Quantifying and relating land-surface and subsurface variability in permafrost environments using LiDAR and surface geophysical datasets

Abstract

The value of remote sensing and surface geophysical data for characterizing the spatial variability and relationships between land-surface and subsurface properties was explored in an Alaska (USA) coastal plain ecosystem. At this site, a nested suite of measurements was collected within a region where the land surface was dominated by polygons, including: LiDAR data; ground-penetrating radar, electromagnetic, and electrical-resistance tomography data; active-layer depth, soil temperature, soil-moisture content, soil texture, soil carbon and nitrogen content; and pore-fluid cations. LiDAR data were used to extract geomorphic metrics, which potentially indicate drainage potential. Geophysical data were used to characterize active-layer depth, soil-moisture content, and permafrost variability. Cluster analysis of the LiDAR and geophysical attributes revealed the presence of three spatial zones, which had unique distributions of geomorphic, hydrological, thermal, and geochemical properties. The correspondence between the LiDAR-based geomorphic zonation and the geophysics-based active-layer and permafrost zonation highlights the significant linkage between these ecosystem compartments. This study suggests the potential of combining LiDAR and surface geophysical measurements for providing high-resolution information about land-surface and subsurface properties as well as their spatial variations and linkages, all of which are important for quantifying terrestrial-ecosystem evolution and feedbacks to climate.

Resumé

La portée de la télédétection et des données géophysique de surface pour caractériser la variabilité spatiale et les relations entre la surface du terrain et les propriétés de la subsurface a été étudiée sous tous ses aspects dans l’écosystème de la plaine côtière d’Alaska (USA). Dans cette région, sur un site où la surface du sol est dominée par des polygones, une série de données se recoupant a été collectée, incluant : données LiDAR; géoradar, tomographie électromagnétique et résistivité; profondeur de la couche aquifère, température, teneur en humidité, texture, teneur en carbone et en azote du sol; et cations du fluide des pores. Les données Lidar ont été utilisées pour établir les cotes géomorphiques, qui peuvent indiquer un drainage potentiel. Des données géophysiques ont été utilisées pour déterminer la profondeur de la couche aquifère, la teneur en humidité du sol, et la variabilité du pergélisol. L’analyse par agglomérat des données LiDAR et des attributs géophysiques ont révélé la présence de trois zones spatiales ayant une distribution similaire des propriétés géomorphiques, hydrogéologiques, thermales et géochimiques. La correspondance entre la zonation géomorphique basée sur LiDAR, la couche aquifère selon la géophysique et la zonation permafrost, met en lumière la relation significative entre ces compartiments de l’écosystème. Cette étude montre le potentiel d’une combinaison des mesures LiDAR et des mesures géophysiques de surface pour fournir une information haute résolution sur les propriétés de surface et de subsurface du sol aussi bien que sur leur variations spatiales et liens, toutes étant importantes pour quantifier l’évolution de l’écosystème terrestre et les réponses au climat.

Resumen

Se exploró el valor de los sensores remotos y de los datos geofísicos de superficie para caracterizar la variabilidad espacial y las relaciones entre la superficie y las propiedades subsuperficiales en un ecosistema de planicie costera en Alaska (EEUU). En este sitio, un conjunto anidado de medidas fue colectado dentro de una región donde la superficie estaba dominada por polígonos, incluyendo: datos LiDAR; datos de radar, electromagnéticos, y tomografías de resistividad eléctrica; profundidad de la capa activa, temperatura del suelo, contenido de humedad del suelo, textura del suelo, contenido de carbono y nitrógeno en suelo; y cationes del fluido de poros. Los datos LiDAR fueron usados para extraer los indicadores geomórficos, que posiblemente indican un drenaje potencial. Los datos geofísicos fueron para caracterizar la profundidad de la capa activa, el contenido de humedad del suelo y la variabilidad del permafrost. En análisis de cluster de los LiDAR y los atributos geofísicos revelaron la presencia espacial de tres zonas, que tenían una única distribución de propiedades geomórficas, hidrológicas, térmicas y geoquímicas. La correspondencia entre la zonación geomórfica basada en LiDAR y la capa activa basada en geofísica y la zonación del permafrost destaca la vinculación significativa entre estos compartimentos del ecosistema. Este estudio sugiere el potencial de la combinación LiDAR y las mediciones geofísicas de superficie para proveer información de alta resolución acerca de las propiedades de la superficie y de la subsuperficie así como su variación espacial y su articulación, todos los cuales son importantes para cuantificar la evolución del ecosistema terrestre y las reacciones con el clima.

摘要

用来描述地表和地下性质的空间变异性和两者之间的关系的遥感和地面地球物理数据已在阿拉斯加(美国)的一个沿海平原生态系统进行了探讨。在本次研究场地的一个表面呈多边形的区域收集到了一套测量数据,包括激光雷达数据;探地雷达数据,电磁和电阻断层扫描数据;活性层深度,土壤温度,土壤水气含量,土壤质地,土壤碳和氮的含量;以及孔隙流体阳离子数据。激光雷达数据用来提取地貌指标,这可能指示出潜在的排泄。地球物理数据用来刻画活性层的深度,土壤水气含量和永久冻土的变化特征。通过对激光雷达和地球物理数据属性的聚类分析发现了三个在地形,水文,热和地球化学性质分布上存在异常的空间区域。基于激光雷达测量的地貌分区与基于地球物理数据的活性层和永久冻土分区之间的对应关系突出了这些生态系统分区间的紧密联系。本次研究表明可以通过结合激光雷达和地表地球物理测量来为地表和地下的性质以及它们在空间上的变化和关系提供高分辨率的信息,所有这些对于量化陆地生态系统的演化和对气候变化的反应都是非常重要的。

Resumo

Num ecossistema da planície costeira do Alaska (EUA) foi explorado o valor da deteção remota e de dados de geofísica de superfície para caracterizar a variabilidade espacial e as relações entre propriedades da superfície do terreno e da subsuperfície. Neste local, inserido numa região onde o terreno superficial é dominado por polígonos, foi recolhido um conjunto agregado de medições, incluindo: dados de LiDAR; dados de geoadar, eletromagnéticos e de tomografia de resistência elétrica; profundidade da camada ativa, temperatura do solo, teor de água no solo, textura do solo, teor de carbono e azoto no solo; e catiões no fluido poroso. Os dados LiDAR foram usados para extrair dimensões geomórficas que potencialmente indicam o potencial de drenagem. Os dados geofísicos foram usados para caracterizar a profundidade da camada ativa, o teor de humidade no solo e a variabilidade no permafrost. A análise grupal de atributos LiDAR e geofísicos revelou a presença de três zonas espaciais que tinham distribuições únicas de propriedades geomórficas, hidrológicas, térmicas e geoquímicas. A correspondência entre o zonamento geomórfico baseado no LiDAR e a zonação da camada activa baseada na geofísica e do permafrost, demonstra a significativa conexão entre estes compartimentos do ecossistema. Este estudo sugere o potencial da combinação de medições de LiDAR e de geofísica de superfície para fornecer informação de alta resolução acerca das propriedades da superfície do terreno e da subsuperfície, assim como sobre as variações espaciais e conexões, sendo todas elas importantes para a quantificação da evolução do ecossistema terrestre e as retroações com o clima.

Appendix: LiDAR and surface geophysical background

Brief descriptions of the LiDAR, ground penetrating radar (GPR), electrical resistance tomography (ERT) and electromagnetic methods are provided in this section.

LiDAR data

LiDAR data were collected over the Barrow site, Alaska, by AeroMetric on October 4, 2005 at an altitude of approximately 600 m above mean ground elevation using an Optech 70 kHz Airborne Laser Terrain Mapper (ALTM 30/70) on board a twin engine Cesna 310 aircraft. The system was configured with a differential global positioning system (DGPS) and 200-Hz inertial measurements units (IMU) were used to improve the accuracy of ground data. LiDAR data were processed by PND Engineers Inc. The data were post-processed utilizing Optech’s REALM software, which first computes a 2-Hz Post Processed Kinematic DGPS trajectory, then integrates the IMU data for a final smoothed best estimate of trajectory (SBET). SBET data were then integrated with the LiDAR pulse data to obtain a final x, y, z “point cloud” dataset. Classification of point cloud data for bare-earth was performed using Terrascan software. The horizontal and vertical accuracy is approximately 0.30 and 0.15 m respectively. A digital elevation model at 0.5-m spatial resolution was created using GRASS software by importing x, y, z point cloud data and performing linear interpolation.

Ground penetrating radar (GPR)

GPR methods use electromagnetic energy at frequencies of ∼10 MHz to 1 GHz to probe the subsurface. At these frequencies, the separation (polarization) of opposite electric charges within a material that has been subjected to an external electric field dominates the electrical response. GPR systems consist of an impulse generator which repeatedly sends a particular voltage and frequency source to a transmitting antenna. The most common ground surface GPR acquisition mode is surface common-offset reflection, in which one (stacked) trace is collected from a transmitter-receiver antenna pair that is pulled along the ground surface. When the electromagnetic waves in the ground encounter a contrast in relative dielectric permittivity (also known as dielectric constant), part of the energy is reflected and part is transmitted deeper into the ground. The reflected energy is displayed as 2D profiles that indicate the travel time and amplitude of the reflected arrivals; such profiles can be displayed in real time during data collection and can be stored digitally for subsequent data processing.

The velocity of the GPR signal can be obtained by measuring the travel time of the signal for various known separation distances between the transmitter and the receiver. For surface GPR, this is accomplished by successively moving the transmitter and receiver apart at specific increments to yield what is called a common midpoint (CMP) gather. Analysis of the arrival time of the reflections in this CMP gather can be performed to estimate the radar propagation velocity. This velocity can be used to convert the GPR profiles, which are recorded as distance versus travel time, into distance versus depth sections. A review of GPR methods applied to hydrogeological applications is given by Annan (2005).

The propagation phase velocity (V) and signal attenuation of the electromagnetic wave are controlled by the dielectric permittivity (or dielectric constant, κ) and the electrical conductivity of the subsurface material through which the wave travels. At the high frequency range used in GPR, the velocity in a low electrical conductivity material can be related to the dielectric permittivity, as

$$ \kappa \approx {{\left( {\frac{c}{V}} \right)}^2} $$

(Eq. 1)

where c is the propagation velocity of electromagnetic waves in free space (Davis and Annan 1989). Due to the sensitivity of dielectric permittivity to moisture content (Birchak et al. 1974; Topp et al. 1980), the travel time and thus velocity of the radar wave are largely controlled by water content. A petrophysical relationship developed using Barrow soils that are similar to those of the study site to relate volumetric water content (θ) to dielectric permittivity is (Engstrom et al. 2005):

$$ \theta =-2.5+2.508\kappa -0.03634{\kappa^2}+0.0002394{\kappa^3} $$

(Eq. 2)

However, due to the typical presence of significant organic materials and variable freeze states in Arctic soils, other factors must also be considering when using dielectric measurements to quantify soil-water content (e.g., Watanabe and Wake 2009).

Electrical resistance tomography (ERT)

Electrical resistivity methods are probably more frequently used for shallow subsurface studies than any other geophysical method. Resistivity is an intrinsic property of a material indicating its ability to resist electrical current flow; it is the inverse of electrical conductivity. At low frequencies measured, energy loss via ionic and electronic conduction dominates. Ionic conduction results from the electrolyte filling the interconnected pore space (Archie 1942) as well as from surface conduction via the formation of an electrical double layer at the grain-fluid interface (e.g., Revil and Glover 1998). Most resistivity surveys utilize a four-electrode measurement approach, where current is injected between two electrode and electrical potential difference measured between two others, while varying the location of electrodes along the profile and the distance between them (e.g., Binley and Kemna 2005). Modern multi-channel geoelectrical equipment decrease acquisition time by injecting current through two electrodes and measuring the potential difference (voltage) signal between several pairs of electrodes and using electrodes alternatively as both current and potential electrodes, a method now referred to as electrical resistivity tomography (ERT). A review of this method is provided by Binley and Kemna (2005).

Data quality is typically initially assessed through creating an apparent resistivity (pseudo-section) section, which is developed following Ohm’s Law with information about the injected current, the measured potential difference and the geometric factor (which is a function of the electrode configuration) and through assuming uniform subsurface conditions. Further processing involves the estimation of the spatial distribution of resistivity that reproduces in a given range of uncertainty the measured data. Inversion of ERT data typically involves iterative minimization of the misfit between measured and calculated data by optimizing two- or three-dimensional electrical resistivity models (e.g., Kemna 2000; Ramirez et al. 2005; Guenther et al. 2006).

For the inversion of the ERT data described in the previous section Electrical resistance tomography (ERT), the discretization included 0.05-m thick cells for the shallowest 0.2 m, and further 0.1-m thick cells until 0.8-m depth, 0.25-m thick cells until 5-m depth, and 0.5-m thick cells below. The horizontal discretization is 0.25 m (half the electrode spacing). The modeling grid was defined to be much larger than the region of interest to ensure reliable inversions. Minimal smoothing was applied. Inversion of only the measurements collected when the distance between the closest injection and potential electrode was equal to or smaller than four times the distance between the injection electrodes gave a lower error of 8 % but revealed very identical shallow variations; this indicates that the highest source of error is associated with the imaging of deepest structures. No corrections for temperature dependency were made, although it is recognized that correcting resistivity to a reference temperature of 20 °C would lead to lower resistivity values than the values considered here (e.g., Hayley et al. 2007).

Electromagnetic (EM) data

Controlled source inductive EM methods consist of injecting a time- or frequency-varying current in a transmitter coil to create a primary EM field that travels to a receiver coil via paths above and below surface. Governed by Maxwell’s equations, the created EM field induces eddy currents in any conductors, which creates a secondary magnetic field. Attributes of this secondary magnetic field, such as amplitude, orientation, and/or phase shift, can be measured by a receiver coil. By isolating these attributes from those of the primary field signal, information about the subsurface electrical conductivity distribution can be inferred (e.g., McNeill 1990; Telford et al. 1990). A review of EM methods for shallow subsurface investigations is given by Everett and Meju (2005).

A frequency domain EM method that is used in this study for shallow subsurface investigations is the EM38 (e.g., McNeill 1980; Geonics 2009), which is a ground conductivity meter that operates at a frequency of 14,500 Hz using transmitter receiver coils oriented vertically or horizontally and with an offset distance of 1 and 0.5 m. Because this method does not require contact with the ground, data can be acquired very quickly. EM38 data are often displayed as maps of apparent electrical conductivity to highlight lateral variation over large area over an averaged depth interval. More, using different frequency and orientation EM datasets collected over the same region, the data can also be inverted to obtain a model of electrical conductivity (and magnetic susceptibility) distribution that best reproduce the data. Several inversion approaches have been successfully developed for EM measurements (e.g., Farquharson 2000; Triantafilis and Santos 2010; Minsley et al. 2012a).

When the EM38-MK2 system is used (as described in the previous section LiDAR and geophysical datasets and analysis) in the vertical mode with 1-m coil spacing, the signal is expected to be most sensitive to variations at 0.4 m below ground surface; variations below 0.2 but above 2 m from ground surface contribute to 75 % of the response. When using a horizontal mode with a 1-m separation, the contribution from near-surface material is large and decreases almost monotonically with depth, leading to 75 % response contribution from the zone above 1 m depth (McNeill 1980). When the coils are separated by 0.5 m, the contribution depths described in the preceding are approximately half.

The acquired electromagnetic data were inverted at each location to obtain a 1D-subsurface-resistivity model using the least-squares based algorithm EM1DFM (Farquharson 2000). The parameters used in the EM inversion code enforces a fixed trade-off parameter that controls how the objective function is minimized, and that enforces a layered model to be as close as possible to a starting layered model while fitting the data in a given range of uncertainty. The model was defined with 11 resistivity layers of various thickness (from top to bottom, in m: 0.25, 0.1, 0.1, 0.3, 0.3, 0.3, 0.3, 0.3, 0.3, 0.3, 0.2, infinity) and the starting model was defined with 100 Ohm.m for the two top layers, 3,000 Ohm.m for the next seven layers and 50 Ohm.m for the two deepest layers. These model parameters were defined on prior information gained from the probe-measured base of the active layer and ERT profile. Although the obtained mean absolute difference between the calculated and measured EM data is very small (<2 %), the calculated model is only one of several possible solutions.