Mr. Qinke Yang, Northwest University, China. Global Symposium on Soil Erosion (GSER19), 15 - 17 May 2019 at FAO HQ.

1. Interpretation of Landuse and Soil
Conservation Measures in the Pan-Third
Pole area using High Resolution Images
Qinke Yang,Mengyang Zhu, Chunmei Wang, Baoyuan Liu,
Xing Wei3, Guowei Pang1, Yuru Li1, Xiang tu1
qkyang@nwu.edu.cn

3. Content
1 Background
2 Methodology
3 Results & analysis
4 Conclusion

4. table 1 main researches on regional soil erosion mapping
1 Background
(1) Researches on regional soil erosion mapping
Author Date Method
Resol./map
scale
Extent
Data
accessibility
Zhu, et al, 等 1965, 1999 manual map 1:15 million China available (1)
Olderman 1991, 1994 manual map 1:10 million Glogal available (2)
Batjes 1996 RS, GIS, model 55km Gloga not available
Lu & Yu 2002 RS, GIS, RUSLE 1000m Australia not available
Yang, et al 2003 RS, GIS, RUSLE 1000m Gloga not available
Panagos, et al 2015 RS, GIS, RUSLE 1000m Pan-EU not available
Bosco, et al 2015 RS, GIS, RUSLE 1000m Pan-EU not available
Teng, et al 2016 RS, GIS, RUSLE 250m Australia not available
Borrelli, et al 2017 RS, GIS, RUSLE 250m Australia available (3)
USGS 2018 Sampling+(R)USLE 1:10k US not available

5. Two issues
* Get erosion rate accurately and rapidly in sampling unit
* Make a regional erosion map, with local geo-variation
First issue will be discussed in my presentation
China, 2011 Australia, 2017 EU, 2015 US, 2018
Global, 1991 Global, 2011 Global, 2017

6. 2 Methodology
(1) Research area and source data: Pan-third Pole area,
Google Images (0.5m), Global landcover 30m (GLC30).
(2) Designing the sampling unit
stratified unequal probability systematic sampling; 20880 units
shape -- watershed (hill) or square box, 0.5-3km2
0.5 long * 0.75 lat, 55km
70%
Sampling unit and image qualityresearch area

7. (3) Interpretation
Features: landuse, soil conservation measures
Method: visual plot; ArcMap, Python based plog-in tool, etc
cut polygon, input attributes Discussing the method
interpreting review and discussion

8. (4) Field Check
Southern Tibet of China: 7 watersheds have been surveyed
by UAV, landuse and measures interpreted.
Northern Thailand: 15 units has been checked in field.
Road
Cropland
Forest
Grass
Water
Villages
Bare land
Google and UAV Image Interpreted resultssampling unit
UAV survey
field tip road
boundary of Tibet, China
sampling unit
field tip road
Chiang Mai
Tibet, China
Northern Thailand

9. 9
Gullied terrain, grassland with
shrub; cropland nearby
Close display photo save to album
search route Use a mobile app (in Chn),
photos with time & position
info have been taken in the
filed, for image features
difficulty to be plotted in lab;
the photos have been used to
improve method.
red circle: place of photo

10. (5) Calculation of water soil erosion rate
Model: Chinese Soil Loss Equation (CSLE)
sampling unit + regional erosion factors, R, K, LS, NDVI, etc
R, K, LS, NDVI/NPV
CSLE (Liu, 2003)
Landuse/measures Water Erosion map
0.22 44.16
t/hm2.a
Just complete some, on going,…

11. Global erodibility (K) mapping
Using 7.5 arc SoilGrids and profile database & 3 algorithms;
global k has being calculated, for next step work.
EPIC-k
USLE-k
RUSLE2-k
A tool developed by us

12. IDN00509 India
IND02484 India
PHL01367 PhilippinesCHN02678 China
CHN09307 IndiaMMR01819 Burma EGY04056 Egypt
CHN09184 China
3 Results and analysis
Results: 1:10k, map of landuse & cons. measures for sampling units

13. CHN06303 China
poly attribute table
arc attribute table
Sampling unit database
Attributes: LU, BM, EM, etc
Format: ESRI Geodatabase
Record no: 20880
Layers: landuse/measures
DEM, slope, image, …

14. In 11 sampling sites (cover 100 sampling units)
Histogram similarity index are more than 0.75, except Tibet plateau of
China, which mainly because poor interpretation of grassland and bare land.
Differences mainly in forest, grassland, water, and bared land, need
sufficient knowledge for the area.
Analysis 1 -- Similarity of interpreted Landuse and GLC30

15. In Tibet, the mean Kappa is 0.7, mean accuracy is 80.04%.
In Thailand, the Kappa greater than 0.5 (except one unit is
0.4), and the average accuracy reaching 85.78%
Analysis 2 – accuracy of interpreting based on field check
Southern Tibet plateau, China Northern Thailand
Kappa
Kappa
Accuracy/%
Accuracy/%
1 2 3 4 5 6 7 8 9

16. Analysis 3 – applicability
of the interpreted results
Maps of water erosion rate has been generated (some units);
The rate is close to observation (eg. in Thailand), and literature;
Erosion rate was over estimated if soil conservation measure
not taken into account (A0, potential erosion);
Landuse ET (0.75) A0 (18.7) A (9.2)
Landuse ET (0.87) A0 (45.8) A
(27.7)
Landuse ET (0.66) A0 (26.5) A (16.5)
Sampling unit in Tibet
Sampling unit in Loess PlateauSampling unit in Thailand
* A0 = RKLSB A = RKLSBET

17. 4 Conclusion
(1) sampling survey of regional landuse and soil
conservation measures can be completed quickly
and accurately based on high resolution image and
visual interpretation.
(2) Soil loss rate for sampling unit can be calculated
accurately based on the interpreted results & CSLE.
(3) Need automatic interpretation, more international
cooperation in future.

18. Thanks for your attention
Any comments & suggestions, email me please
Qinke Yang, Northwest University, China
qkyang@nwu.edu.cn