SEA Working Paper 99/07

Understanding Monitoring of "Sustainability Indicators" by Farmers: A Case Study of Groundwater Monitoring Under Threat of Dryland Salinity

Sally P. Marsh, Michael P. Burton, David J. Pannell

Agricultural and Resource Economics, University of Western Australia, Nedlands WA 6907

Abstract

Dryland salinity is one of the most pressing land management problems in Western Australia. A number of projects are in progress to provide a more comprehensive picture of the location and extent of potential saline areas in the landscape. Associated with some of these projects, a large number of bores (piezometers) have been installed or are being installed throughout the agricultural area to provide information on depth to groundwater and changes in water levels over time. These bores provide information about whether and when the ground water will reach the surface, causing losses of agricultural production through salinisation of soils. Using data from the Jerramungup Land Conservation District (LCD) we explore factors influencing the behaviour of farmers in monitoring or not monitoring their bores. In 1989, 110 bores were sunk in 7 catchments in the Jerramungup LCD. Monitoring responses were initially exceptionally high, with 96 percent of bores observed in 1990, but then fell steadily to 44 percent by 1997. Our statistical analysis indicates that the probability that a bore will be monitored decreases with time but is influenced by the depth to groundwater, the salt stored in the soil and the bore’s location. As well as these physical factors, we explore some of the sociological and economic factors that influence bore monitoring behaviour. For reasons related to the economic theory of information, the initial groundwater monitoring is relatively valuable. Monitoring is more likely to continue when farmers are clearly able to link the collected information with land management practices, such that the information is of potential economic value.

Key words: sustainability indicators, environmental indicators, resource monitoring, economics of information, hydrology, dryland salinity, Western Australia

Introduction

The clearing of native vegetation to support conventional agriculture in Australia based on annual crops and pastures has reduced the amount of water captured and transpired by vegetation and led to rising water tables. Because the groundwater in most regions is naturally highly saline, dryland salinity is one of the most pressing land management problems in Australia (Anon., 1999) and particularly so in Western Australia (Anon., 1996). However the exact degree and distribution of salinity threat across the landscape is still being investigated. In Western Australia a number of projects are in progress to provide a more comprehensive picture of the current location and potential extent of salininised land. Associated with some of these projects, a large number of bores (piezometers) have been installed and are being installed throughout the agricultural region.

Data from the drilling of piezometers provides information to hydrologists and farmers on factors such as the salt stored in the soil, groundwater conductivity, depth to water and depth to bedrock. If farmers continue to monitor the depth to water, the bores provide information about whether and when the ground water will reach the surface, causing losses of agricultural production through salinisation of soils. This information is one of many possible ‘sustainability indicators’; that is, "environmental attributes that measure or reflect environmental status or condition of change" (Smyth and Dumanski, 1993).

Using data from the Jerramungup Land Conservation District (LCD), where 110 bores were drilled in 1989/90, we explore the factors which influence farmers to keep monitoring groundwater levels. After an initial very high monitoring response, farmers now monitor less than 50 percent of these bores. Disadoption is usually associated with a perception that the practice was not useful, or is no longer relevant (Rogers, 1995). In their review of the application of sustainability indicators in agriculture, Glenn and Pannell (1998) argue that the value of a sustainability indicator is directly related to its potential to improve decision making. In other words, an indicator is as valuable as the information it provides is useful. Additionally, they conclude that many sustainability indicators are strongly technical in focus, with no close link to management. They comment that:

There have been attempts so far to persuade farmers to monitor and use sustainability indicators but apparently the attempts so far have failed. Given the lack of a management focus, this is not surprising. (p. 4)

Bores to monitor groundwater levels are a component of many catchment plans, and many are still being installed by farmers. They represent a sizeable investment by both farmers and society (through government funding). The value of that investment is directly related to the usefulness of the information that can be gained from monitoring groundwater levels. The information from monitoring is potentially useful to farmers, scientists and policy makers, but the responsibility for maintaining the monitoring rests with the farmers. Scientists require information from long-term monitoring to calculate trends in groundwater changes (George, 1999). This investigation will contribute to an understanding of the factors that influence farmers to continue to monitor groundwater levels.

Background

The Jerramungup LCD is located on the south coast of Western Australia within the south-west agricultural region. Like many of the lighter lands in Western Australia, the area is a comparatively new farming district. Some agricultural development took place in the western part of the district in the 1920s as established farming areas around Katanning, Ongerup and Gnowangerup pushed further east (Twigg, 1987). However, the great majority of the district has been settled and cleared since the 1950s, firstly under the War Service Land Settlement Scheme in the 1950s and then in the 1960s when 35,000 hectares were allocated to Conditional Purchase blocks (Davis, 1997). Twigg (1987, p. 12) comments that settlement that occurred under the War Service Land Settlement Scheme at Jerramungup/Gairdner River was "perhaps the largest land clearing venture in Australia at that time".

Dry seasons in the early 1980s resulted in the district experiencing severe wind erosion problems and this, together with an awareness of the insidious nature of increasing dryland salinity, provided the impetus for the formation of a Soil Conservation District Advisory Committee in 1983 (Twigg and Lullfitz, 1990). The Jerramungup Land Conservation District Committee (LCDC) which grew from this beginning was one of the first LCDCs to form in WA. The first catchment group within the LCD, Jacup, formed in 1984, and the second, Corackerup, in 1989. Both these catchments had specified objectives which indicated their concern about dryland salinity (Davis, 1997).

In 1989 the LCDC obtained funding from the National Soil Conservation Program (NSCP) for a network of piezometers to monitor groundwater levels. The original impetus to set up the monitoring scheme came from individuals within this functioning LCDC who were anxious to raise awareness. They were concerned that "some people thought that they didn’t have a problem …. they didn’t believe that they would have a saline watertable" (Jerramungup LCDC members, pers. comm., 1999). The project was supported strongly by Agriculture Western Australia (AGWEST) who were responsible for the drilling of the bores and the collection of the initial data, and committed to providing feedback on the bore data to farmers. There is also an indication that the LCDC were interested in linking groundwater monitoring to farm management. Davis (1997) reports that in an application for continuing funding to the NSCP in 1989/90 for Farm Planning on a District Basis, the specific objectives of the project include measuring the effect of implemented farm works on the watertable through a piezometer network funded by another NSCP project.

In 1989/90, 110 bores were sunk on 81 farms in 7 catchments in the Jerramungup LCD - Gairdner/Bremer, Carlawillup, Needilup North, Corackerup/Ongerup/ Nawainup, Fitzgerald, Jerramungup North, and Jacup. The LCDC was keen to involve as many farmers as possible so most farms only had one bore drilled. AGWEST had no preconceived specific location site for the bores and farmers were consulted about location and encouraged to be present when the bore was drilled. "Involvement was the biggest thing we wanted so bores went where farmers wanted" (Jerramungup LCDC members, pers. comm., 1999). A consequence of this was that many bores were not ideally sited and, for example, placed low in the landscape. The bores were drilled to bedrock wherever possible and as each bore was drilled, samples of the cuttings were collected at regular intervals. Depth to water was measured from the top of the pipe. One water sample was taken from each bore soon after drilling for water quality measurement. A number of bores were ‘dry’ when initially drilled and no water for quality measurement could be taken. Farmers were (and still are) sent quarterly reminders from the local LCDC coordinator to read their piezometers and the information passed on to Agriculture Western Australia for data interpretation. The project had what can only be described as an initial exceptional response, with close to 100 percent of the bores being monitored.

The first detailed feedback on the bores was given to farmers in 1992. The salt profile associated with each bore was presented in a graphical format. The salt attributed to each profile was calculated and expressed as tonnes per hectare and kilograms per cubic metre, the latter measure taking account of the depth to bedrock and giving a measure for average Total Soluble Salt. For bores not drilled to bedrock, the assumed depth to bedrock was used and the last electrical conductivity reading was extrapolated to the assumed depth. Comment was made on the Total Soluble Salt as compared to other bores in the Jerramungup catchment. Information on groundwater readings was given back to the farmers in a graph format and comments made about the depth of the groundwater and any early trend. For example, comments on a particular bore state:

The plot of water level shows seasonal fluctuations superimposed on a rise of around 1 m. Further data is required to confirm that this rise is part of a long term trend. As the water level is within 2 m of ground level there is imminent danger of land degradation in the vicinity of this bore. … Regular monitoring of water level and water quality is strongly recommended. (Greenham, 1992).

A plot of bore water level over time has been made available to farmers each year since 1992. By 1993, 82 bores had sufficient data to enable trends in groundwater levels to be estimated and these trends were presented at the 1993 Jerramungup Agricultural Science Exposition (JERAC). This community-organised science expo brings farmers, researchers, advisers and others together at the one venue to share their experience and knowledge. The trends that were displayed at JERAC were not encouraging, but not surprising to AGWEST hydrologists. Although not intended to be so, the data could have been frightening to individual landholders whose farms were potentially under threat. For example, analysis of the bore data indicated that:

The average rate of rise has been 14 cm/year. This represents a rise of about one metre every seven years, although individual bores were rising by up to one metre every year. Of particular concern, the average depth of the watertable was only 6.5m. … On average, there is over 2,500 tonnes of salt stored under each hectare in the Jerramungup region. Some areas have over 10,000 t/ha. This salt is being dissolved by the rising groundwaters resulting in their average salinity being 2703 mS/m or 14,867 mg/L. This is almost half as saline as sea water (35,500 mg/L). (McFarlane and Ryder, 1993)

The data was also presented on a catchment basis, and this clearly illustrated that trends in some catchments were worse than others. These catchment differences are discussed later in this paper.

At JERAC in 1994, AGWEST presented data from the bores on a landform rather than catchment basis in the form of salinity hazard maps. The maps illustrated that salinity in some areas would be harder to control, with less options available to instigate salinity management strategies. There was some negative reaction by a few farmers to this public disclosure of what was considered sensitive information. For example, there was concern about the potential effect of such information on land values. Because of these concerns a field trip was organised and issues and management options were discussed.

The Upper Gairdner area of the Jerramungup LCD became a Focus Catchment in 1996. The bore monitoring data had suggested that the catchments in this area had higher rates of rise of groundwater levels and higher salt storage and salt concentrations (McFarlane and Ryder, 1993). A Focus Catchment is a catchment designated by AGWEST to receive extra inputs of money and personnel over a limited period (usually three years) to address land management issues in the catchment. The catchment must have an active and enthusiastic catchment group. Catchments with some of the 1989/90 bores in this Upper Gairdner area include Jerramungup North, Corackerup/Ongerup/Nawainup and Needilup North.

By 1993 the number of bores being monitored had fallen to 74 percent and by 1995 it was 52 percent. This approximate monitoring level has continued until the present. Although this is considerably less than the original virtually 100 percent monitoring rate it still represents a high on-going monitoring rate by many standards. Since 1989/90 more bores have been installed in better locations in conjunction with new projects (40 in the Upper Gairdner, 20 in the Fitzgerald), but similar to the existing piezometer network, not all are regularly monitored (Daniel, 1999). Some farmers have been experimenting with new farming systems incorporating perennial pastures and have become district and state ‘champions’ of these changed systems. The Jerramungup LCDC is still very active.

Methodology and analysis of the data

Our analysis focused on investigating the reasons for the drop-off in the level of monitoring whilst explaining the generally high level of initial and on-going monitoring. We used two approaches to do this. Firstly, we conducted Probit analyses using the statistical package STATA to relate the probability that an individual bore would be monitored to the physical characteristics of the bore as described by data such as salt storage, depth to groundwater, etc. Probit analysis is a form of multivariate regression analysis used when the dependent variable is a dichotomous variable with the value of either 1 or 0. In this case we consider an index variable, Y, which takes a value of 1 if the bore is a monitored at a specific time and 0 otherwise. We believe that a set of technical and socioeconomic factors (x), loosely derived from underlying theory, might explain that decision, so that:

The function F should be defined such that the probabilities generated are well behaved, and the normal distribution provides that restriction, giving the Probit model:

(1)

where f and F are the standard normal density and distribution functions respectively.

We expected that the monitoring behaviour of farmers would be explained, at least partially, by characteristics of the initial drilling and on-going trend data from the bore. Our hypotheses were:

• The probability that a bore would be monitored would increase with increased water level, with increased salt storage and groundwater conductivity readings.
• A rising trend in water levels in bores would increase the probability that a bore would be monitored.
• Differences in the monitoring rates between catchments would be explained by the physical characteristics of the bore data.

Secondly, as we did not expect that all the monitoring behaviour would be explainable using the physical bore data we also spoke to AGWEST personnel, the Jerramungup LCDC Coordinator and Jerramungup LCD farmers using a semi-structured interview technique with open questions. Some reasons for monitoring behaviour that were suggested to us by these interviews were subsequently tested statistically. Similarly, ideas originating from results of the statistical analysis were discussed in the interviews to get responses to them.

Description of the data

We were provided with both the initial physical data taken when the bores were drilled and quarterly water levels readings for individual bores (if taken) from 1989. An initial analysis of the data was published by Martin (1992). Additionally, we had access to trend analyses conducted by AGWEST in 1993 and 1996, and by CSIRO in 1999. As previously stated, the data showed that monitoring responses were initially high, 96 percent of the bores were monitored in 1990, but this fell to 74 percent by 1993 and then further to 44 percent by 1997 (see Figure 1). As some farms had more than one bore we investigated whether the monitoring percentage was different when expressed in terms of percentage of farmers monitoring, and Figure 1 illustrates that it is essentially similar. The percentage of farmers monitoring bores varies by catchment. In 1998 it varied from 36 percent in the Gairdner/Bremer/Carlawillup catchment to 70 percent in the Corackerup/Ongerup/Nawainup catchment (see Table 1).

Figure 1: Bore monitoring response in the Jerramungup LCD (110 bores on 81 farms were installed in 1989/90)

The physical data associated with the bores varied greatly between the catchments (see Table 2). This reflected different land forms, soil types and climate variables (McFarlane and Ryder, 1993). The trend analysis done by AGWEST in 1993 showed that only 16 percent of bores (of those with sufficient water level readings) had falling water levels. On average, water level in the bores was rising at the rate of 14 cm per year, although some were rising at rates of greater than 60 cm per year. Jacup and Needilup North catchments showed the highest rate of rise (see Table 2). The trend analysis done by AGWEST in 1996 showed that 37 percent of bores had falling water levels. A preliminary analysis done by CSIRO in 1999, using a different methodology to estimate groundwater trends ( Shao et.al., 1999), estimates that of 68 Jerramungup bores with sufficient readings only 10 percent show an overall falling trend. Another 18 percent of the bores however are measuring shallow watertables with strong seasonal fluctuations where the water is within one metre of the surface, and the remainder display a rising trend of variable type (Crossing, 1999).

Table 1 Percentage of farmers monitoring bores by catchment and year*

* To count as monitoring, farmers must monitor at least one bore once in the year

Table 2 District groundwater data in 1993* (Source: McFarlane and Ryder, 1993)

* The groundwater trends for the Fitzgerald and Needilup North districts may not be accurate as they are based on only six or seven well-monitored bores.

Statistical results

A number of variables were defined for the purpose of the Probit analysis. Our dichotomous Yes=1/No=0 dependent variable was defined as whether or not the bore was monitored in each quarter (February, May, August, November) for the years 1989 to 1998. The first reading for each bore was ignored in the analysis as it represents the ‘test’ reading done at installation rather than a decision by the farmer to monitor. We were aware that there could be a number of practical reasons why a bore might not be monitored. For example, the bore might be dry or have been damaged so that water level could not be read, but our data does not allow us to distinguish which, if any, bores are not monitored for such reasons. No socio-economic data is available to be included in the statistical analysis, only technical data related to the bore readings and the bore location are included. The independent variables investigated are listed in Table 3.

Table 3 Independent variables used in the Probit analyses

(*) Only these variables vary across time for each bore: SALTSTORE etc. relate to measurements made at the initial reading of the bore.

From casual inspection of the data, we hypothesised that the probability of reading a bore will decline over time: this may be due to failure of the bore, a loss of interest in the project, or a perception that there is no further information of value to be gained from monitoring. Given the different dates at which bores were drilled, the measure of time elapsed is conditioned on the date of the first reading, which occurred when the bore was installed. However, we anticipated that the severity of the problem (i.e. higher water tables and increased salt) would increase monitoring. The appropriate measurement of these variables was something explored within the analysis, by including levels and changes in distance to groundwater, and alternative definitions of salt load. One problem faced in the analysis is that once a bore is not monitored there is no information generated on water levels. We therefore define the measure of groundwater level as that at the most recent reading, and the change in water level as the most recent recorded change in water level, prior to the current quarter. We also explored the possible interaction between salt load and depth to water, on the expectation that high or rising water tables may not have so great an impact on monitoring response, if they have a low salt load.

Two results are statistically very robust across all specifications. Changes in water level are not associated with monitoring behaviour, while water levels are, and it is the total measure of salt storage which is the most significant variable, and not ground water conductivity or average salt concentrations. All of these variables were available to farmers at the start of the monitoring process. In theory, total salt storage will not be a good estimate of the potential salt problem, as it is a function of the distance to bedrock, and will not necessarily be a good indication of potential impacts on farm productivity. However, despite the 3 measures being highly correlated (ranging from 0.5 to 0.8) it is salt storage which appears to be the variable which influences monitoring behaviour.

As a general modelling strategy, quadratic terms were included to allow for flexibility in the response function. Furthermore, the coefficients for GWLEVEL and (GWLEVEL)² were allowed to vary as a function of (logged) salt storage. Both TIME and (TIME)² are used, and a dummy variable was also included to identify if there was any change in monitoring after the public presentation of results in 1993. Other significant variables were the rainfall in the previous quarter, catchment and season dummies and distance from the sea.

The results from the final specification are reported in Table 4 below. As noted, dF/dx reports the change in probability of monitoring following a unit change in the exogenous variable, or, in the case of dummy variables, a switch from 0-1. In each case all other variables are at mean levels. This gives some indication of the magnitude of the effects. These measures are not reported for variables that have quadratic or interaction terms, as the individual marginal impact has no sensible interpretation in those cases.

Table 4 Results of the Probit analysis, Number of observations = 3615, Pseudo R² = 0.1847

Z is the ratio of coefficient to standard error, P the significance level. Standard errors corrected for clustering by bore. dF/dx is the change in probability of monitoring, for a discrete change of dummy variable from 0 to 1, or for a unit change in other variables, all other variables measured at their mean.

Baseline catchment is Gairdner/Bremer, baseline season Jan-March.

As a result of the analysis we have to reject our second and third hypotheses, that changes in the water levels in bores would increase the probability that a bore would be monitored and that differences in the monitoring rates between catchments would be explained by the physical characteristics of the bore data. Change in water level was not significant in any specification used and there are still significant catchment effects, even allowing for the physical data available to us. Bores located in the Corackerup/Ongerup/Nawainup, Jacup and Carlawillup catchments are significantly more likely to be monitored than those in the baseline catchment, Gairdner/Bremer.

We have identified some unexpected impacts: rainfall in the previous period is negatively correlated with monitoring, which may be due to a view that high rainfall biases the readings, and therefore it is not worth doing, or may be some more complex interaction with water level. Bores are less likely to be monitored in May and November: times which coincide with peak workloads on farms for sowing and harvesting. There is also a very robust relationship between monitoring and distance to the sea. Given the geographical distribution of bores across catchments, this is more than a "catchment specific" variable, and must be identifying some hydrological aspect of the problem, but at the moment it is unclear what this may be.

Interpretation of the impacts of time, water level and salt storage is complicated by the non-linear and interaction terms included in the model. The effects are shown in Figures 2 and 3 for representative bores. Figure 2 gives the evolution of the probability of monitoring as time elapses, assuming the bore was first monitored in quarter 1 1989. This figure shows a relatively constant rate at the start (with the quadratic function giving a slight rise) but with the onset of a decline at around 9 quarters. The step in the function is the large negative impact of the 1993 dummy, which is strongly significant, and which seems to suggest that the public presentation of the results did have an impact on farmers’ perceptions of the value of further monitoring. The probabilities then decline further with time.

Figure 3 gives the relationship of water depth on monitoring, for 3 different levels of salt storage. Here the interaction between salt storage and the quadratic leads to distinct changes in behaviour. At higher levels of salt storage there is a confirmation of the hypothesis that if the water table is deep, the incentive to monitor is low. It also indicates a possible effect of very high water tables leading to reduced monitoring, as the problem becomes self-evident, or overwhelming. For a wide range of depths, higher salt load is associated with higher rates of monitoring. At lower salt loads the shape of the curve is inverted, but there is a tendency for low salt loads to be associated with lower probabilities of monitoring. At the tail of the distribution this is reversed, but it should be noted that there are relatively few bores that have actual observations in this range (e.g. there are no observations with salt levels less than 300 and depths to water exceeding 17m). The nature of the quadratic generates the result that all curves pass through the two fixed points, irrespective of load, and this may also be biasing the estimate of the response function. There may well be benefit in exploring more flexible specifications for the interaction.

It should also be noted that there are a number of limitations to the statistical approach employed here. Firstly, the standard probit model assumes that the error terms are independent, but in this case we have repeated observations on the same bores, and so this may not be true. We have allowed for that to some extent by estimating robust standard errors, which assumes independence between clusters (defined here as bores), but is robust to assumptions about within cluster correlation (StataCorp., 1999). However, the overall results have been remarkably robust to other statistical specifications (such as a Random Effects Probit model, or explicitly modelling the structure of within-bore correlation of errors). An alternative approach that may also be fruitfully explored is to treat each year as an observation, and apply a count model to the number of times the bore is monitored in each year. This may overcome a problem of farmers selectively deciding to monitor at a low frequency each year, but continuing to monitor.

Overall, the physical data relating to the bore is an incomplete predictor of whether or not a bore will be monitored; the explanatory power is only approximately 18 percent using a psuedo R², defined as 1-L₁/L₀, where L₁ and L₀ are the log likelihood values for the full and constant only models. The distribution of actual versus predicted monitoring of bores generated by the model (assuming a 50 percent cutpoint) is reported in Table 5.

Table 5. Predicted v. actual monitoring

Although illustrative of how the model works, such a table, or estimates of the proportion of correct predictions, should not be used as a measure of the goodness of fit of the model (Veall and Zimmermann, 1996). Instead we report one of the "R²" type measures, given by

(2)

where p_ij is the fraction of times the realisation was outcome i when the model predicted outcome j, and p_~j is the fraction of times alternative j is predicted. s _n is positive for a model with any predictive power, and bounded at the upper limit by unity. A value of 0.434 is returned, which indicates a relatively high level of fit.

Our analysis has to some extent mechanistically modelled the process of the fall-off in bore monitoring, but it doesn’t explain why people drop out or keep going. The TIME and '93 dummy variables don’t give us any idea why the monitoring has stopped, they just describe how it does. The following section explores this issue further.

Interview responses

We spoke to AGWEST personnel, the LCDC coordinator, and a number of Jerramungup LCD farmers using a semi-structured interview technique and obtained the following impressions, which as such are unsupported by ‘rigorous’ data. The interview questions focused on teasing out reasons for the initial high success rate of the monitoring project, reasons for continuing to monitor and failing to monitor, the value of monitoring and why there might be differences in monitoring behaviour between catchments. Responses to all the interviews as interpreted by the interviewer were returned to the interviewees for comment and feedback.

The initial high success rate of the monitoring project

The level of involvement and commitment by both the Jerramungup LCDC and AGWEST to the bore monitoring project has been discussed in the Background section of this paper. It was indicated to us that involvement of farmers in the project was obtained by key individuals in each area who had been ‘fingered for the job’ by the LCDC. Reporting results from a survey conducted in two Jerramungup catchments in 1993, Davis (1997, p. 57) says that "most participants cited the same 2 or 3 people as the key person who invited them to join the Farm Planning Project". This project was instigated at a similar time to the bore drilling and indicates that the recollection of how involvement was achieved is likely to be correct. There were however some mixed responses about the level of overall participation, with one comment being that the "initial project was ill-conceived, some people weren’t asked, the bores were not well-sited".

The value of monitoring and reasons for continuing to monitor

Farmers considered that the monitoring project had been useful in contributing to a general awareness of the extent of saline rising groundwater within the LCD. It was also evident that farmers had learnt from monitoring their bores. Examples of the learning that had occurred include:

• Some hydrological concepts were better understood. One farmer explained he had previously perceived the ‘watertable’ as something flat (like a table) and had not understood how it could be higher then the creek without running out the bottom. There was also a realisation by some that recharge happens throughout the landscape and "there is no magic recharge area".
• Farmers who had plotted their rainfall against the groundwater level had gained an appreciation of the fluctuations in groundwater level caused by seasonal rainfall variation.
• One farmer indicated that he had learnt that surface water management was not going to be enough to control a rising watertable. "I thought that annual pastures and surface water management would be enough but the piezometers have shown otherwise."
• Some farmers had begun to experiment with how they could use bores and realised that they didn’t all have to be drilled to bedrock. More piezometers are being used; one farmer had installed 20 piezometers, some shallow, as a transect across a lucerne paddock.

Some farmers considered that monitoring is useful for management and that data from regular monitoring starts to "tell a story". Some had started treatments: for example, sowing lucerne on problem patches or instigating surface water management strategies. One farmer indicated that he had stopped monitoring but plans to start again in conjunction with management options. A number of farmers said that "if piezometers form part of the land management plan it is likely they will then be read more frequently". However there was seemingly some farmers who could not see any further value for them in monitoring. This was evident in comments such as "What are we to do with this information?", and "Measuring is most use for the LCDC not for us".

AGWEST are seemingly still interested in the bore data but there were some mixed messages about whether quarterly data from this district is particularly needed for regional hydrology purposes. AGWEST say farmers should be monitoring for themselves, farmers say they are monitoring for the catchment or AGWEST. This is possibly the consequence of AGWEST’s considerable initial involvement being perceived by farmers as the main reason for the project. The current focus and interest within AGWEST is on assessing management options, particularly ‘high water use’ systems. Groundwater monitoring is an integral part of this program.

Reasons for not monitoring

Reasons given for not continuing to monitor were heavily focused on ‘practical’ explanations. For example: the bore being in the wrong place, the bore dry or water table very low, the bore being damaged (e.g. by stock), or that they didn’t have the measure in the vehicle when they drove past. As a result of our visit and discussions with Jerramungup farmers, the LCDC suggested to us that it would be useful for AGWEST to revisit and service bores as their expertise was needed to service and check salt levels, etc. The LCDC was prepared to coordinate such an activity.

Another reason put forward for not monitoring was that with only one piezometer the monitoring job is too small and so gets easily overlooked. To test this idea we introduced a multiple bore variable (MULTI) into the analysis, but this did not significantly affect the regression. Most bores, however, in this data set are single ones on the property, with only 11 properties having 2 bores, 6 properties having 3 bores and one having 5 bores. It was also pointed out that properties get sold and the new owners may not continue with monitoring as they were not originally involved in the project.

Other reasons focussed on the value of the information from monitoring. A comment was made that data from one bore is not representative and that "a lot of piezometers were needed to tell the truth". Both AGWEST personnel and farmers suggested that a succession of drier years may have reduced the perceived need to monitor. The early signs of salinity are often linked to waterlogging, and after a wet year in 1993, 1994 had been dry, and rainfall in 1995 average and less than average in 1996 . It was suggested a common attitude was that when there’s "not so much water around what’s the point of monitoring". To attempt to test this idea we introduced quarterly rainfall data from the Jerramungup Post Office to the statistical analysis. The rainfall during the immediately preceding quarter did significantly affect the probability of monitoring, but not in the way that had been suggested. Higher rainfall in the preceding quarter reduced the probability of monitoring.

AGWEST personnel commented that farmers seem less interested now in feedback (e.g. of groundwater trends) than earlier in the project. Responsibility for feedback on groundwater trends has now been devolved to the LCDC coordinator, who is also responsible for sending out the quarterly reminders to monitor. Farmers who had not been monitoring were recently asked if they still wanted to be sent quarterly reminders and some replied no. It is possible that some farmers don’t choose to pass on groundwater information as they don’t want the data to be public knowledge.

Both AGWEST staff and the LCDC coordinator said that the salinity issues in the Jerramungup LCD had caused some feelings of despondency and despair amongst farmers and their families. This was evidenced by people becoming distressed at meetings, and speaking of their feelings of hopelessness. There was also evidence of some hostility towards professionals working in the Landcare area and anger at the salt ‘predictions’. An earlier report, the Fitzgerald/Mallee Rd project, was mentioned as having predicted dire consequences which failed to eventuate to the extent predicted.

Reasons for the differences in monitoring behaviour between catchments

There are quite distinctive soil and water differences between the catchments. The soils vary in the district from north to south, ranging from well-drained silty sediments at Gairdner/ Bremer/Carlawillup to poorly drained heavy clays at Corackerup/Ongerup/Nawainup. Rainfall is higher near the coast and water tables tend to be lower (McFarlane and Ryder, 1993). These soil and water differences are reflected in the characteristics of bore data, such as depth to water, salt storage and groundwater conductivity, reported in Table 2. However, the statistical analysis showed that there are differences in monitoring behaviour that can not be explained by differences in the technical bore data that we investigated. Farmers suggested that there are different types of water problems in different areas, such as water erosion, inundation, and waterlogging, that might affect the perceived value of information from monitoring groundwater levels. This may provide an explanation of the significance of the DISTANCE variable in the statistical analysis, as these problems are related to landform, soil and climatic factors.

Additionally, it is evident there are social differences between the catchments. As discussed briefly in the Background section, areas within the district were settled at different times, resulting in different social groups. In contrast to communities around Jerramungup which settled since the 1950s, the settlement in Needilup happened early in the century and there have been families in this area since 1912 (Twigg, 1987). The recently settled communities often have a strong sense of unity and common purpose. Davis (1997, p.43), for example, comments that:

"The Jacup catchment group had its origins in the cooperation of the Conditional Purchase settlers who worked together to clear the land and fight the government to obtain the infrastructure and community facilities they needed. There is a strong sense of community in Jacup".

Finally, some of the catchments, namely Jerramungup North, Corackerup/Ongerup/Nawainup and Needilup North, formed part of the Upper Gairdner catchment which became a Focus Catchment in 1996. It is possible that this has affected farmers’ bore monitoring behaviour. Certainly, Table 1 suggests that a greater percentage of bores in the Corackerup/Ongerup/ Nawainup catchment are monitored than in other catchments, but this is not consistent across the other two catchments involved in the Focus Catchment. Fitzgerald and Jacup catchments are likely to become a Focus Catchment in the near future and it is possible that anticipation of this has contributed to the higher monitoring percentage seen in Fitzgerald in 1998 (see Table 1). Jacup has maintained a consistently relatively high monitoring percentage.

Discussion

What is the value of monitoring groundwater levels?

For individuals the first value in monitoring lies in a greater awareness of the salinity threat and how it relates to their land and the district - "they believe the data if they measure it". The monitoring carried out by farmers in the Jerramungup LCD, combined with the feedback and interpretation that was provided by AGWEST, allowed farmers to quickly become aware of the threat posed to the district by saline rising groundwater. The second value clearly evident is that groundwater monitoring can result in a substantial degree of learning, both of hydrological processes and also learning that can lead to monitoring being perceived as a useful management tool. Monitoring can be used as a management tool in two different ways:

1. to assess the effect of a particular management treatment, and
2. as a indicator of when a particular management tool (e.g. lucerne phase of a rotation) needed to be implemented (i.e. as a tool to know when to act).

For catchments, groundwater monitoring has the potential to create a district awareness that is necessary to gather local support for district initiatives to obtain funding and support to address salinity issues. Once that funding has been obtained, continued monitoring serves a number of purposes. It provides information to funding bodies and government agencies that addresses accountability requirements, such as data that plots district trends, records the response to different management options, and contributes hydrological information to large scale projects. Further to this it helps in ‘creating an impression’ of awareness and willingness-to-act that attracts both outside expertise and further funds for a range of Landcare and production purposes.

Why do farmers continue to monitor groundwater levels?

That bores with higher water levels and higher salt readings have a higher probability of being monitored indicates that farmers continue to monitor groundwater levels because they are concerned. However, discussions with farmers indicate that the most powerful reason to continue monitoring is if the monitoring is linked to management options, such as lucerne or surface water management. Associated with this is a desire in some cases to "prove a point", especially if it is against conventional wisdom or the law. There are farmers who wish to clear further areas of their land (an action currently prevented by law) and anxious to demonstrate that tagasaste, lucerne or other perennial alternatives will substitute hydrologically for native vegetation.

Farmers also continue to monitor bores out of habit and/or a feeling of responsibility. Many have a genuine interest in the figures and are keen to discuss them with hydrologists and other professionals. Continued monitoring often provides links to expertise and individuals who wish to use the data for research reasons. Finally, there are peer and social reasons which influence farmers’ monitoring behaviour. One example of this which might lead to increased monitoring is provided by the location of the farm within a Focus Catchment.

Why do some farmers seemingly fail to monitor groundwater levels?

There are a range of practical reasons why individual bores are not monitored. Given that, our analysis suggests that bores in situations where the salinity threat is less serious (i.e. lower water levels, lower salt storage in the soil) are monitored less frequently.

It also appears to us that farmers cease to monitor a bore because they have obtained all the information they want or can see how to use. The value of any information is related to its ability to reduce uncertainly about a situation (Pannell, 1999). There are two possibilities in this situation. Firstly, uncertainty about the situation may be quickly reduced following a small number of readings of groundwater levels. In this case, awareness that groundwater was saline and rising was achieved with 3 to 4 years of the commencement of the project. It might then be perceived that there is no further need to monitor, or that monitoring may only need to be done infrequently (e.g. not quarterly or even yearly). This awareness, and not the actual public release of information in 1993, may be the reason for the fall-off in monitoring after 1992, and the significance of the ’93 dummy variable in the probit analysis. Secondly, uncertainty about the relationship between groundwater levels and on-farm strategies may not be reduced by monitoring; that is, the information is not useful to farmers in a tactical sense. In that case there is little point (for farmers) in monitoring after initial awareness needs are met.

Associated with the awareness that results from initial monitoring, there appears to be psychological reasons that dissuade some farmers from further monitoring. There is a limit to how much "continual bad news" people can take, especially if they feel disempowered and unable to act to solve the problem. We suggest that this could particularly be a problem for farm women who take on the responsibility for monitoring bores. Even if alternative farming systems exist, the stress, learning and risk associated with changing farm practices should not be forgotten (Marsh, 1998, Pannell, 1999).

Finally, there are undoubtedly peer or social reasons that influence farmers’ bore monitoring behaviour. We have not investigated this in any real depth, but have indicated some of the factors that could conceivably play a part.

Conclusions

The Jerramungup LCDC has been recognised for their Landcare efforts, winning the National Landcare Award for Landcare groups in 1991. Despite the focus of this paper on reasons for failure to monitor, an enviable initial and continual bore monitoring response has been achieved in the Jerramungup LCD. A number of factors have contributed to this. Firstly, a high degree of community ownership of the program was achieved through the efforts of the Jerramungup LCDC which focussed on involving as many farmers as possible in the project and allowing them to decide where the piezometers were located. Other projects recognise the importance of involving people at early planning stages for maintenance of a monitoring program (Kenny, 1998). Nicholson (1996) also reports that farmer committees recognise that involvement at any level is preferable to non-involvement, a concept not often accepted by agencies. In this case, more suitable bore siting was sacrificed for the principles of involvement and ownership. Other key reasons for the success of the Jerramungup program have been the commitment of AGWEST to providing support and feedback to the project, and the co-ordinating and motivating role played by the LCDC coordinator.

Our analysis shows that the physical characteristics of the bore data have some influence on monitoring. Data related to the bore such as higher water level and higher salt storage appear to influence the likelihood of a bore being monitored more frequently. The interaction between water level and salt storage is also significant. This makes intuitive sense as it is rising water levels in soils with high levels of salt that poses the most serious salinity threat. Interestingly, despite a positive correlation between salt storage (dependent on the depth to bedrock and expressed in tonnes per hectare) and both ground water conductivity and average salt storage (standardised for depth and expressed in kilograms per cubic metre), both the latter variables were not significant if substituted in the regression for salt storage. We suggest that the high figures quoted for salt storage may have a powerful influence on a farmers’ perception of the potential salinity threat.

Groundwater monitoring does appear to be a powerful awareness tool, but some farmers discontinue monitoring even though they have a rising saline water table. Pannell (1999) suggests that the usefulness of information is related to its ability to reduce uncertainty. The information from groundwater monitoring can reduce uncertainty about the nature of the groundwater fairly quickly and dramatically, and any reason to continue monitoring then needs to be linked to information that further reduces uncertainty. For this reason strong and clear links to management options make continual monitoring make sense to farmers, as suggested by Glenn and Pannell (1998) and Kenny (1998). Farmers in the Jerramungup LCD who spoke enthusiastically about the value of continued monitoring were evaluating farming systems options such as lucerne, perennial grasses and surface water management. They talked of having a "piezometer in every paddock". It is not clear how many bores are being read with a management perspective but we would suggest not many as yet. Many piezometers are still being installed with awareness issues and regional hydrology being perceived as providing the motivation for long-term farmer monitoring. There is a need to involve farmers in R&D related to the implementation of high water use systems on farms and linking this to groundwater monitoring.

Social and psychological factors appear to be major influences on both failure to monitor and continuing to monitor. However, we have not explored these factors in any real depth. Given this caveat, it appears that social factors are responsible for the differences in monitoring behaviour between catchments, and that feelings of discouragement or despair can dissuade farmers from continuing to monitor. A growing awareness of the problem and an inability to see a way forward puts people in a vulnerable position; both to inappropriate responses (e.g. promises of quick fixes) and despair/hopelessness. We see evidence that support people are coming under increasing pressure as the considerable threats posed by rising saline groundwater is recognised in rural communities. Unaddressed, this will lead to problems associated with ‘burn out’ and a consequent loss of community capacity at a time when it is increasingly needed. The social impacts in rural communities of the salinity problem need to be recognised and addressed. This includes at a very minimum linking-in hydrologists and LCDC coordinators with appropriate existing services.

It may also be possible that farmers are monitoring bores but choosing not to make that information available to AGWEST. There are many valid reasons why farmers may perceive that this information should remain confidential to themselves. Our analysis shows a significant fall-off in monitoring after 1992. This could be related to the public release of information in 1993, or it could be related to a perceived reduction in the value of monitoring after awareness needs had been met. We have no conclusive evidence that the public release of information affected bore monitoring in the Jerramungup LCD but suggest that the ownership of regional data that comes from farmer bore monitoring is a key issue to address. Ownership of data and conditions for its release have already been raised as issues in other catchments in WA (George, 1999). Kenny (1998) suggests that permission should be sought for data to be disclosed.

There was evidence of some confusion regarding the reason to monitor after initial levels, trends and awareness are achieved. It is important to consider exactly to whom the information is useful (Glenn and Pannell, 1998; Kenny, 1998)). It may be that the long-term monitoring of many bores situated on private land is of more interest to regional hydrologists than to individual farmers. If this is the case, then issues related to continued monitoring need to be addressed. Bores will only be monitored in the long term by farmers if the information they get from doing so contributes to their ability to make better decisions about their farm business. This means being able to clearly link the collected information with land management practices, such that the information is of potential economic value.

Acknowledgements

The authors acknowledge the contribution to this work of Don McFarlane and Arjen Ryder (Agriculture Western Australia), Carolyn Daniel (Jerramungup LCDC) and farmers in the Jerramungup LCD.

References

Anonymous, 1996. Salinity: A Situation Statement for Western Australia. A Report to the Minister for Primary Industry and Minister for the Environment, WA Government, November 1996.

Anonymous, 1999. Dryland Salinity and its Impact on Rural Industries and the Landscape. Occasional Paper Number 1, Department of Industry, Science and Resources, Commonwealth of Australia, Canberra.

Crossing, L., 1999. Agriculture Western Australia, pers.comm.

Daniel, C. 1999. Jerramungup LCDC Coordinator, pers.comm.

Davis, J., 1997. "The Effectiveness of Public Funding for Landcare: a Case Study from Jerramungup, Western Australia. Thesis submitted for the degree of Master of Science (Natural Resource Management) in the Faculty of Agriculture at the University of Western Australia.

Glenn, N.A. and Pannell, D.J., 1998. "The Economics and Application of Sustainability Indicators in Agriculture". Paper presented at the 42^nd Annual Conference of the Australian Agricultural and Resource Economics Society, University of New England, Armidale, Jan 19-21, 1998.

George, R., 1999. Agriculture Western Australia, pers.comm.

Greenham, K., 1992. "Collation and interpretation of Jerramungup LCDC bores, 1992". A report compiled for the Jerramungup LCDC and Western Australian Department of Agriculture.

Kenny, N., 1998. "Avon Water Quality: Sustainable Community Water Quality Monitoring", in Proceedings of the National Waterwatch Conference Getting better at getting wet. University of Adelaide, Roseworthy, 21-24 July 1998, pp. 159-164.

Land and Water Resources Research and Development Corporation website – www.lwrrdc.gov.au/ndsp/projects/phase-1.htm

MacFarlane D. and Ryder, A., 1993. "Groundwater Trends in the Jerramungup Region. A report compiled for the JERAC EXPO, October 1993.

Marsh, S.P., 1998. "Researchers and the Sustainability Message". Agricultural Science 11: 4, 29-32.

Martin, S., 1992. Groundwater investigations in the Jerramungup Shire, Technical Report No. 122, Division of Resource Management, Agriculture Western Australia, Perth.

National Land and Water Resources Audit website – www.nlwra.gov.au

Nicholson, C., 1996. "Woady Yaloak – What can we learn". In Proceedings of the Australasia Pacific Extension Network Conference Beyond Technology Transfer, University of Melbourne, December5-6, 1996.

Pannell, D.J. 1999. Social and Economic Challenges in the Development of Complex Farming Systems, Agroforestry Systems 45(1-3): 393-409.

Rogers, E.M., 1995. Diffusion of Innovations. The Free Press, New York.

Shao, Q., Campbell, N.A., Ferdowsian, R. and O’Connell, D., 1999. Analysing Trends in Groundwater Levels. Report Number CMIS 99/37, CSIRO Mathematical and Information Sciences, Floreat, Western Australia.

Smyth, A.J. and Dumaski, J, 1993. FESLM: An international framework for evaluating sustainable land management. World Soil Resources Report 73. Food and Agriculture Organisation of the United Nations, Rome, Italy.

StataCorp., 1999. Stata Statistical Software: Release 6.0. College Station, TX: Stata Corporation, pp. 65-79.

Twigg, B., 1997. "Settlement and Landuse History Fitzgerald – Western Australia". Independent study contract at Murdoch University, Perth.

Twiff, B. and Lullfitz, B., 1990. "Landcare in W.A. – A look back … then forward". Australian Journal of Soil and Water Conservation 3: 54-55.

Veall, M.R. and Zimmermann, K.F. 1994. "Goodness of fit measures in the Tobit model", Oxford Bulletin of Economics and Statistics 56: 485-499.

Citation: Marsh, S.P., Burton, M.P. and Pannell,D.J. (2001). Understanding farmer monitoring of a ‘sustainability indicator’: Depth to saline groundwater in Western Australia. In: A. Conacher (ed.), Land Degradation, Kluwer, Dordrecht, pp. 207-222. (SEA Working Paper 99/07, Agricultural and Resource Economics, University of Western Australia).

The SEA News home page is now at http://www.crcsalinity.com.au/newsletter/sea/

Copyright © 1999 S. Marsh, M. Burton and D. Pannell
Last revised: June 16, 2013.