Summary of Rissman and Carpenter’s ‘’Progress on Nonpoint Pollution’’

Progress on limiting nonpoint source pollution is slow for many reasons, but limiting pollution is still a worthy endeavor.
water quality

September 9, 2019

We all want clean water, right? Clean water to drink, to swim in, to fish in, to use for irrigation of crops, to use in industry, or even just to look at. So “why are we making little progress on nonpoint source pollution in water quality?”

This is the essential question posed by Rissman and Carpenter (2015) 1 as they proceed to explain what they see as the obstacles and opportunities in the cycle of scientific research and policy making surrounding water quality. The context of this whole discussion is related to diffuse nutrient pollution from agriculture or municipalities or even forests and natural lands, and the negative effects such nutrients (nitrates, phosphates, etc.) have on surface water quality downstream.

1 Rissman, A.R., and S.R. Carpenter. 2015. Progress on Nonpoint Pollution: Barriers & Opportunities. Daedalus 144(3): 35-47. doi: 10.1162/DAED_a_00340.

Improvements in water quality, through decades now of research and regulations, have not manifested according to earlier expectations. The complexity of the natural systems involved already overwhelm the ability of scientists to pull apart and make sense of what is causing what. Consider the complex interactions of soil with climate with topography with rivers and lakes with biological process with people and development all with the diffuse nature of the pollution with legislation and political pressures with historical and legacy effects…and on and on. High levels of complexity, of course, increase the time it takes to get research done and to make policies and decisions. So while research and policy making is trying to press forward, time itself is working to increase the complexity of the water quality systems in that climate and land use and other factors might change over the same time frame.

Achieving higher standards of water quality is complex. What factors specifically challenge the capacity of scientific research to be useful in predicting and assessing the management of nonpoint source pollution and policy effectiveness? Time lags and legacy effects are obstacles. A time lag would manifest itself, for example, in the time it takes for a new management practice to be adopted by land managers, or in the time it takes an institution to update and revise policies to reflect new data. A legacy effect might manifest, for example, as a lake continues to show poor water quality characteristics despite a quantifiably lower pollution input. This was the case for Lake Mendota local to the authors in Wisconsin. Working for decades to overcome legacy effects does not necessarily line up well with the shorter term goals and budgets of the people involved in the effort.

Another major factor that limits the success of science to predict or assess management and policy effectiveness is the spatial heterogeneity within and between watersheds. Policies that worked in one watershed with certain soil types and land uses and so on may not prove as effective in another. Even within a watershed, heterogeneity can make it difficult to find causality. The reality of large, heterogeneous watersheds is somewhat at odds with the small research studies done in small, homogeneous areas over short time frames. The research is needed, but extrapolating to larger and more complex watershed just increases the uncertainty of decisions being made.

As mentioned previously, simultaneous changes in multiple drivers has been another reason limits to nonpoint source pollution is complex and slow. Examples of drivers that have changed over the past 30-40 years include: changes in land use (e.g. conversion of agricultural fields to residences), increases in impervious surfaces, climate shifts (e.g. increasingly frequent large precipitation events), and changes to agricultural practices (e.g. conservation tillage). That last one is a good example of how reducing tillage on farm lands brings trade-offs. The reduction in tillage has reduced sediment erosion, thus improving water quality. Simultaneously, the reduction in tillage has, from the standpoint of soil phosphorus, increased the concentration of P at the soil surface. This in turn offsets the benefits of conservation tillage to some degree, as phosphates from the vulnerable soil surface continue to degrade surface waters 2.

2 Duncan, E.W., D.L. Osmond, A.L. Shober, L. Starr, P. Tomlinson, et al. 2019. Phosphorus and Soil Health Management Practices. Agricultural & Environmental Letters 4(1). doi: 10.2134/ael2019.04.0014.

Snowmelt in Wallsburg, Utah

Snowmelt in Wallsburg, Utah

No discussion of scientific shortcomings would be complete without the following reason for slow progress in effective policy-making and practice: lack of monitoring. I’ve heard it said:

“When performance is measured, performance improves. When performance is measured and reported, the rate of improvement accelerates.”

The concept of measuring (and ideally measuring performance) is embedded in the essence of science, I think. But when measuring gets complicated or the thing measured gets complex or some other factor limits some aspect of the ability of the scientist to measure, uncertainty and risks in the policy-making increase.

If the goal to monitor water quality in a given watershed only comes as a result of observations of diminished water quality downstream, then at the time monitoring begins, a lack of historic data will muddle the picture of progress. Historic data can be used to create a baseline against which all new data can be compared. Another significant challenge to monitoring is the lack of a reference watershed, a nearby watershed of similar characteristics that can be paired with the targeted watershed. The targeted watershed undergoes some changes due to policy and management, and the resulting water quality in that target watershed can be compared to the reference watershed which had no changes in policy or management. Any improvements in water quality can then be attributed to the changes, and not to random chance or some external factor (e.g. less rain).

In my thesis research, the target watershed of study had no reference watershed. We observed some temporal changes in water quality, but could not confirm how that stacked up against water in other nearby watersheds. At least the outflow point had mostly regular, monthly water quality measurements going back a couple decades. But except for the outflow and a couple other stream locations, most other sampling locations lacked historic data. This increased the complexity of data interpretation.

Establishing paired watersheds is a suggested way by the authors to reduce uncertainty in assessing policy effectiveness. In this same vein, there must also be a long-term commitment to water quality monitoring. Why? Again, multiple drivers of water quality can change simultaneously over time. Additionally, long-term monitoring provides new inputs to models used, which in turn affect the models and future decisions. Long-term monitoring also reduces data gaps, which preserves the meaningful interpretation of temporal data.

On the topic of models, the authors clearly point to the strengths and uses of models in water quality research. Models are used for: estimating pollution sources, predicting the efficacy of management changes, prioritizing locations, and determining compliance. Models are not “truth machines”, but are pretty good at averages and generalizations. So in looking at a model, a land manager like a farmer may have reasons to believe the model does not apply to their farm. They may think their farm is not average. If the farmer has good enough reasons for this, then a policy including his land may have to be flexible or adaptable to his farms unique needs.

Yet from the realm of social science we know that people more often tend to view their own situation as not-average, and also tend to underestimate the risks and challenges the face. In this light it becomes more clear why some stakeholders in a watershed may be reluctant to trust or embrace the model put forward by the science team. So it must be understood, the authors point out, that models are part of the long-term, incremental, and iterative effort of science and policy to implement, assess, and update decision made to water quality. While models are never perfect nor completely useful, they are better than guessing or hoping. Modeling struggles, too, to fill in the gaps left by qualitative and hard-to-define characteristics in which policy makers might be interested. Models should be updated periodically with new data. Using old water quality models that fit less and less well the current state of a watershed is a potential source of mistrust and reluctance from watershed stakeholders.

Has the list of obstacles to improving water quality been exhausted yet? Of course not! Institutional bias, for one, suggest that the vested interests of an institution will sway their interpretation of results. Data that does not align with pre-existing mission goals of an organization may not be given the same weight or attention. Progress in research and policy outcomes tends to be an important factor in an organization securing future funding, which in turn secures jobs. The sharing of data, or rather the lack thereof, may limit progress as those with unfavorable data may not be inclined to share it for a concern that such data might be used to target certain individuals or groups. This is certainly the case in agriculture. Farmers produce valuable resources (food!) but an externality of pollution is its byproduct.

In conclusion, the ultimate goal in water quality science and policy should really be to make better decisions, the authors claim. A performance-based approach has strengths over a behavior-based approach as is commonly pursued. In short, to improve water quality and to make progress on nonpoint pollution, we ought to:

1. engage a broad, cross-section of society to make sure that scientific pursuits align with public values, thus ensuring, too, that results will have meaningful interpretation

2. use the best science and models, with historic data and reference watersheds for comparisons

3. create and maintain realistic expectations about response, variability, and uncertainty so as to keep all involved grounded throughout the long haul.