Networked Soil Sensors

We have been field testing coupled soil sensors with conventional minirhizotron for the past beginning in October of 2005. The preliminary data indicate that this approach provides reasonable data, but just as importantly, show new ecological responses that cannot be generated using conventional approaches. We located 5 nodes along a transect underneath the NIMS transect. Each node had an individual meteorological station for precipitation, air temperature, relative humidity and barometric pressure to calculate CO₂ and vapor pressure deficit. In each node, we placed replicated minirhizotron tubes (3 per node) and sensor probes (2 sets per node) at 2, 8 and 16cm for CO₂, temperature, and water content. From these units, we can observe roots, rhizomorphs, and hyphae (we attempt to observe these weekly, and in daily campaigns), measure soil temperature and moisture, and calculate soil respiration, using vertical gradient differentiation. For this report, we will show all data using the Marshall model (Tang et al. 2005) for calculating respiration. This model calculates fluxes as a function of differential CO₂ concentrations, temperature, moisture, and integrates a soil porosity factor determined from the soil texture measurements.

Trials to date indicate that our determination of soil respiration are reasonable, both in terms of absolute amounts, and in terms of temporal patterns. Soil respiration estimates was tested against the Licor 8500 and found to be reasonable and against estimates of total respiration, using preliminary measurements from Mike Goulden’s tower measurements.

Just as important, we can observe responses to events. For example, following each rainfall event, we see responses in both the net flux (from Goulden, <0 means a drawdown of CO₂ from the atmosphere, or net photosynthesis) and in respiration (>0 means CO₂ respired back to the atmosphere.

We were also able to observe a hysteresis response in respiration to soil cooling (fall-winter) and soil warming (spring). Similar patterns could be observed in respiration in response to soil moisture, and in organisms (roots, hyphae, rhizomorphs) in response to both temperature and moisture.

In the figure (above), green denotes the fall cooling, where flux (J): J= 2.926 - 0.426T + 0.024T², r²=0.57, p<0.0001,  and blue denotes spring warming, where: J=0.437 + 0.339T – 0.008T², r²=0.928, p<0.0001

One critical issue that our data expose is the incredible variability that exists from point to point. The arrayed network thus becomes extremely powerful in understanding ecosystem dynamics. For example, the figure below illustrates the difference among sensor sets in soil respiration. Two explanations can explain differences between locations. First, sensor failure. But the calibrations show no malfunction. The second is that there are differences between locations. Indeed, we can ascribe the differences to such factors as exposure, water fluxes, hydraulic lift, and other biological and physical characteristics. These are being further analyzed at this time. When we examine each node and each sample time:

J= 0.175 * e^(0.101T) * e^{[(19.151?) + -51.449(?2)]}, r²=0.881, p<0.0001

Automated Minirhizotron (AMR) Status:

The AMR is currently being bench tested in the lab at JR. The images are of exceptionally high quality, with ready visualization of roots and hyphae. Some technical issues have been uncovered involving machine (PC) dependent driver conflicts that are currently being resolved.  Final mechanical drawings are being reviewed and software is being finalized.  We had a slight setback because the tubes received were not of the proper optical quality (due to lack of notation on the initial drawing).  The correct tubes should arrive within the next three weeks.  End caps and O-rings are now in stock, so once the proper tubes arrive, we will begin placing them in the ground to begin the seasoning process.

Minirhizotron Image Analysis

There are thousands of unanalyzed minirhizotron images across the research community and automated image acquisition will only increase that problem. Thus, there is a need to develop automated software to analyze images of roots and fungal hyphae. Our initial foray into automation was to perform scale-space analysis to identify linear structures in the images (Lindeberg 1998) and then learn a classifier (a support vector machine, Cortes and Vapnik 1995) to discriminate between roots, fungal hyphae and soil. We also experimented with simple models of temporal changes by registering adjacent images and thresholding their difference in order to eliminate regions with insignificant changes. More accurate and flexible deformation models must be developed, along with temporal models that can account for uneven temporal sampling and large-scale changes.

Scale-space analysis (Lindeberg 1994) holds promise in this area since structures of interest are defined and detected in a continuum of resolutions by finding maxima of differential operators in space and scale. Multi-resolution analysis with wavelets (Mallat 1989), curvelets and ridgelets (Starck et al. 2002) are also natural candidates for bottom-up processing and feature extraction in minirhizotron images. However, it is clear that such low-level processing will not be sufficient, as a significant amount of prior knowledge is necessary in order to correctly classify structures of interest. Therefore, we foresee the necessity to integrate such low-level, bottom-up processing with explicit statistical (generative) models that can be learned from labeled data in order to enforce prior knowledge from the top down.

At this point, monitoring roots seems to hold considerable promise. For the fungal hyphae, we are limited by the quality of the resolutions from conventional minirhizotrons, and anticipate a dramatic improvement when the AMR are deployed.

Future Directions

Work planned in the coming year: Tubes for the AMR are planned for placement this month at the James Reserve. We anticipate that AMR units will be ready for instillation by summer. Nitrate and ammonium sensors have been ordered and will be expected in the soil by summer or late spring. Continued measurements of respiration as a function of temperature, moisture, and soil organisms measured using the conventional minirhizotron are expected.

People

Faculty:

• Michael F. Allen, Plant Pathology/Biology, UC Riverside

Researchers:

• Chris Glover, CCB, UC Riverside
• Einay Mayzlish, CCB, UC Riverside
• Anna Wrona, CCB, UC Riverside

Technical Staff:

• Kunihiro Kutijama, CCB, UC Riverside
• Mike Taggart, UCB, UC Riverside
• Thomas Unwin, CCB, UC Riverside

Graduate Students:

• Brian Fulkerson, UCLA
• Hector Estrada, CCB, UC Riverside
• Alisha Glass, CCB, UC Riverside
• Niles Hasselquist, CCB, UC Riverside
• Nigie Shi, CCB, UC Riverside
• Ayesha Sirajuddin, CCB, UC Riverside
• Rodrigo Vargas, CCB, UC Riverside

Undergraduates:

• Jianina Grace Alora, UC Riverside
• Diana Louie, UC Riverside
• Alborz Mehdizadeh, UC Riverside

External Research Partnerships

Collins, Scott. University of New Mexico. Using soil arrays to study monsoonal processes.