Monthly Archives: November 2016

Technology Can Lead Us Astray

Our iPhones teach us very subtly to have great faith in technology. This leads the public at large to think that technology will solve large issues like greenhouse gases and climate change. But for scientists we should remember that technology must be looked at very carefully when it tells us we have a shortcut to ecological measurement and understanding. For the past 35 years satellite data has been available to calculate an index of greening for vegetation from large landscapes. The available index is called NDVI, normalized difference vegetation index, and is calculated as a ratio of near infrared light to red light reflected from the vegetation being surveyed. I am suspicious that NDVI measurements tell ecologists anything that is useful for the understanding of vegetation dynamics and ecosystem stability. Probably this is because I am focused on local scale events and landscapes of hundreds of km2 and in particular what is happening in the forest understory. The key to one’s evaluation of these satellite technologies most certainly lies in the questions under investigation.

A whole array of different satellites have been used to measure NDVI and since the more recent satellites have different precision and slightly different physical characteristics, there is some problem of comparing results from different satellites in different years if one wishes to study long-term trends (Guay et al. 2014). It is assumed that NDVI measurements can be translated into aboveground net primary production and can be used to start to answer ecological questions about seasonal and annual changes in primary production and to address general issues about the impact of rising CO2 levels on ecosystems.

All inferences about changes in primary production on a broad scale hinge on the reliability of NDVI as an accurate measure of net primary production. Much has been written about the use of NDVI measures and the need for ground truthing. Community ecologists may be concerned about specific components of the vegetation rather than an overall green index, and the question arises whether NDVI measures in a forest community are able to capture changes in both the trees and the understory, or for that matter in the ground vegetation. For overall carbon capture estimates, a greenness index may be accurate enough, but if one wishes to determine whether deciduous trees are replacing evergreen trees, NDVI may not be very useful.

How can we best validate satellite based estimates of primary productivity? To do this on a landscape scale we need to have large areas with ground truthing. Field crops are one potential source of such data. Kang et al. (2016) used crops to quantify the relationship between remotely sensed leaf-area index and other satellite measures such as NDVI. The relationships are clear in a broad sense but highly variable in particular, so that the ability to predict crop yields from satellite data at local levels is subject to considerable error. Johnson (2016, Fig. 6, p. 75) found the same problem with crops such as barley and cotton (see sample data set below). So there is good news and bad news from these kinds of analyses. The good news is that we can have extensive global coverage of trends in vegetation parameters and crop production, but the bad news is that at the local level this information may not be helpful for studies that require high precision for example in local values of net primary production. Simply to assume that satellite measures are accurate measures of ecological variables like net aboveground primary production is too optimistic at present, and work continues on possible improvements.

Many of the critical questions about community changes associated with climate change cannot in my opinion be answered by remote sensing unless there is a much higher correlation of ground-based research that is concurrent with satellite imagery. We must look critically at the available data. Blanco et al. (2016) for example compared NDVI estimates from MODIS satellite data with primary production monitored on the ground in harvested plots in western Argentina. The regression between NDVI and estimated primary production had R2 values of 0.35 for the overall annual values and 0.54 for the data restricted to the peak of annual growth. Whether this is a satisfactory statistical association is up to plant ecologists to decide. I think it is not, and the substitution of p values for the utility of such relationships is poor ecology. Many more of these kind of studies need to be carried out.

The advent of using drones for very detailed spectral data on local study areas will open new opportunities to derive estimates of primary production. For the present I think we should be aware that NDVI and its associated measures of ‘greenness’ from satellites may not be a very reliable measure for local or landscape values of net primary production. Perhaps it is time to move back to the field and away from the computer to find out what is happening to global plant growth.

Blanco, L.J., Paruelo, J.M., Oesterheld, M., and Biurrun, F.N. 2016. Spatial and temporal patterns of herbaceous primary production in semi-arid shrublands: a remote sensing approach. Journal of Vegetation Science 27(4): 716-727. doi: 10.1111/jvs.12398.

Guay, K.C., Beck, P.S.A., Berner, L.T., Goetz, S.J., Baccini, A., and Buermann, W. 2014. Vegetation productivity patterns at high northern latitudes: a multi-sensor satellite data assessment. Global Change Biology 20(10): 3147-3158. doi: 10.1111/gcb.12647.

Johnson, D.M. 2016. A comprehensive assessment of the correlations between field crop yields and commonly used MODIS products. International Journal of Applied Earth Observation and Geoinformation 52(1): 65-81. doi: 10.1016/j.jag.2016.05.010.

Kang, Y., Ozdogan, M., Zipper, S.C., Roman, M.O., and Walker, J. 2016. How universal Is the relationship between remotely sensed vegetation Indices and crop leaf area Index? A global assessment. Remote Sensing 2016 8(7): 597 (591-529). doi: 10.3390/rs8070597.

Cotton yield vs NDVI Index

University Conundrums

Universities in Canada and the United States and probably in Australia as well are bedeviled by not knowing what they should be doing. In general, they all want to be ‘excellent’ but this is largely an advertising gimmick unless one wishes to be more specific about excellent in what? Excellent in French literature? Probably not. Excellent in the engineering that facilitates the military-industrial complex? Probably yes, but with little thought of the consequences for universities or for Planet Earth (Smart 2016). Excellence in medicine? Certainly, yes. But much of the advertisement about excellence is self aggrandisement, and one can only hope that underneath the adverts there is some good planning and thinking of what a university should be (Lanahan et al. 2016).

There are serious problems in the world today and the question is what should the universities be doing about these long-term, difficult problems. There are two polar views on this question. At one extreme, universities can say it is our mandate to educate students and not our mandate to solve environmental or social problems. At the other extreme, universities can devote their resources to solving problems, and thereby educate students in problem analysis and problem solving. But these universities will not be very popular since for any serious issue like climate change, many voters are at odds over what can and should be done, Governments do not like universities that produce scholarship that challenges their policies. So we must always remember the golden rule – “she that has the gold, makes the rules”.

But there are constraints no matter what policies a university adopts, and there is an extensive literature on these constraints. I want to focus on one overarching constraint for biodiversity research in universities – graduate students have a very short time to complete their degrees. Given a 2-year or 3-year time horizon, the students must focus on a short-term issue with a very narrow focus. This is good for the students and cannot be changed. But it is potentially lethal for ecological studies that are long-term and do not fit into the demands of thesis writing. A basic assumption I make is that the most important ecological issues of our day are long-term problems, at least in the 20-year time frame and more likely in the 50 to 100-year time frame. The solution most prevalent in the ecology literature now is to use short term data to produce a model to extrapolate short term data into the indefinite future by use of a climate model or any other model that will allow extrapolation. The result of this conundrum is that the literature is full of studies making claims about ecological processes that are based on completely inadequate time frames (Morrison 2012). If this is correct, at least we ought to have the humility to point out the potential errors of extrapolation into the future. We make a joke about this situation in our comical advice to graduate students: “If you get an exciting result from your thesis research in year 1, stop and do no more work and write your thesis lest you get a different result if you continue in year 2.”

The best solution for graduate students is to work within a long-term project, so that your 2-3 years of work can build on past progress. But long-term projects are difficult to carry forward in universities now because research money is in short supply (Rivero and Villasante 2016). University faculty can piggy-back on to government studies that are well funded and long-term, but again this is not always possible. Conservation ecology is not often well funded by governments either, so we keep passing the buck. Collaboration here between governments and universities is essential, but is not always strong at the level of individual projects. Some long-term ecological studies are led by federal and regional government research departments directly, but more seem to be led by university faculty. And the limiting resource is typically money. There are a set of long-term problems in ecology that are ignored by governments for ideological reasons. Some politicians work hard to avoid the many ecological problems that are ‘hot potatoes’ and are best left unstudied. Any competent ecologist can list for you 5 or more long-term issues in conservation biology that are not being addressed now for lack of money. I doubt that ideas are the limiting resource in ecology, as compared with funding.

And this leads us back in a circle to the universities quest for ‘excellence’. Much here depends on the wisdom of the university’s leaders and the controls on university funding provided by governments for research. In Canada for example, funding constraints for research excellence exist based on university size (Murray et al. 2016). How then can we link the universities’ quest for excellence to the provision of adequate funding for long-term ecological issues? As one recommendation to the directors of funding programs within the universities, I suggest listing the major problems of your area and of the world at large, and then fund the research within your jurisdiction by how well the proposed research matches the major problems we face today.

Lanahan, L., Graddy-Reed, A. & Feldman, M.P. (2016) The Domino Effects of Federal Research Funding. PLoS ONE, 11, e0157325. doi: 10.1371/journal.pone.0157325

Morrison, M.L. (2012) The habitat sampling and analysis paradigm has limited value in animal conservation: A prequel. Journal of Wildlife Management, 76, 438-450. doi: 10.1002/jwmg.333

Murray, D.L., Morris, D., Lavoie, C., Leavitt, P.R. & MacIsaac, H. (2016) Bias in research grant evaluation has dire consequences for small universities. PLoS ONE, 11, e0155876.doi: 10.1371/journal.pone.0155876

Rivero, S. & Villasante, S. (2016) What are the research priorities for marine ecosystem services? Marine policy, 66, 104-113. doi: 10.1016/j.marpol.2016.01.020

Smart, B. (2016) Military-industrial complexities, university research and neoliberal economy. Journal of Sociology, 52, 455-481. doi: 10.1177/1440783316654258