Prettiest pictures of 2014

The next person that says big data, puts a fiver in the "most overused terms in meetings in the year 2014" jar. I am excited about the opportunities of ever larger micro datasets, but even more thrilled by how much thought is going into the visualization of these datasets. One of my favorite macho nerd blogs Gizmodo just put up a number of the 2014 best data visualizations. If you also think that these are of so purdy, come take Sol's class at GSPP where he will teach you how to make graphs worthy of this brave new word.

Source: Gizmodo. 

Predictions, paradigms, and paradoxes

Failure can be good. I don’t mean in the “learn from your mistakes” kind of way. Or in the “fail year after year to even put up a good fight in the Big Game” kind of way. But failure is often the sidekick of innovation and risk-taking. In places with lots of innovation, like Silicon Valley, failure doesn’t have the stigma it has in other places. Because people understand that being new and different is the best way to innovate, but also the best way to fail.

Another area that failure can be really useful is in making predictions. Not that bad predictions are useful in themselves, but they can be useful if they are bad in different ways than other predictions. Then averaging predictions together can result in something quite good. The same way that a chorus can sound good even if each person is singing off key.

We have a paper out today that provides another example of this, in the context of predicting wheat yield responses to warming. By “we” I mean a group of authors that contributed 30 different models (29 “process-based” models and 1 statistical model by yours truly), led by Senthold Asseng at U Florida. I think a few of the results are pretty interesting. For starters, the study used previously unpublished data that includes experiments with artificial heating, so the models had to predict not only average yields but responses to warming when everything else was held constant. Also, the study was designed in phases so that models were first asked to make predictions knowing only the sow date and weather data for each experiment. Then they were allowed to calibrate to phenology (flowering and maturity dates), then to some of the yield data. Not only was the multi-model median better than any individual model, but it didn’t really improve much as the models were allowed to calibrate more closely (although the individual model predictions themselves improved). This suggests that using lots of models makes it much less important to have each model try as hard as they can to be right.

A couple of other aspects were particularly interesting to me. One was how well the one statistical model did relative to the others (but you’ll have to dig in the supplement to see that). Another was that, when the multi-model median was used to look at responses to both past and projected warming at 30 sites around the world, 20 of the 30 sites showed negative impacts of past warming (for 1981-2010, see figure below). That agrees well with our high level statement in the IPCC report that negative impacts of warming have been more common than positive ones. (I actually pushed to add this analysis after I got grilled at the plenary meeting on our statement being based on too few data points. I guess this an example of policy-makers feeding back to science).

Add this recent paper to a bunch of others showing how multi-model medians perform well (like here, here, and here), and I think the paradigm for prediction in the cropping world has shifted to using at least 5 models, probably more. So what’s the paradox part? From my perspective, the main one is that there’s not a whole lot of incentive for modeling groups to participate in these projects. Of course there’s the benefit of access to new datasets, and insights into processes that can be had by comparing models. But they take a lot of time, groups generally are not receiving funds to participate, there is not much intellectual innovation to be found in running a bunch of simulations and handing them off, and the resulting publications have so many authors that most of them get very little credit. In short, I was happy to participate, especially since I had a wheat model ready from a previous paper, but it was a non-trivial amount of work and I don’t think I could advise one of my students to get heavily involved.

So here’s the situation: making accurate predictions is one of the loftiest goals of science, but the incentives to pursue the best way to make predictions (multi-model ensembles) are slim in most fields. The main areas I know of with long-standing examples of predictions using large ensembles are weather (including seasonal forecasts) or political elections. In both cases the individual models are run by agencies or polling groups with specific tasks, not scientists trying to push new research frontiers. In the long-term, I’d guess the benefit of better crop predictions on seasonal to multi-decadal time scales would probably be worth the investment by USDA and their counterparts around the world to operationalize multi-model approaches. But relying on the existing incentives for research scientists doesn’t seem like a sustainable model.

Adapting to extreme heat

Since we are nearing the holidays, I figured I should write something a bit more cheerful and encouraging than my standard line on how we are all going to starve.  My coauthor Michael Roberts and I have emphasized for a while the detrimental effect of extreme heat on corn yields and the implications for a warming planet.  When we looked at the sensitivity to extreme heat over time, we found an improvement (i.e., less susceptibility) roughly around the time hybrids were introduced in the 30s,  but that improvement soon vanished again around the 1960s.   Heat is as susceptible to heat now as it was in 1930.  Our study simply allowed the effect of extreme heat to vary smoothly across time, but wasn't tied to a particular event.

David Popp has been working a lot on innovation and he suggested to look at the effect of hybrid corn adaptation on the sensitivity to extreme heat in more detail.  Richard Sutch had a nice article on how hybrid corn was adopted slowly across states, but fairly quickly within each state.  David and I thought we could use the fairly rapid rollout within state but slow rollout across state as source of identification of the role of extreme heat. Here's a new graph of the rollout by state:

The first step was to extended the daily weather data back to 1901 to take a look at the effect of extreme heat on corn yields over time - we wanted a pre-period to rule out that crappy weather data in the early 1900s results in a lot of attenuation bias but get significant results with comparable coefficients when we use data from the first three decades of the 20th century.

In a second step we interact the weather variables with the fraction of the planted area that is hybrid corn. We find evidence that the introduction of corn is reducing the sensitivity of hybrid corn from -0.53 to -0.33 in the most flexible specification in column (3b) below, which is an almost 40% reduction. Furthermore, the sensitivity to precipitation fluctuations seems to diminish as well. (Disclaimer: these are new results, so they might change a bit once I get rid of my coding errors).

The table regresses state-level yields in 41 states on weather outcomes.  All regressions include state-fixed effects as well as quadratic time trends. Columns (b) furthermore include year fixed effects to pick up common shocks (e.g., global corn prices). Columns (1a)-(1b) replicate the standard regression equation we have been estimating before, columns (2a)-(2b) allow the effect of extreme heat to change with the fraction of hybrid corn that is planted, while columns (3a)-(3b) allow the effect of all four weather variables to change in the fraction of hybrid corn.

In summary: there is evidence that at least for the time period when hybrid corn were adopted that innovation in crop varieties lead to an improvement in heat tolerance, which would be extremely useful as climate change is increasing the frequency of these harmful temperatures.  On that (slightly more upbeat note): happy holidays.

Feeding 9... er... 11 billion people

Demographers have been telling us for a while that global populations will level off to about 9 billion, and this "9 billion" number has indeed become the conventional wisdom -- so much so that one of your trusty G-FEED bloggers actually teaches a (presumably excellent) class called "Feeding Nine Billion".

With the current global population at just over 7 billion, the belief that population might level off at 9 billion has given some solace to folks worried about the "pile of grain" problem, i.e. the general concern that feeding a bunch of extra mouths around the world might prove difficult.  9 billion people by 2100 implies a much slower population growth rate over the coming century than was observed over the last century, and while a scandalous 800 million people in the world continue to go to bed hungry every night, there has been notable success in reducing the proportion of the world population who don't have enough to eat even as populations have skyrocketed.  This success, if you can call it that, has in part to do with the ability of the world's agricultural producers to so far "keep up" with the growing demand for food induced by growing incomes and populations, as evidenced by the general decline in real food prices over the last half century (the large food price spikes in the last 5-7 years notwithstanding).

But a paper last month by Gerland et al in Science (gated version here), straightforwardly titled "World population stabilization unlikely this century", provides some uncomfortable evidence that the magic 9 billion number might be a substantial underestimate of the population we're likely to see by the end of this century.  Turns out that fertility rates have not fallen as fast in Africa as expected:  while the total fertility rate has fallen, the decline has only been about a quarter as fast as what was observed in the 1970s and 80s in Latin America and Asia.  This is apparently due both to slow declines in African families' desired family sizes, as well as a substantial unmet need for contraception. Here's a plot from this paper showing the relatively slow decline in African fertility:

So run the world forward for 85 years taking these slower-than-expected fertility declines into account, and you get population projections much higher than 9 billion.  In fact, the mean estimate in the Gerland et al paper of population in 2100 is 11 billion, with their 95% confidence interval barely scraping 9 billion on the low end and 13 (!) billion on the high end.  In fact, their 95% confidence interval for 2050 barely contains 9 billion. Here's the relevant plot (R users will appreciate the near-unadulterated use of the ggplot defaults):

Figure 1 from Gerland et al 2014, Science

So perhaps David should retitle his class, "11 is the new 9", or, "Feeding 9 billion in the next 20 years", or, "Feeding 11 billion (95% CI, 9 billion to 13 billion)".  In any case, these 2+ billion extra mouths are not entirely welcome news for those worried about the global pile of grain.  These much larger numbers imply that even greater progress needs to be made on improving agricultural yields if we want to (a) keep prices at reasonable levels and (b) not have to massively expand agricultural land use to do it.  Thanks, Gerland et al!  

Estimating the impacts of CO2 Fertilization...

While David is shaking in his boots about Saturday's matchup of the Cal Bears against Stanford (both teams have a shameful 5-5 record so far), I have been spending time perusing NASA's recent explosion of multimedia offerings. The video that caught my attention comes from a recent paper displaying the transport of CO2 across the globe. That got me thinking...

My illustrious co-bloggers have documented extensive evidence that extreme heat is bad for crops. We also know that rainfed crops do not appreciate a lack of or too much rainfall. How do we know this? The G-Feed crowd likes to use econometric methods to estimate dose response functions between yields/output and temperature/precipitation. In order to attach a causal interpretation to the estimated coefficients of these dose response functions, one needs "exogenous" (read roughly random) sources of variation in temperature and rainfall. While we know that the distribution of crops across climate zones is not random, day to day changes in weather can be interpreted as random, if one controls carefully for other confounders. We first made this point in a PNAS paper in 2006 and this has been standard practice subject to a number of well understood caveats.

So we know: Extreme Heat=Bad. Too much or too little water = Bad.

What we do not understand well so far are the impacts of CO2 on crop yields using non experimental data. There are plenty of studies which pump CO2 into open top chambers of fields and measure differences in yields between carbon fertilized and control plots. What we do not have is a good measure of carbon fertilization in a field setting which incorporates farmer behavior. What has prevented me and arguably many others from attacking this problem empirically is the fact that I thought that CO2 mixed roughly uniformly across space and any variation in CO2 is variation over time. This variation is not useful as one cannot empirically separate the impacts of CO2 from other factors that vary over time, such as prices and business cycles.

The video link above makes me want to question that assumption. The model based patterns here show tremendous spatial and temporal variability within a year. This is the type of variation in temperature and precipitation we use to identify their impacts on yield. While I understand that we do not have a great historical dataset of ground level CO2 measurements, I wonder if an interdisciplinary team of rockstars could come up with a meaningful identification strategy to allow us to measure the impacts of CO2 on yields. Not much good will come from global climate change, but we cannot simply measure the bads and ignore the goods. If anyone has any good ideas, I am interested. I got lots of great suggestions on my climate data post, so here's hoping...

The hunger strawman

A few questions are almost guaranteed to come up from an audience whenever I give a public talk, regardless of what I talk about. Probably the most persistent question is something like “Don’t we already produce more than enough food to feed everyone?” or its close relative “Isn’t hunger just a poverty or distribution problem?”

Some students recently pointed me to an op-ed by Mark Bittman in the NY Times called “Don’t ask how to feed the 9 billion” that rehashes this question/argument. It probably caught their attention because I teach a class called “Feeding 9 billion”, and they’re wondering why I’d organize a class around a question they supposedly shouldn’t even be asking. The op-ed has some catchy lines such as “The solution to malnourishment isn’t to produce more food. The solution is to eliminate poverty.” Or “So we should not be asking, ‘How will we feed the world?,’ but ‘How can we help end poverty?’" My first reaction to these kind of statements is usually “Gee, why didn’t anyone think of reducing poverty before -- we should really get some people working on that!” But more seriously, I think it’s really a quite ludicrous and potentially dangerous view for several reasons. Here’s three:

1. To talk about poverty and food production as if they are two separate things is to forget that in most of parts of the world, the poorest people earn their livelihoods in agriculture. Increasing productivity of agriculture is almost always poverty reducing in rural areas. The 2008 World Development Report explains this well. Of course, the poor in urban areas are a different story, but that doesn’t change the critical global link between underperforming agriculture and poverty.
2. Food prices matter, even if they are low enough that many of us barely notice when they change. If you go to a market, you’d of course rather have hundreds of dollars in your pocket than a few bucks. But if you are there with a few bucks, and you’re spending half or more of your income on food, it makes a big difference whether food prices are up or down by, say, 20%. If you could magically eliminate poverty that’d be great, but for a given level of poverty, small changes in prices matter. And if productivity of agriculture slows down, then (all else equal) prices tend to rise.
3. Maybe most importantly, there’s no guarantee that past progress on keeping productivity rising and prices low will continue indefinitely, especially if we lose sight of its importance. There’s a great deal of innovation and hard work that goes into simply maintaining current productivity, much less continuing to improve it. Just because many remain hungry doesn’t mean we should treat past successes as failures, or take past successes for granted. And just because we have the technology and environment to feed 7 billion, it doesn’t mean we have it to feed 9 billion (at least not on the current amount of cropland, with some fraction of land going to bioenergy, etc.).

When Stanford had Andrew Luck, we didn’t go undefeated. The football team still had some weaknesses and ended up losing a couple of games, sometimes because of a key turnover or because we gave up too many points. Nobody in their right mind, though, concluded that “the solution to winning football games isn’t to have a good quarterback, it’s to have a good defense.” That would be the wrong lesson to learn from the Andrew Luck era. In other words, it’s possible for more than one thing to matter at the same time. (Incidentally, this year Stanford football has produced more than enough points to be a great team; they just haven't distributed them evenly across the games.)

Similarly, nobody that I know is actually claiming that the only thing we have to worry about for reducing hunger is increasing crop production. That would be idiotic. So it’s a complete strawman to say that the current strategy to reduce malnourishment is simply to raise yields in agriculture. It’s part of a strategy, and an important part, but not the whole thing.

I’m not sure why this strawman persists. I can think of a few cynical reasons, but I’m not really sure. To paraphrase a joke a student told me the other day: there’s really only one good use for a strawman. To drink, man.

Indian crops and air pollution (Guest Post by Jen Burney)

Greetings, loyal G-FEED Readers,

My colleague (and former postdoctoral mentor!), V. "Ram" Ramanathan and I have a paper out this week in PNAS (here) on the impacts of short-lived climate pollutants (SLCPs) on Indian crop yields over the past 30 years (1980-2010). Anticipating the publication, the G-FEED crew invited me to guest post this week. Thanks, guys! It's a fantastic opportunity to talk about the paper itself as well as what we know more generally about air pollution and agricultural impacts. I'm biased, but I think this topic is going to be increasingly important in the coming years. And it feels fitting to discuss it here, as G-FEEDers have been active contributors to understanding the relationship between air quality and crop yields: this work builds on some papers by Ram and Max (here and here), speaks to current work by Wolfram and co. on ozone impacts in the US, and benefitted from feedback from Marshall and David at various points.

First, some background. When we think about anthropogenic climate change, we've usually got carbon dioxide (CO₂), and maybe methane (CH₄), nitrous oxide (N₂O), and CFCs/HCFCs on our minds. These greenhouse gases are basically homogeneously mixed in the atmosphere, and they have very long atmospheric lifetimes (with the exception of methane, centuries). They're the reason that, even if human emissions ceased instantaneously tomorrow, the earth would still see another degree or so of warming. However, the big boon in climate science in the past decade or so has been our increasing understanding of the role that short-lived species play in the climate system. These compounds are more conventionally considered pollutants, and include things like aerosol particulates, ozone precursor compounds, and short-lived greenhouse gases (e.g., tropospheric [surface] ozone and HFCs). The one thing they have in common is their relatively short atmospheric lifetimes, but -- as their physical and chemical properties vary widely -- their mechanisms of impact on our climate system differ. In some cases we don't yet have a full handle on their impacts: for example, uncertainty in cloud-aerosol interactions represents one of the biggest uncertainties remaining in GCMs. (Take a look at the radiative forcing attribution figure from the IPCC AR5 for fun.)

What we do know is this: black carbon is the second or third largest contributor to present warming (after CO₂ and perhaps methane, depending on how you partition its impact – it is part LLGHG and part SLCP, since it’s also an ozone precursor). And the four main warming SLCPs combined (BC, tropospheric ozone, methane, and HFCs) have contributed around half of present warming (again, it depends on how you partition methane’s impacts). Moreover, two of these – ozone and BC – have some clear pathways of impact on crops beyond their impacts through temperature and precipitation. Ozone is directly toxic to plants, and BC cuts down on surface radiation, which should negatively impact photosynthesis. And both ozone and BC are very spatially heterogeneous – they are not well-mixed, and so local emissions could very well be expected to have local impacts.

So…given these realities we took the unabashedly empirical approach and threw available data at the problem. We ran a panel regression analysis that included growing season and crop-area averaged temperature, precipitation, and pollutant emissions on the RHS, along with various time controls. Our unit of analysis was major rice- and wheat- producing states in India, and we looked at data from 1980-2010. We find that relative yield loss (RYL) for climate and air pollution (weighted average for India) is 36% for wheat. That is, yields in 2010 would have been 36% higher absent climate and pollution trends over the past 3 decades. Moreover, most of that RYL is due to pollution (90%). (Our estimate for rice RYL is 20% but is not statistically significant.) This suggests a few things: first, cost benefit analyses for air pollution mitigation programs will likely look a lot better in a place like India if agricultural benefits are included. Second, in the near term, cleaning up the air could help offset T- and P-related yield losses.

How should we think about these results in comparison to previous work? First, what we find in terms of climate (T & P) impacts is slightly smaller than the coarser-scale findings for S. Asia by David and Wolfram et al (here). (They find -5-6%, we have -3.5%.) These are probably not statistically different, but one way to think about this is that, absent pollution variables on the RHS, we’d expect the coefficients for T and P to include some of the impacts of SLCPs. The same holds for the time controls. Wolfram and others have made the case that these panel analyses should include quadratic time trends of one form or another to account for an empirical leveling off of yields. Biophysical limits alone make this a totally sensible proposition, but it’s also likely that the “leveling off” trend that has been observed (more on that by David, here) includes the previously unexplored impact of pollution.

Some previous studies have explored the impact of individual pollutants on yields. Max et al looked at black carbon indirectly, by examining the impact of surface radiation on yields (here and here). They see the signature of atmospheric brown clouds on precipitation and thus rice yields, but no direct radiative impact. My best guess is that: (1) their metric of total radiation contains a combination of two effects – a reduction in direct radiation from black carbon and an increase in diffuse radiation from scattering aerosols like sulfates. Plants can often use diffuse radiation more effectively for photosynthesis. (2) More important, rainy-season rice is probably the hardest place to see this signal, as precipitation clears particulates out of the air. We put both BC and sulfur dioxide (the main precursor of sulfate aerosols, which are scattering, or net cooling) in our model to account for these two effects independently, and we see significance for BC’s impact on wheat but not rice, as would be expected based on seasonal concentrations.

On the ozone side of things, a handful of papers have used atmospheric chemistry models to estimate surface ozone concentrations at a given point in time. Those concentrations have then been used in conjunction with exposure-response (E-R) relationships from test plots to estimate yield losses. In theory, if you really do have a good handle on surface concentrations and a valid E-R function, this would be the way to go. The preferred method is AOT40 – accumulated hours of exposure over 40ppbv. But this number is extremely sensitive to crop cultivar, estimated concentrations, when/how long you count exposure, and other management factors. So these kinds of estimates have very large error bars / uncertainties. (See here, here, here, and here.) A few of those studies suggest RYL in 2000 to be 15%-28% for India for ozone though – a comforting similarity in magnitude.

For me, the nice part of this analysis was that we really took a physical approach – we put stuff on the right hand side in forms that made physical sense, and it turned out to hold up to all the (sometimes nonsensical) things reviewers asked us to do. Best of all, the results agree at least in magnitude with some other approaches.

In terms of impact, I hope this paper helps bring agricultural benefits into the conversation about air pollution mitigation. On the methods side, we did a lot of work to think about how to detangle air pollution and climate impacts based on what we can actually measure – an exercise fraught with nonlinearity. I hope some of our work can help guide future efforts to estimate pollution (and pollution policy) impacts. Readers of this blog are well familiar with threshold-y temperature and precipitation effects, but the pollution impacts and mitigation landscape is worse. A real heterogeneous treatment effects nightmare, where everything is co-emitted with other stuff that all acts in different directions with different thresholds. (See FAQ 8.2, Figure 1 from the IPCC AR5 below for a hilarious, if depressing, rendition.) 

I mentioned the competing impacts on radiation from absorbing and scattering aerosols above, but another interesting example is in ozone formation. Ozone depends on both the absolute and relative concentrations of NOx and VOCs. We had evidence (satellite and ground) that we had both types of ozone regimes (NOx-sensitive or NOx-saturated) in our study area. We used the VOC/NOx ratio to account for that; it should probably be standard practice in these kinds of econometric analyses if you’re doing anything involving NOx or ozone.

All that horn-tooting aside, our analysis was limited in a few ways, and I’m excited to push it in some other directions as a result. A few thoughts:

First, we use emissions, and not concentrations, of pollutants. The main reason is that there aren’t any long-run records of pollutant concentrations in India (or most places, for that matter), and we need better satellite-aided algorithms for extrapolating station data to get these kind of reliable exposure maps (the main task consuming my time right now). So we use emissions inventories of aerosols and ozone precursors as proxies for concentrations. Of course, these inventories are also just estimates, and in some cases (e.g., here) have been shown to be wayyy off (particularly low for black carbon). So I’m looking forward to using better exposure proxies. Ideally, one would look at both emissions estimates (the policy-relevant variable), and concentrations (the physically-relevant variable) together.

Second, we’re also just statistically limited. Emissions of all pollutants have been going up fairly monotonically, and there’s not a ton of signal there for estimation. Going to smaller scales doesn’t make sense from a physical perspective (then you have to worry about transport). So the best thing to do would be to run this same kind of analysis in different countries. I’m particularly excited to look at different pollution regimes – think biomass dominated versus coal dominated, etc. Hopefully I can convince some G-FEEDers that this would be an awesome collaboration idea (ahem, Sol).

I’ll leave off there, though I’m always happy to discuss this stuff if there are any questions. Thanks to G-FEED for having me, and to all of you for reading this tome.

Don’t forget to vote!

- Jen

Guest posting on G-FEED

We're going to try something new out on G-FEED, which is to invite colleagues for guest posts when they have a new paper that is relevant to the topics we cover. Not only will this help to obscure how infrequently we manage to post, but it will provide some fresh perspectives. And hopefully it's a good chance for people to explain their work in their own words, without having to make a commitment to long-term posting. There is a lot of great work out there, and to paraphrase George Costanza, if you take everything the community has ever done in our entire lives and condense it down into one blog, it looks decent!

So without further ado, first up is Jen Burney, who hails from UCSD and has a new paper in PNAS on ozone and crop yields in India...

Fanning the flames, unnecessarily

This post is a convolution of David's post earlier today and Sol's from a few days ago.  Our debate with Buhaug et al that Sol blogged about has dragged on for a while now, and engendered a range of press coverage, the most recent by a news reporter at Science.  "A debate among scientists over climate change and conflict has turned ugly", the article begins, and goes on to catalog the complaints on either side while doing little to engage with the content.

Perhaps it should be no surprise that press outlets prefer to highlight the mudslinging, but this sort of coverage is not really helpful.  And what was lost in this particular coverage were the many things I think we've actually learned in the protracted debate with Buhaug.

We've been having an ongoing email dialog with ur-blogger and statistician Andrew Gelman, who often takes it upon himself to clarify or adjudicate these sorts of public statistical debates, which is a real public service.  Gelman writes in an email:

In short, one might say that you and Buhang are disagreeing on who has the burden of proof.  From your perspective, you did a reasonable analysis which holds up under reasonable perturbations and you feel it should stand, unless a critic can show that any proposed alternative data inclusion or data analytic choices make a real difference.  From their perspective, you did an analysis with a lot of questionable choices and it’s not worth taking your analysis seriously until all these specifics are resolved.

I'm sympathetic with this summary, and am actually quite sympathetic to Buhaug and colleagues' concern about our variable selection in our original Science article.  Researchers have a lot of choice over how and where to focus their analysis, which is a particular issue in our meta-analysis since there are multiple climate variables to choose from and multiple ways to operationalize them.  Therefore it could be that our original effort to bring each researcher's "preferred" specification into our meta-analysis might have doubly amplified any publication bias -- with researchers of the individual studies we reviewed emphasizing the few significant results, and Sol, Ted, and I picking the most significant one out of those.  Or perhaps the other researchers are not to blame and the problem could have just been with Sol, Ted, and my choices about what to focus on.

How journalism works (or doesn't)

One reason we started this blog was the frustration of being misrepresented in media coverage of topics we work on. It can be hard for people to grasp just how frustrating it can be. You spend time talking to a journalist until they seem to get what you're saying, they go off and write the story, and then only about half the time do they check back to see if the quotes they attribute to you are right.  Having words put in your mouth is often compounded by other issues, like a "he said, she said" tone that can make issues appear much more contentious than they should be (see Sol's last post)

Another case in point - the other day I took a call from a reporter for the Guardian who said she was working on a story about which crops are threatened by climate change. I thought I was pretty clear that when we talk about impacts we are never talking about complete eradication of the crop. But today I see their story is "8 foods you're about to lose due to climate change"!

When she asks about CO2 I say it's absolutely clear that it has a benefit, it's just a question of whether it's enough to counteract the bad stuff that happens with climate change. That turned into:

One major issue is carbon dioxide, or CO2. Plants use the gas to fuel photosynthesis, a fact that has led some analysts to argue that an increase CO2 is a good thing for farming. Lobell disagrees, noting that CO2 is only one of many factors in agriculture. “There’s a point at which adding more and more CO2 doesn’t help,” he says. Other factors – like the availability of water, the increasing occurrence of high and low temperature swings and the impact of stress on plant health – may outweigh the benefits of a CO2 boost.

What happens over time is you learn to be a little more aggressive with reporters, but that only helps so much. And also you learn to stop answering your phone so much, and to stick with the handful of reporters you think do a really good job. It's sad but true.

What's especially annoying, though, is when people see these stories and start attributing everything it says to you, as if you wrote it, picked the headline, etc. (I see some tweets today saying I'm trying to spread fear about climate change.) The irony is when I give talks or speak on panels I'm more often than not accused of being a techno-optimistic, both about climate change and food security in general. I actually am quite optimistic. About food. Just not about journalism.