• No results found

Section 2.3 provided a review of ground-based soil sensors, appropriate for on-the-go soil mapping, with most emphasis placed on those relevant to the present water use efficiency study. Other sensors exist that have not been discussed, e.g., microwave (surface roughness) (Richard et al., 2008) and magnetics (iron content) (Maier et al., 2006) sensors.

In reviewing the literature, the electromagnetic sensors stand out as the most popular and useful sensors for general soil characterization, to this point in time. For this reason the EM sensor will be used in the present study to quantify and map soil variability at the farm-scale. EM data can successfully be interpreted on a basis of soil texture and soil type (e.g., Sudduth et al., 2005), and extrapolated to estimate available water-holding capacity (e.g., Kitchen et al. 2008). Soil moisture sensors, e.g., time domain (TDR) and frequency domain (e.g., capacitance sensors) have been successfully employed for accurate on-the-spot estimation of soil moisture (e.g., Evett et al., 2006). However, on-the-go applications of capacitance sensors for soil moisture assessment have been less successful (e.g., Whelan et al., 2008; Lammer et al., 2008). Optical and radiometric on-the-go sensors have also been used for accurate soil moisture assessment (e.g., Mouazen et al., 2005). A Vis-NIR DRS sensor will be trialled in the present study for field soil analysis of soil moisture. One advantage of Vis-NIR DRS sensors is that soil spectral data can be interpreted for a number of soil properties at the same time (e.g., soil moisture, carbon and nitrogen) (Viscarra Rossel & McBratney, 1998). Vis-NIR DRS has also been employed to assess a wide range of other soil properties, including soil pH, cation exchange capacity, clay, sand and phosphorus (Adamchuk, 2008). In addition on-the-go ion-specific electrodes have been used for

soil pH and K status (Adamchuk et al., 2005), although there are some operational difficulties to be overcome (Lobsey et al., 2008).

Ground-based gamma spectrometers have recently become available to soil researchers, and collected data have been widely interpreted. For example, the gamma sensor trade-named The Mole is used to produce a wide range of soil parameter maps. Such maps are used for precision agriculture strategies, which could include irrigation scheduling.

A number of examples have been given of the simultaneous use of more than one sensor, e.g., EM and gamma sensors (Doolittle & Collins, 1998; Wong et al., 2008; Taylor et al., 2008). Dual sensors commonly improve soil property predictions (e.g., Taylor et al., 2008), because they respond to more than one variable (Adamchuk, 2008). Dual sensors have the increased potential, compared with a single sensor, to compensate sparsely sampled field measurements and estimate their spatial distribution at high resolution in complex field situations without the need for expensive and extensive direct sampling and measurements (Wong et al., 2008). Given that one of the most critical aspects of soil testing is actually obtaining representative soil samples (Adamchuk et al., 2004), the ability of on-the-go soil sensing to map soil variability, for (i) guided soil sampling and (ii) guidance for installation of on-the-spot soil monitoring, e.g., installation of embedded soil moisture sensors for guiding irrigation scheduling, is a significant advance in soil parameter mapping.

Taylor et al. (2008) employed GPS, EM and gamma radiometry to acquire seven data layers at a Scottish farm (elevation, ECvertical., EChorizontal., Gammatotal counts, GammaK counts, GammaUcounts, GammaTh counts). These datasets were collated into one file, and k-means cluster analysis was performed to generate 6 classes, which were then targeted for soil sampling. The six classes can also be used as management zones for variable rate application of, for example, seed, fertilizer or irrigation water, and also targeted for installation of sensors, for example, for on-the-spot monitoring of soil moisture.

Future challenges for development of new soil sensors include deciding on the correct sensor combinations that may be region-specific, as well as more automated sensor data processing techniques that require less subjective expert input (Adamchuk, 2008).

On-the-go soil sensing and mapping is used for real-time and map-based variable rate application. Richard et al. (2008) employed microwave and laser sensors for surface roughness, a load cell for soil resistance, and a capacitance probe for soil moisture, all attached to the rear of a tractor. They discussed the feasibility of real-time manipulation of tillage depth and speed, as well as seed rate, based on real-time sensor data. Map-based variable rate application has been adopted widely over the last decade, with a prescription map concept for variable rate applications (e.g., Godwin, 2003).

Thompson et al. (2007) and Blonquist et al. (2006) used on-the-spot soil moisture sensors to guide irrigation scheduling. Thompson et al. (2007) monitored soil moisture every 30 minutes in a greenhouse replicated plot trial, using capacitance sensors. Actual commencement of crop water stress was indicated by the first statistically significant difference in midday leaf water potential (LWP), measured by sampling a leaf at midday and assessing soil water potential (Soil Moisture Co Model 3005, Santa Barbara, CA, USA). This was matched to soil moisture conditions, “apparent daily crop water uptake” (ADCWU) and “daily soil water loss” (DSWL). DSWL is calculated over 24 hours and ADCWU over day-light hours. As the leaf goes into stress ADCWU becomes smaller.

Blonquist et al. (2006) monitored soil moisture with one Acclima Digital TDT transmission line sensor in a trial plot, and the sensor was used to directly control the irrigator, switching on the irrigator when soil moisture dropped below a threshold value. While they examined the potential water savings of such a system, there was no discussion of how to cope with spatial variability when such a system is scaled up to field and farm scale. The map-based concept, made possible by on-the-go soil sensor systems, and discussed by Adamchuk (2008), addresses the issue of spatial variability that will occur at some sites.

Currently soil property maps, developed from soil sensing systems, are largely site- specific. Obviously a universal sensor calibration for one soil property is ideal – but rarely achieved. Some advances have been made in this direction for laboratory-based Vis-NIR soil sensing by Brown et al. (2006) based on a dataset of 3768 samples

collected from four continents; and by Viscarra Rossel et al. (2008) based on >7000 spectra contributed from 23 countries.

The potential use and application of these new high resolution proximal soil sensing methods is exemplified by an international collaborative effort DIGISOIL in the European Union (DIGISOIL, http://eusoils.jrc.ec.europa.eu/projects/DIGISOIL/; 2009). DIGISOIL is a consortium of European scientists who are developing new in situ and proximal sensing technologies to assess soil properties and degradation indicators that can be used for the production of high quality geo-referenced soil maps (Grandjean et al., 2008). DIGISOIL intends to improve in situ and proximal measurement technologies with integration into digital soil mapping (DSM), because of the obvious need for high resolution and accurate maps of soil properties. The stated core objective is to explore and exploit these technologies to answer societal demand to address soil degradation issues and to benefit from ecological and economic functions of soils. DIGISOIL will specifically develop and test the most relevant geophysical technologies for mapping soil properties using GPR, EMI, seismic, magnetic and airborne hyperspectral methods. This project and its stated aims exemplify an obvious need for high resolution and accurate maps of soil properties to address soil degradation issues noted around the world (e.g., Lal., 2009; Mu et al., 2009; Basset-Mens et al., 2009; Williams, 2004; Grandjean et al., 2008; Adamchuk, 2008), as well as the need for sustainable use of natural resources (Williams, 2004; Jury & Vaux, 2007; Swaminathan, 2007).

This literature review has listed available proximal sensing technologies and has outlined how they can be used at the farm-scale for site-specific management, allowing more precise application of inputs with more efficient use of natural resources, e.g., freshwater for irrigation. It has identified the relevant sensors to be used in the present research project. Opportunities also exist to establish correlations between sensor measurements and environmental issues such as erosion, compaction, organic matter decline, and salinisation.

These new methods for in situ soil analysis and mapping will provide new information for interpretation and management of soil and landscape performance. For example, a soil reflectance spectra obtained by a radiometric sensor in the field is a unique

signature for that soil, and advances in its analysis will be accompanied by new opportunities to assess soil properties and suites of properties in a completely new way, making it possible to monitor and compare performance and sustainability of soils and landscapes under existing and projected practice.

CHAPTER THREE

Proximal Sensing of Soil and Pasture at

Forest-to-Farm Land Use Change Sites

The rationale for the research reported in this chapter was to investigate the potential of two proximal sensing methods: on-the-go soil EM mapping, and backpack Vis-NIR field spectroscopy for (i) quantitative mapping of soil variability and (ii) rapid field analysis of selected soil and plant properties, respectively. Seven sites were selected in the Taupo-Rotorua region of North Island, New Zealand, where recent conversion of plantation forest to pastoral farming provided an opportunity for assessing soil changes and pasture development during the initial years after tree removal. The conversion of forest to pasture involves major disturbance of the soil profile during tree clearing and stump removal frequently resulting in highly variable soils for pasture establishment. This was therefore a suitable case study to investigate proximal sensing methods for assessing soil variability at the farm-scale. The central concept of geostatistics the variogram and kriging for analysis of the spatial structure of soil properties is discussed in this chapter.

Some of the research reported in this chapter has been presented at two conferences and accepted as one journal paper:

Hedley CB, BH Kusumo, ID Sanchez, M Hawke, MJ Hedley and MP Tuohy. 2006. Forest to Farm Conversions – a Field Characterization of Soil Heterogeneity relating to Pasture Production. p 42. In New Zealand Soil Science Society 2006 Conference Proceedings “Soils and Society”. 72pp.

Hedley CB, BH Kusumo, ID Sanchez, MJ Hedley M P Tuohy and M Hawke. 2007. Proximal sensing of soil and pasture development under “forest to farm” land use change. p 158-169. In Designing Sustainable Farms: Critical aspects of soil and water management. (Eds. L.D. Currie and L.J Yates). Occasional Report No.20. Fertilizer and Lime Research Centre, Massey University, Palmerston North, New Zealand. 527pp. Hedley CB, Kusumo, BH, Hedley MJ, Tuohy M and G Arnold. 2009. Soil C and N sequestration and fertility development under land recently converted from plantation forest to pastoral farming. New Zealand Journal of Agricultural Research 52: 443 453.

Abstract

Current forest-to-farm land-use changes are causing concern for (i) reduced forest sinks for carbon (C) sequestration, (ii) water quality, and (iii) increased pressure on water resources for irrigation. The challenge is to design new pastoral farms within the limitations of their environs. However, for this process to be achieved, the impacts of land-use change on soils and their environment must first be measured.

Sites were selected under recently converted pastures (1-yr, 3-yr, 5-yr conversions) and permanent pasture at three farms in the Taupo-Rotorua Volcanic Zone. Two proximal sensing techniques: (i) on-the-go soil electromagnetic (EM) mapping, and (ii) backpack Vis-NIR field diffuse reflectance spectroscopy, were trialed for (i) simultaneously mapping soil apparent electrical conductivity (ECa) and elevation as a measure of soil variability and (ii) rapid in situ field analysis of soil carbon (C), soil nitrogen (N), soil moisture content ( ) and herbage N.

Soil fertility results (0–7.5 cm soil depth) show rapid increase in Olsen P, with soils reaching their optimum agronomic range within 3–5 years after conversion, at two of the three farms. Results suggest a soil C sequestration rate (0 15 cm soil depth) of 6.0 T ha-1 yr-1, and a soil N sequestration rate of 0.48 T ha-1 yr-1 for the first five years after conversion at two of the farms. Decreasing C:N ratios with time since conversion reflect improved fertility status, and imply that in initial years of pasture establishment, N leaching to freshwater is reduced due to its immobilization into soil organic matter.

ECa sensor data were used to investigate soil variability using variogram analysis and kriging. The resulting soil ECa maps are available for defining soil management zones to allow site-specific management, e.g., variable rate fertilizer application and precision irrigation. Analysis of zones defined on a basis of soil variability relating to slope and aspect differences at two of these hill country sites, showed that north-facing slopes had highest C:N ratios, lowest Olsen P and lowest herbage N, suggesting less soil development, lower fertility and a reason to reduce fertilizer input to this zone. Diffuse reflectance spectra were used successfully to predict soil C, N and (R2 predicted = 0.75; 0.86 and 0.70 respectively), and herbage N (R2 predicted = 0.67).

Our research suggests forest-to-farm land-use change, with inputs of N, P, K and S to soils, allows significant soil C and N sequestration for at least 5 years after conversion. Further research is therefore required using control sites monitored over a number of years to confirm this preliminary study. The design of these new pastoral farms can be facilitated by proximal sensing tools, which allow real-time mapping of soil properties across the landscape.