Soil organic matter sensing with an on-the-go optical sensor

Comparison between artificial neural network and partial least squares for on-line visible and near infrared spectroscopy measurement of soil organic carbon, pH and clay content

“…Many studies have reported on-line measurements of OC or soil organic matter (OM) with vis-NIRS at different geographic scales (e.g. Bricklemyer and Brown, 2010;Kuang and Mouazen 2013b;Kweon and Maxton, 2013;Mouazen et al, 2007;Muñoz and Kravchenko, 2011). Less success was reported for online measurement of soil pH (Kodaira and Shibusawa, 2013;MarinGonzález et al, 2013;Mouazen et al, 2007;Tekin et al, 2013), whereas very limited studies on CC were reported (Bricklemyer and Brown, 2010;Mouazen et al, 2013).…”

“…Among available techniques, the visible and near infrared (vis-NIR) spectroscopy proved to be the most capable technology for on-line characterisation and quantification of within field variation in soil properties (Bricklemyer and Brown, 2010,b;Kuang and Mouazen, 2013a,b;Kweon and Maxton, 2013;Mouazen et al, 2007;Shi et al, 2006;Shibusawa et al, 2001). Similar to OC, along with soil total carbon, soil total nitrogen, soil moisture content, CC has been clarified into the category of soil properties with direct spectral response in the near infrared range (Kuang et al, 2012;Stenberg et al, 2010).…”

Spiking regional vis-NIR calibration models with local samples to predict soil organic carbon in two High Arctic polar deserts using a vis-NIR probe

“…The spectrophotometer spectra were trimmed to eliminate scan insensitive regions (B440 nm and !2200 nm) and the region influenced by the grating between the two spectrometers (950Á1150 nm) (Kweon and Maxton 2013). Samples from each site were divided into a calibration dataset (n 075%) for model development and a validation set (n 025%) to internally validate the models for a more representative estimation of the prediction performance than obtained by crossvalidation (Varmuza and Filzmoser 2009;Soriano-Disla et al 2014); samples were randomly divided with the condition that organic carbon values of the test set fell within the range of the calibration dataset values.…”

“…Calibration models based on laboratory collected spectra on ground and dried samples yield good prediction results for SOC and other soil properties (refer to reviews Bellon-Maurel and McBratney 2011;Soriano-Disla et al 2014). For field condition soils, a variety of approaches for estimating SOC have been tested including using moist samples (Chang et al 2005;Stenberg 2010;Nocita et al 2013), simulated in situ methods (Morgan et al 2009), static in situ methods (Ben-Dor et al 2008;Kusumo et al 2008;Kweon et al 2009;Wenjun et al 2014) and on-the-go in situ methods (Mouazen et al 2007;Christy 2008;Kweon and Maxton 2013). Particularly useful in a remote field environment is the novel static in situ method developed by BenDor et al (2008) where a spectrometer with a probe attachment is used to collect soil spectra down the profile of a drilled hole.…”

Applications of artificial neural networks in three agro-environmental systems: microalgae production, nutritional characterization of soils and meteorological variables management

“…Color is an easily measurable physical attribute of soil through colorimeters, color sensors, computer scanners and smartphone cameras Kirillova et al, 2017;Stiglitz et al, 2016;. The difference in soil color has often been used in the soil classification (Maejima et al, 2000), different studies indicate that the OM influences the soil color, darker soils have more OM (Kweon and Maxton, 2013). Iron oxides show both redness and yellowness (Vodyanitskii and Kirillova, 2016), other properties such as clay and humus contents also is related with the redness of the soil (Hu et al, 2014).…”