Search

Abstract: We used an object-oriented approach in conjunction with the Random Forest algorithm to classify agricultural practices, including tillage (till or no-till (NT)), crop intensity, and grassland-based conservation reserve (CR). The object-oriented approach allowed for per-field classifications and the incorporation of contextual elements in addition to spectral features. Random Forest is a classification tree-based advanced classifier that avoids data over-fitting associated with many tree-based models and incorporates an unbiased internal classification accuracy assessment. Landsat satellite imagery was chosen for its continuous coverage, cost effectiveness, and image accessibility. Classification results for 2007 included producer''s accuracies of 91% for NT and 31% for tillage when applying Random Forest to image objects generated from a May Landsat image. Low classification accuracies likely were attributed to the misclassification of conservation-based tillage practices as NT. Results showed that the binary separation of tillage from NT management is likely not appropriate due to surface spectral and textural similarities between NT and conservation-type tillage practices. Crop and CR lands resulted in producer''s accuracies of 100% and 90%, respectively. Crop and fallow producer''s accuracies were 95% and 82% in the 2007 classification, despite post-senesced vegetation; misclassification within the fallow class was attributed to pixel-mixing problems in areas of narrow (<100 m) strip management. A between-date normalized difference vegetation index approach was successfully used to detect areas having “changed” in vegetation status between the 2007 and prior image dates; classified “changed” objects were then merged with “unchanged” objects to produce crop status maps. Field crop intensity was then determined from the multi-year analysis of generated crop status maps. [Copyright &y& Elsevier]

Abstract: There has been growing interest in the use of diffuse reflectance as a quick, inexpensive tool for soil characterization. Some studies, using techniques like Partial Least Squares (PLS) regression of 1st derivative spectra have reported predictive accuracies for soil Organic C (OC) and Inorganic C (IC) that approach the analytical limits of standard laboratory measures. We applied 1st derivative Visible and Near-Infrared (VNIR) reflectance PLS regression modeling to soil samples obtained from six sites with similar soils across three counties in north central Montana, with five completely random 30% test sets selected for model validation. We obtained–relative to estimated SEL (Standard Error of Laboratory reference measurements) of 1.07 and 0.97 g kg−1 for OC and IC, respectively–SECV (calibration Standard Error of Cross-Validation) values of 1.04–1.20 and 1.54–1.63 g kg−1, and SEP (validation Standard Error of Prediction) values of 1.09–1.27 and 1.43–1.63 g kg−1. These results, together with validation RPD (Residual Prediction Deviation) values ≥2, could suggest a stable, effective PLS calibration that could be applied to similar soils in the same physiographic region. However, when we attempted to predict soil C for each of the six sites in turn using the remaining five sites for calibration, the models failed completely at two of the six sites and gave inconsistent results at a third site despite pre-screening for spectral similarity. “One-off” local calibrations for this study required ∼20–35% of the full samples, which could be prohibitively expensive for many applications. The results of this study demonstrate that “pseudo-independent” validation (random selection of non-independent test samples) can overestimate predictive accuracy relative to independent validation. The spatial structure of calibration and validation samples matters a great deal. Greater care needs to be taken to ensure that validation samples are independent to a degree that matches intended model use. [Copyright &y& Elsevier]