This blog post discusses the seminar given by Bob Hirsch of the US Geological Survey (USGS), at the IAN Seminar Series on November 23, 2010.

The analytical method used by Bob Hirsch was clarified. Bob explained that he used nutrient concentrations that were estimated daily every day over a 31 year period, and these concentrations were converted to fluxes by multiplying nutrient concentrations by discharge (water flow) rates. In the example of the Patuxent River, the trends in total phosphorus fluxes declined from the period 1978 to 1990; following 1990, the total phosphorus fluxes did not decline, rather remained steady. High total phosphorus concentrations did occur post-1990, as a result of high discharge years. The Hirsch method served to smooth out sudden changes and the time period integration is affected by the period over which the integration is conducted.

There were several other trends detectable in the data, including an increased minimum phosphorus in the Rappahannock River and an upward trend in nitrate in the Choptank River. In the Choptank, it is the leaching of groundwater nitrate into the river that is likely the reason for increasing nitrate. The Patuxent River watershed is being converted from farming into suburban development, which could result in higher total phosphorus runoff. Bob Hirsch drew the analogy of Lake Erie, where point source phosphorus loadings decreased due to upgrades in sewage treatment historically, but more recent trends are for a re-eutrophication. This recent trend could be attributed to the prevalence of no till agriculture in which the farms become more efficient at retaining water and sediment on the land. The application of manure or fertilizer on land leads to a build up of phosphorus levels in the soil. The fields in Ohio tend to be tiled, providing efficient runoff, and dissolved phosphorus levels increase, while total phosphorus increases due to soil retention. Bob Hirsch talked about the trade-offs in nutrient management between high flow and low flow runoff. Managing for high flow could compromise low flow nutrient management, and vice versa.

The question of whether or not it was valid to analyze short term trends (e.g., 2-3 year milestones) using this analytical approach was raised. Bob responded that his method was not useful over short time periods. Five to ten years of data were needed, and while the 'tweaks' at the end of the graphs are interesting, they are not meaningful. Bob also pointed out that long term problems need to have long term trend perspectives. In order to have shorter term data interpretation, we would need to have daily flows and concentrations measured. Monitoring technology improvements, particularly by the US Geological Survey, for daily nitrate monitoring could allow this analysis in the future. Unfortunately, the technology for continuous phosphorus monitoring is still lacking. The development of sensitive analytical techniques targeting short term trends is needed.

Bob Hirsch discussed the importance of retaining data at the finest temporal resolution possible. For example, if fifteen-minute data are available, then rather than integrating into twenty four hour intervals, thus losing approximately a hundred times the resolution, preserving the data in fifteen minute intervals should occur. Bob also indicated that the smaller the watershed, the shorter the time interval needed.

Bob noted that we increasingly need an analytical tool box that goes beyond standard statistical analyses. For example, the use of quadratic equations are problematic (e.g., banana plots) due to the mathematical tendency for the data to drive an inflection into the data. We need to tailor the data handling approaches to new data integration questions. As fertilizer applications change and sewage treatment technologies change, there will be time lags which will need to be analyzed. Bob pointed out that analyses of Mississippi River fertilizer applications led to a 7-10 year lag in the receiving water quality.

The question of whether this methodology could be adapted to tidal reaches was posed. Bob pointed out that what would be needed would be accurate estimates of residence time. The method did not work when applied to Lake Champlain phosphorus levels, with data spikes occurring in both high flow and low flow years. Non linear biogeochemical trends in estuaries could confound the analyses, for example when sediment oxidation levels leads to an 'iron curtain' which traps and binds the dissolved phosphorus vs. sediment reducing conditions in which the 'iron curtain' is removed and dissolved phosphorus is rapidly released into the water column.

The issue of climate change confounding the data interpretation was discussed. Bob explained that the flow normalization assumes that July 4 1970 flow relationships were the same as July 4 2010 flow relationships. The 'stationarity' of flow is implicit, and the construction or removal of dams would influence the flow relationship. The climate scenario of warmer, wetter winters in the mid-Atlantic could influence the importance that snow melt contributes to flow, but the small changes due to climate change will be overwhelmed by the changes brought on by development. Bob pointed out that an interesting analysis would be that of suspended solids trends as a result of the dams being filled with sediments. A recent paper comparing minimum and maximum flows in which wet climates led to a reduction of maximum flows. This non-intuitive data result did not correspond with climate predictions. Bob pointed out that there are several orders of magnitude differences in these temporal analyses of flow and concentrations.

The question of whether or not increased 'flashiness' of flows due to increased impervious surface area and/or climate change would affect the analysis was raised. For example, would the increased impervious surfaces represented by Washington, D.C. affect the entire downstream Potomac River? Bob pointed out that the scale needs to be considered, so if river flow is affected, then indeed the trend would affected. On a related note, the flood plain connectivity with the river as it could affect this trend analysis was questioned. Bob remarked that it remains to be seen as to whether this would affect trends.

For future analyses, Bob indicated that suspended solids analysis would be 'messy', due to variability, but scientifically important and interesting. The issue of data detection limits needs to be dealt with, due to some data reported as 'less than' values. Changes in analytical thresholds with technological improvements over time need attention.