Category Archives: Risk Management

On ENSO and the timing of food shortages

Much has been written about how the El Niño Southern Oscillation (ENSO) affects temperature and precipitation globally (these impacts are often referred to as ENSO teleconnections). In two recent studies (one on climate teleconnections, one on their impacts on crop production), my coauthors and I try to bring attention to the predictable interannual evolution of ENSO, which is often overlooked in discussions of food security although not in the climate science or climate forecasting communities. In particular we look to highlight the fact that ENSO events follow a predictable pattern in which La Niñas (cold ENSO events) only occur following El Niños, and often persist for two years. While not all El Niños develop into La Niñas, the pattern is still noteworthy, particularly when one considers the implication of La Niñas for food security in North and South America.

Here we need to be precise about a few terms. I’m going to talk about three main concepts: 1.ENSO creates global teleconnections, which means that the risks posed by ENSO are correlated in space, 2. ENSO has a characteristic multi-year evolution, which means risks posed by ENSO are correlated in time, and 3.The sign and timing of the anomalies (i.e. do good years follow bad or vice versa) are critical for food security.

So to save you some time (you are after all reading this blog post rather than the paper) I’ll skip to our conclusions. 1.ENSO poses a correlated risk across much of North and South America. This means that plentiful harvests in North America tend to coincide with good harvests in major producing regions of South America (some notable exceptions mentioned in the paper). 2.The characteristic multi-year evolution of ENSO is apparent not only in patterns of temperature and precipitation, but also in yield anomalies. While not all states are influenced by ENSO in the Americas, many of the major production areas for wheat, maize and soybeans are. And in significant portions of these major production areas there is a multi-year progression of yield anomalies attributable to ENSO. 3. El Niños bring favorable growing conditions while La Niñas increase the probability of crop failures. This point is crucial when we reconnect it to the life-cycle of ENSO (i.e. La Niñas only occur following El Niños). In other words, despite the spatial correlation in the risk (remember the global teleconnections), the temporal correlation brings a silver lining: Poor Pan-American production years are likely to occur following years of above expected production. I don’t think I need to hammer home why this is positive from the perspective of food stocks and food security.

So that’s the new analysis. Some reason to worry (largely ignored correlations in the risks posed by ENSO) and some information that may be used to improve food security. There’s obviously much more in the paper, including the importance of soil moisture for each crop, so I encourage you to give them a read!

How (and whether) to Disseminate Climate Forecasts

One topic that I’ve been interested in for a while now, but haven’t yet had the chance to explore in any depth is the way in which we disseminate drought forecasts. In this blog post I’d like to look a little further into how we disseminate, what we disseminate, and whether it makes a difference. The short version is this: we have thought quite a bit about what we provide, but surprisingly little about whether it is effective (cost or otherwise).

Numerous studies have explored the way in which we provide information (see here and edited collections here and here). They have made some real advances in shedding some light on how farmers’ make use of climate forecasts, as well as the estimated impact (i.e. whether farmers changed their responses in the context of a workshop participation).

As it turns out, farmers’ are quite capable of understanding and acting on probabilistic information. For forecasters, this is good news. One question that I would be interested in exploring further is whether the workshops required to train farmers to use forecasts is cost effective. This question relates both to the initial cost, and to the question of information retention. When testing principles in the context of daylong participatory workshops, we are unable to address issues such as usage retention (particularly following forecasts that do not match the eventual seasonal totals).

A related question, raised by a colleague of mine here at IFPRI, is whether we should really be providing the information to individual farmers or if it is more effective to provide the information to regional met agencies. Again, the question is not whether farmers are capable of using the forecasts, but rather whether providing them directly is cost effective in the long run.

The most pressing question, however, is in many ways the most obvious: do climate forecasts improve yields? A rigorous study (read randomized control) of the real-world implications of climate yields is badly needed as a means of addressing whether climate forecasts are effective. Although I understand the desire to provide a high quality product (accurate forecast) in a reliable manner, it is past time that we begin discussing the hard evidence of cost effectiveness.

Much time has been dedicated to studying climate forecasts, but surprisingly little has been invested in understanding what role climate forecasts are likely to play in improving livelihoods. We can’t afford to silo these questions any longer.

Operational Drought Monitoring

To reduce the risk to food security posed by drought, it’s crucial that we develop systems able to disseminate accurate information in a timely manner. Although doing this may seem straightforward in an era of near-ubiquitous satellite measurements and increasingly high-powered computational models, there are challenges relating both to the analysis and dissemination of information in an operational context. I’ll explore the challenges to producing operational drought forecasts and monitors in this post and write about the challenges of disseminating those forecasts in my next post.

Broadly, most operational drought systems may be subdivided into those that monitor current conditions (and perhaps make historical readings available) and those that provide projections for future conditions. Coincidentally, while doing some research for this post I started to compile a list of available hydrological monitors and data portals, which you can find here. I focused on drought, but included a few others as well.

Drought monitoring, despite the availability of near real-time satellite data, faces great challenges of data availability. In fact, methods using satellite measurements currently perform comparably to those using only a handful of gauge stations. Satellite measurements are additionally challenged with a short climatology, and inconsistencies between products. Never-the-less, these products provide unparalleled coverage for regions with sparse in-situ measurements (as is often the case in developing countries), where drought monitoring has a crucial role to play in maintaining adequate levels of food security.

Although it is often described as a “slow onset” phenomenon, the development and evolution of drought can be remarkably difficult to predict. Part of this stems from an incomplete understanding of the oceanic forcings of drought (ENSO), and part of it stems from the inherently chaotic nature of the atmosphere. A chaotic systems sensitive to initial conditions creates an environment in which errors propagate through models and forecasts quickly diverge from one another. Shukla et al., 2013 analyzed the skill of a forecast as it relates to either (1) the initial conditions of the model or (2) the forecast skill, and found that the skill was dependent on both region and time of the year.

One means of improving both the monitoring and the forecasting of drought is to further explore the limitations to current model skill under different climate regimes. While Smith et al., explored times of the year and regions, the underlying factors of substance are the moisture-temperature-atmosphere regimes. It shouldn’t be ignored that during a drought, the region of interest (which may normally be energy-limited) will be abnormally arid (and therefore potentially moisture-limited). This shift may in fact mean that where the initial condition of soil moisture was once not a limiting factor for forecast skill, it may become one during the forecast of drought recovery. In that sense, forecasting the onset as opposed to the recovery of a drought may be two problems with distinct characteristics.

A second aspect of the drought system that warrants further exploration is the dynamics of vegetation during the evolution of multi-seasonal droughts. Previous studies have pointed towards dynamic vegetation as one source of increased interannual variability in precipitation, however, the impact of this dynamic vegetation on evapotranspiration and therefore on the atmospheric boundary layer may also play a significant role in determining how a drought develops. This is particularly true during multi-year droughts when drying of the soil occurs more deeply than during one-season droughts.

On Irrigation, Groundwater and Drought

Recently groundwater has been getting a lot of attention. Particularly, scientists have focused on characterizing our groundwater use over the last few decades, and producing estimates of future water availability. Many aquifers are being depleted at alarming rates, some projected to become depleted as early as the end of the century (Jay Famiglietti has done a lot of good work on this topic, see his articles in Nat Geo here). If we continue along the path we are currently on, the question is not if we will deplete our groundwater resources, but when.

Given this reality, it seems appropriate that we begin to model management scenarios in which groundwater plays either a substantially reduced role, or no role at all in providing water for irrigation. That is no small task when you consider that in the U.S. groundwater provides 61% of total water used for irrigation (Siebert et al., 2010). These scenarios may involve switching to more efficient irrigation technology (drip irrigation as opposed to our current water-intensive practices) or reducing the irrigated area altogether.

Irrigation has a substantial impact on regional climate, and the implications of altering our current irrigation practices need to be studied further. It has been well documented that irrigation provides a regional cooling effect, a property that alleviates heat-stress in crops during droughts and heat waves (see van der Velde, 2010). Less well understood is whether switching to drip irrigation will increase water efficiency at the expense of this regional cooling effect. It is worth exploring whether during heat waves plants are in greater need of the cooling effect or the water provided by irrigation.

An alternative management scenario – allowing portions of fields to lay fallow due to insufficient water for irrigation – may have a time-dependent component, particularly when considering the time required for transitional vegetation to take root. If multi-year droughts lead farmers to leave large percentages of their fields fallow all-at-once, the vegetative landscape is likely to look very different than if the transition is gradual. Vegetation cover has an effect on atmospheric dust loading (think dust bowl) and on the latency with which the landscape reacts to drought (see work by Ning Zeng on this subject).

So, back to the point: how do we dynamically incorporate potential management decisions into climate projections? We need to conduct analyses that provide an envelope of uncertainty around management decisions, and a few scenario analyses in which different management options are explored in detail. Part of understanding the future of the hydroclimate is understanding how we are playing a dynamic role in that system and to what degree our actions will impact our hydrologic future.

Information dissemination, financial systems and resiliency

The idea of resiliency in the face of climate change has been a popular idea lately, but a “resilient system” is difficult to define and a slippery concept at best.  Resiliency does not describe any one mechanism, nor does it describe a policy. Rather, resiliency describes the interaction between components of a system under stress. As a hydroclimate scientist, the idea of “resiliency” is particularly of interest to me in the context of agriculture.

New Security Beat is running a series of articles on resilient agriculture in the face of climate change. The most recent article begins with a point that is not often emphasized: resiliency is natural. As the author notes, “any effort to build resilience must begin with a deep understanding of existing strengths and adaptive mechanisms and make every effort to keep them intact”. Agricultural systems are a complex weave of formal and informal relationships (both financial and otherwise), which all play a part in determining the resiliency of the system. As such, agriculture cannot be approached from a single scale or a single perspective. The remainder of this post focuses primarily on financial and information-driven initiatives, but these are by no means the only perspectives at play in such systems.

From a financial perspective, we have been fairly good about discussing mechanisms that function across scales. On the micro-level there has been extensive study of micro-loan structures both formally through development banks and informally through community organizations. On a macro scale there has recently been interesting developments in the successful implementation of parametric weather-linked crop insurance. The IRI recently partnered with Oxfam and Swiss Re (among others) in a fascinating pilot study of weather-linked crop insurance for Northern Ethiopia, which I would encourage you to read. These two financial mechanisms, although applied in different manners, are complementary means to the same end.

From an information dissemination perspective we have had a somewhat more lopsided approach to providing support for the development of resilient systems. Many recent developments have pioneered exciting top-down information dissemination programs by partnering with local meteorological offices to issue growing season forecasts to farmers (please do watch the short video in the link, it is a great example of why such programs are invaluable). Despite these recent advances, we have too often overlooked the potential added value of low-tech community-scale information systems. Supplementing regional forecasts with local information is not a new concept, but it seems to be a discussion that is too-often missing from the academic literature. As a scientific community we need to consider not only how we can provide valuable forecasts, but also how those forecasts will interface with existing community-scale information systems.

Finally, looming in the substratum of any discussion involving climate change is the notion that no system is static, and no future certain. Although these are topics for another post, I’ll briefly point to an interesting article, which describes an increasing number of pastoralists in East Africa taking up farming due to the insecurities of a pastoral life in times of drought. If this trend continues, the dynamics of the East African agricultural system will change significantly and projections of growing season precipitation will become that much more valuable.