Aarvark Percolation Tests

Aardvark permeameter

Test procedure

The Aardvark permeameter (Soilmoisture Equipment Corp., USA) is a constant-head permeameter to measure the soil saturated hydraulic conductivity K_sat. A test hole is drilled with a 10 cm diameter hand auger to the desired depth and percolation rates are measured through oberserving the volume of supplied water infiltrating into the bottom and side walls of the test hole over time by means of a permeameter module ensuring a constant water level in the test hole (water is consistently added in a constant-head method). After a steady infiltration rate is reached, the resulting K_sat can be determined.

Sites

Sites used for Aarvark percolation tests and respective T-values

ID	Location	County	T-value
CC	Crecora	Limerick	35.6
CF	Coolflugh	Cork	15.0
DL	Drumlargan	Meath	41.0
JT	Jealoustown	Meath	46.1
KM	Kilmallock	Limerick	13.3
LU	Lusk	Dublin	NA
MC	Mullincross	Louth	NA

A total of 18 permeameter test were conducted on 7 unique sites in 5 Irish counties (Cork, Dublin, Limerick, Louth, Meath) from April 2015 to May 2017. For all sites, traditional T-test percolation results were availabe obtained from T-tests performed according to the EPA guidelines prior to the Aardvark tests.

Geometry sensitivity analysis

According to the manufacturer, several methods are available to compute K_sat values from the percolation rate. Using the USBR 7300-89 procedure (well permeater method), three cases have to be distinguished:

• Case I: The water table is sufficiently low (to be exact, the distance between the historical high water mark and the water surface in the well is at least three times the height of water in the well).
• Case 2: The water table is moderately low (to be exact, the distance between the historical high water mark and the water surface in the well is less than three times, but at least equal to, the height of water in the well).
• Case 3: The water table is close to the well bottom (to be exact, the distance between the historical high water mark and the water surface in the well is less than the height of water in well).

The respective equations to be used to compute the hydraulic conductivity depending on the groundwater table and test geometry are available in the Aardvark manual and are implemented as a function in R as follows:

calcPercFactor = function(d, z, h, g) {
  w = 9 + 0.003 * (z + h - 18.5)
  x = 2*w/d
  y = (g - z + w)/w
  f = ifelse(y>3 | is.na(g), 
           1440* ((log(x + sqrt(x^2+1)) - sqrt(x^2+1)/x + 1/x)/(2*pi*w^2)),
         ifelse(y<=3 & y>=1,
           1440* ((log(x)/(1/6 + y/3))/(2*pi*w^2)),
         ifelse(y < 1, 
           1440* ((log(x)/(y + y^2/2))/(2*pi*w^2)),
         NA)))
  return(f)
}

This function takes the well diameter d, the well depth z, the water reservoir elevation h, and the depth g of the water table as input and returns the normalized percolation factor f for each site-specific measurement setup. To obtain K_sat, the test-specific f has to be multiplied by the steady-state percolation rate. In all percolation tests in this study the groundwater water was sufficiently low (case I; established though trial holes). A sensitivity analysis to visualize the influence of site-specific factors on f only needs to focus on the distance H = z + h between the well bottom and the water reservoir, and the well diameter d.

For any given set of test geometry, the well diameter is the most crucial influencing factor on the accuracy of obtaining correct percolation factors. The obtained f vary widely across different well diamaters d, but not over H.

Test results

The percolation test results are shown below. For each percolation test, the change of reservoir water level V is graphed over time t as elapsed minutes since start of the test. The test was stopped when a constant percolation rate has been observed over several readings (constant gradient dV/dt).

Each test was conducted in a separate test well and variations in percolation rates represent spacial heterogeneieties of the natural soil. However, as in the traditional T-test, sometimes these variations can be rather large, as in the case of Jealoustown (JT). The first two test showed a very low percolation, while as the thirs test was more in line with the expected percolation rates (from previous T-test results). The sites in Lusk and Mullincross both failed the traditional T-test.

Calculating K_sat

According to the manufacturer, K_sat should be calculated using the percolation rate from the last two readings of a test. This, in practise, is easy to do and, assuming a constant percolation rate has been reached, also delivers an acceptable result. However, during the field trials it has been noted that there will always be slight fluctuations in percolation rates, especially when readings are taken in relatively short time intervals. There are two main reasons for this: firstly, even under saturated conditions, there might be natural fluctuations or inhomogeneities in the surrouding soil causing, temporarily, changes in observed percolation rates; and secondly, the scaling of the provided water container has a resolution of 100 ml and reading might only be accurate enough to the nearest 50 ml. Especially in slowly percolating conditions, this might make it difficult to mark the exact time of the water level passing such a mark. A revised water container design (long and thin, like a graduated cylinder) and/or automated weighting of the water reservoir would improve the accuracy of the test.

Several different approaches are presented to compare the determination of percolation rates.

• Endpoint Method: Percolation rates are calculated from the last two recorded readings. This method assumes a constant percolation rate is reached and the two last recorded readings are representative for the overall steady-state percolation in the test well.

• Linear Method: Percolation rates are calculated using a linear best-fit model using all but the initial two recorded readings. The initial readings might be biased towards higher percolation rates following the expanding wetting front in the soil and might not be representative for fully saturated conditions. This method requires at least 3 readings. In case of 3 available readins, it is equivalent to the endpoint method.

• Cut-off Method: Percolation rates are calculated using a linear best-fit model using the first two consecutive readings with a difference in percolation rate lower than 10%. This model is similar to the endpoint method, but employs an early cut-off rule to find the initial time when the percolation rate could be assumed constant (to a certain degree).

Overview of percolation rates [ml/min] per test

	Percolation Rate [ml/min]			Ksat [cm/d]
Test ID	Endpoint Method	Linear Method	Cut-off Method	Endpoint Method	Linear Method	Cut-off Method
CC-a	2.50	2.88	2.50	4.81	5.53	4.81
CC-b	13.33	11.16	10.00	23.80	19.91	17.85
CC-c	9.26	7.36	13.64	18.61	14.78	27.41
CC-d	2.50	2.64	2.67	5.19	5.49	5.53
CF-a	20.00	24.51	25.00	39.64	48.57	49.55
DL-a	6.67	6.67	6.67	12.02	12.02	12.02
DL-b	8.33	8.33	8.33	15.64	15.64	15.64
JT-a	0.97	NA	NA	1.93	NA	NA
JT-b	0.67	NA	NA	1.38	NA	NA
JT-c	33.33	31.67	32.50	64.47	61.25	62.86
KM-a	13.29	14.03	NA	23.81	25.14	NA
KM-b	20.00	18.55	19.74	38.29	35.52	37.79
KM-c	13.37	14.52	13.37	24.74	26.86	24.74
KM-d	21.67	23.65	21.67	43.46	47.44	43.46
KM-e	11.43	12.55	11.33	23.81	26.16	23.61
LU-a	2.38	3.22	NA	4.73	6.39	NA
LU-b	0.83	0.60	NA	1.65	1.18	NA
MC-a	0.43	0.43	NA	0.83	0.83	NA

Overview of percolation rates [ml/min] per site

		Percolation Rate [ml/min]						Ksat [cm/d]
		Endpoint Method		Linear Method		Cut-off Method		Endpoint Method		Linear Method		Cut-off Method
ID	Site	mean	SE	mean	SE	mean	SE	mean	SE	mean	SE	mean	SE
CC	Crecora	6.90	2.67	6.01	2.03	7.2	2.77	13.10	4.80	11.43	3.58	13.90	5.41
CF	Coolflugh	20.00	NaN	24.51	NaN	25.0	NaN	39.64	NaN	48.57	NaN	49.55	NaN
DL	Drumlargan	7.50	0.83	7.50	0.83	7.5	0.83	13.83	1.81	13.83	1.81	13.83	1.81
JT	Jealoustown	11.66	10.84	NA	NaN	NA	NaN	22.60	20.94	NA	NaN	NA	NaN
KM	Kilmallock	15.95	2.04	16.66	2.01	NA	NaN	30.82	4.19	32.22	4.23	NA	NaN
LU	Lusk	1.61	0.77	1.91	1.31	NA	NaN	3.19	1.54	3.78	2.60	NA	NaN
MC	Mullincross	0.43	NaN	0.43	NaN	NA	NaN	0.83	NaN	0.83	NaN	NA	NaN
* NaN values in the mean represent sites where the specified method did not succeed to calculate a percolation rate. NaN values in the SE represent sites where only a single measurement was used to calculate the percolation rate.