 |
| Titel |
Mapping tillage operations over peri-urban croplands using a synchronous SPOT4/ASAR ENVISAT pair and soil roughness measurements |
| VerfasserIn |
Emmanuelle Vaudour, Nicolas Baghdadi, Jean-Marc Gilliot |
| Konferenz |
EGU General Assembly 2014
|
| Medientyp |
Artikel
|
| Sprache |
Englisch
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 16 (2014) |
| Datensatznummer |
250089686
|
| Publikation (Nr.) |
 EGU/EGU2014-3896.pdf |
|
|
|
|
|
| Zusammenfassung |
| Tillage operations (TOs) affect nutrient uptake, carbon sequestration, water and CO2
exchanges in soil, and therefore impact soil ecology together with biophysical processes such
as soil erosion, leaching, run-off and infiltration. They are critical for parameterizing complex
dynamic models of carbon and nitrogen. This study done in the framework of the
Prostock-Gessol3 project presents an approach for mapping TOs of bare agricultural fields
over a peri-urban area characterized by conventional tillage system in the western suburbs of
Paris (France), combining synchronous SPOT4 and ENVISAT/ASAR images (HH and HV
polarizations).
Spatial modeling relied on 57 reference within-field areas named “reference zones” (RZs)
homogeneous for their soil properties, constructed in the vicinity of 57 roughness
measurement locations and spread across 20 agricultural fields for which TOs were known.
Soil roughness expressed as the standard deviation of surface height (Hrms) was estimated on
the ground with a fully automatic photogrammetric method based on the processing of a
set of overlapping pictures taken from different viewpoints from a simple digital
camera all around a rectangular frame. The relationship was studied between the
mean backscattering coefficient of the ASAR image and Hrms choosing a limited
set of 28 RZs, on which successive random selections of training/validating RZs
were then performed; the remaining 29 RZs were kept for validating the final map
results.
Six supervised per-pixel classifiers were used in order to map 2 TOs classes
(seedbed&harrowed and late winter plough) in addition to 4 landuse classes (forest,
urban,crops and grass, water bodies): support vector machine with polynomial kernel
(pSVM), SVM with radial basis kernel (rSVM), artificial neural network (ANN), Maximum
Likelihood (ML), regression tree (RT), and random forests (RF). All 6 classifiers were
implemented in a bootstrapping approach in order to assess the uncertainty of map
results.
The best results were obtained with pSVM for the SPOT4/ASAR pair with producer’s and
user’s mean validation accuracies of 91.7%/89.8% and 73.2%/73.3% for seedbed&harrowed
and late winter plough conditions, respectively. Whatever classifier, the SPOT4/ASAR pair
appeared to perform better than each of the single images, particularly for late winter plough:
producer’s and user’s mean validation accuracies were 61.6%/53.0% for the single
SPOT4 image; 0%/6% for the single ASAR image. About 73% of the validation
agricultural fields (79% of the RZs) were correctly predicted in terms of TOs in
the best pSVM-derived final map. Final map results could be improved through
masking non-agricultural areas with land use identification system layer prior to
classifying images. These results show promises for further monitoring of TOs
within a crop production campaign, in the perspective of the ESA Sentinel-1/2 time
series. Such knowledge of cultural operations is likely to facilitate the mapping of
agricultural systems which otherwise proceed from time-consuming surveys to farmers. |
|
|
|
|
|
|