineations were replaced by brown soils, cinnamon soils, aquic cinnamon soils, castanozems, and fluvo-aquic soils on the 1:10 000 000 scale map.


Figure 8
View larger version (18K):
[in this window]
[in a new window]
  		Fig. 8. Changes in the estimates of SOC (soil organic C) stocks for leached cinnamon soils using the median method as map scale decreases from 1:2 500 000 to 1:10 000 000. Arrows indicate the map generalization process as map scale decreased, and the two SOC stock values in each shaded box were based on some delineations of a map unit being preserved while others were merged into other map units during the map generalization process; however, all delineations of the leached cinnamon soils map unit were merged into cinnamon soils, aquic cinnamon soils, etc., as map scale decreased from 1:4 000 000 to 1:10 000 000).

 
Other Influences on Soil Organic Carbon Estimates Caused by Different Scales
Changes in soil map scales may affect the number of profiles entered into the estimation process of SOC stocks. Profiles belonging to some map units may be excluded from the estimation process as the map decreases in scale, unless steps are taken to correlate the data to a different existing soil dataset. Such profiles are not linked to delineations of the map units according to the soil type name if the soil type information represented by these map units is eliminated during the map generalization process. For example, "leached cinnamon soils" was eliminated on the 1:10 000 000 scale map, and profiles belonging to this map unit were excluded from the estimation. Soil map scales also have an influence on the choice of linking methods used to aggregate SOCD values of soil profiles to represent the SOCD values of map units. For example, the basic map units on the 1:500 000 soil map were soil families. There were several soil profiles (soil series) with different SOCD values than the map units (soil families) in which they were included. To some degree, the PKB method reflects the variability within a map unit, and although such variability is based on pedological professional knowledge, it is probably constrained by the scale of a map. For example, there are only 27 delineations on the 1:10 000 000 soil map. The PKB method is almost the same as the mean method in such a situation. To preserve the full set of soil data, modelers must make certain that the soil type names used by the profiles are identical to the map units before linking the soil profiles to a map unit, so that the soil properties will be combined successfully with the spatial information according to the soil type name. If names of soil types between two soil maps of different scales are not identical, the correlation between the soil type names must be established beforehand (i.e., the 1:4 000 000 soil map used in this study). Such correlation processes probably impart some influence on the estimated results. Another problem caused by the change in soil map scale is the difficulty in dealing with inclusions of the delineations on soil maps of different scales. Such delineations have the same spatial positions and similar shape, but they are classified into different soil types on different soil maps. For example, some delineations of lithosols (soil great group) on the 1:500 000 soil map of the Hebei Province are called loessial soils (soil great group) on the 1:1 000 000 soil map. This same problem was found by Galbraith et al. (2003) in a mismatch between soil orders on different scales of maps in the same county. Moreover, methods of SOCD determination within an individual soil profile can also affect the estimated results, as can variability of SOCD within the same taxa (Galbraith et al., 2003); however, it was not considered in this study because the estimation methods of each profile used in this study were identical.

Changes in soil map scale cause changes in the number of profiles entering into the estimation process and the choice of the linking methods used to combine soil properties with spatial extent. The difficulties in dealing with inclusions on soil maps of different scales and the correlation process of names of soil types between soil map units and soil profiles affect the estimation of SOC stocks, too. These factors, such as changes in the number of profiles and linking methods, difficulties in dealing with inclusions, and correlation processes, in turn affect the process of assigning SOCD values to soil maps of different scales. As a result, the differences in SOCD values of the delineations obtained with maps of different scales and the change in total land area of each soil type together affect the final results of the estimation of regional SOC stocks.


   CONCLUSIONS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
REFERENCES
 
When the median, mean, and PKB methods were applied to soil maps at scales ranging from 1:500 000 to 1:10 000 000, similar trends in the estimation of SOC stocks were observed as map scale decreased from 1:500 000 to 1:10 000 000. The SOC stocks estimated using the 1:4 000 000 scale soil map were higher than those from the 1:10 000 000 or 1:2 500 000 scale maps. If the SOC stocks estimated by the 1:4 000 000 scale soil map were ignored, a persistent increasing trend could be observed as the map scale increased from 1:10 000 000 to 1:500 000. As the map scale changed from 1:10 000 000 to 1:500 000, the SOC stocks obtained with the mean method were always higher than the values of the median or PKB methods. Thus, the influence of soil map scale on the estimation of SOC stocks in the Hebei Province resulted mainly from the map generalization process.

Very high or low SOCD values in several profiles linked with a map unit had a remarkable influence on SOC estimation when the mean method was applied. The median method may be more reasonable than the mean method because the influence of an extremely high value can be partially ignored by the median method; however, both the median and mean methods assume that each soil series represented by the profiles of a map unit has the same extent. The PKB method was more advantageous than either the median or mean methods because the PKB method applied different land area extents to each soil series than the median and mean methods did, and the difference was determined by linking profiles to different areas of a map unit using information recorded in Ding (1992) and pedological professional knowledge.

Estimates of SOC stock should consider not only the total SOC stock but also the SOC stock at different spatial locations; from this point of view, the 1:500 000 scale soil map has the most detailed spatial information of SOC among the five soil map scales considered in this study, and the PKB method can reflect the spatial variability of SOCD within a map unit to some degree. Therefore, the 1:500 000 soil map combined with the PKB method is probably the best one for SOC stock estimates in Hebei Province.


   ACKNOWLEDGMENTS
 
Funding provided by Knowledge Innovation Program of CAS (ISSASIP0201), The National Natural Science Foundation of China (No. 40471081) and the National Key Basic Research Support Foundation of China (G1999011810).

Received for publication May 10, 2004.


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 





This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via ISI Web of Science (2)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Zhao, Y.
Right arrow Articles by Wang, H.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Zhao, Y.
Right arrow Articles by Wang, H.
Agricola
Right arrow Articles by Zhao, Y.
Right arrow Articles by Wang, H.
Related Collections
Right arrow Carbon Sequestration
Right arrow Hillslope Analysis
Right arrow Pedology

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
The SCI Journals Agronomy Journal Crop Science
Vadose Zone Journal Journal of Plant Registrations
Journal of Natural Resources
and Life Sciences Education		 Journal of
Environmental Quality