Crop models and their associated errors

In any climate change research there is a certain aspect of uncertainty associated with the climate models used, this uncertainty is often much larger than the uncertainty associated with the crop model used (Mereu et al., 2015). Having said that Lobell suggests that ignoring errors, when adopting a crop model, can distort crop sensitivity to rainfall by a factor of 2 or more.

Watson and Challinor (2012) undertook a research paper in which they manipulate the source of error in a General Large Area Model (GLAM) to measure the influence of RMSE of Groundnut yield. The comparison of error associated with climate data (including mean temperature and rainfall initially inputted into the model) with crop yield inputs for calibration purposes to establish which created the largest Root Mean Square Error. Rainfall posed to be the have the most significant influence on the skill of the model. This is quite expected as rainfall is often the limiting factor within semi-arid regions. In particular misrepresentation of inter seasonal variation of rainfall (and temperature) creates more error than that generates through systematic bias of climate simulations. However due to the nature of climate it is important to remember the generalisation of these results from the study site, Gujarat India needs high consideration.

RMSE was increased by 143% upon the manipulation of yield data, because of this the paper concludes both potential and actual crop yield measurements should be incorporated as decreasing this gap between potential and actual yields are ever the focus of agriculture practices. Without this observed data (used within the calibration process) it is impossible to estimate the total error associated with simulations produced by the model and have confidence in results.

Figure 1: demonstrates the RMSE of each of the parameters under scenario A in the study. The ranges shows the greatest error is associated with incorrect precipitation data.

Therefore it is essential that these errors are incorporated as the sensitivity of crops to climatic factors can have implications for the overall assessment of climate change on food security and consequently the adaption strategies implemented (Challinor 2009). 

In the face of Climate Change: Part 4

Carrying on from last week’s blog, in which I discussed the applicability of models to realistic scenarios, I have begun to explore the utilization of crop modelling in policy, in this case for climate change adaption. Gaun et al (2015) argue the results of their recent study can aid the design of future rain fed agriculture strategies in West Africa.

The paper, “What aspects of future rainfall changes matter for crop yields in West Africa?”, discusses the relationships between variable rainfall patterns (as a result of climate change) and Sorghum crop yields in the region. A stochastic rainfall model (based on the CMIP5 predictions) was adopted to drive 2 crop models (2 models were used to reduce model dependency and ensure confidence in the simulated results).

Figure 1: Crop Yield change for increased (+30%) and decreased (-30%) frequency of rainfall events

The model ran scenarios under varying Mean Annual Precipitation, intensity, and frequency to differentiate the effects of each factor. There is a general academic consensus that, as a result of climate change, rainfall patterns will change. However, uncertainty (associated with gridded climate change predictions) creates a challenge when estimating which aspect of rainfall patterns will have the most detrimental impacts on crop yields (Sorghum in this case). The study’s results indicated that a reduction in Mean Annual Precipitation will have the most widespread influence. In drier areas (Bellow 600mm/year), however, intensity prevails over alterations to frequency (Figure 1). Increased frequency, in rainfall events, showed a reduction in the total volume of water reaching deep soil layers, causing greater evaporation from surface and soil stores and thus reducing the amount of water available for crops and reducing yields. An assumed reduction in runoff, in this scenario, does not outweigh losses through evaporation in the soil moisture, thus the paper concludes increased intensity could be more beneficial for crop yields.

In relation to policy implication and further academic research, this paper provides justification for the focus on seasonal total rainfall in agricultural management, yet calls for attention to be brought to intensity and frequency in regions of low rainfall (ie. West Africa). Furthermore this is especially important in relation to adaption to climate change, by prioritising which aspects of rainfall variability will have the greatest impacts on crop yields and consequently food security (see my previous post on the sustainable development goals for more details). This subject is very topical at present, yet I have started to wonder about the consistency in results from paper to paper… are academics in consensus about the influence of climate change on crop yields? Or is this just another topic with high levels of uncertainty? Stay tuned to find out.

In the fact of climate change: Part 3

This is a fantastic video which summaries the some of the impacts of climate change on crop yields. This is also discusses the implications for food security! Enjoy!

In the face of Climate Change: Part 2

Whilst completing my weekly reading I strolled upon a paper which stopped me in my tracks. A crop model? My favorite!

 Lobell and Burke (2008) ‘On the use of statistical models to predict crop yield response to climate change’. Within the Paper Lobell and Burke discussed the potential effects of climate change on crop yields for 198 sites in Sub Saharan Africa. The paper adopts the hypothesis, that a 2 degree temperature rise and 20% precipitation reduction will be the prominent implications of climate change. As always there will be uncertainty associated with climate change predictions but this argument lies beyond the scope of this blog post. The results, of the study, indicated (based on median emissions predictions) a 2 degree warming would result in a 14.4% loss in crop yields, whereas the effect of reducing precipitation (by 20%) would only produce a 5.8% yield reduction. This seems counter intuitive when considering the importance of rainfall in agricultural production. However temperature trends are large in relation to historical yield variability, thus becoming more prominent than precipitation trends when using the model CERES to analyse crop responses to climate change.

Methods:

The paper adopted the following methodology. A CERES-Maize model was used to simulate historical measurements of maize yield variability. The model was run under the climate change conditions to simulate the impact of a 2 degree temperature increase and 20% reduction of precipitation in maize yield variability. A statistical model was run alongside to provide a comparison of results.

Thoughts and reflections:

In relation to the adopted methods of the study, evidence suggests there are areas which contain increased levels of uncertainty. Firstly results do not consider any form of socio-economic response, all parameters are purely physical. An integrated assessment aims to tie together crop yield and socio- economic factors and can increase relevance (to a specific region) by incorporating issues such as how yield may differ in repose to adaptive measures.(Challinor et al., 2007). Fischer et al. (2002) use an integrated assessment to combine the use of climate change scenarios and resulting price trends in the global crop market.

Secondly, historical measurements of crop yield variability (within the Lobell and Burke study) were simulated within a CERES model (based on climate parameters and soil moisture conditions). In general modelling terms, a historic data set of observed variability (in this case I would expect historically observed crop yields), is initially inputted into the model. Observed data is also used to calibrate models to reduce uncertainty and ensure results are realistic. The use of observed historical data does not seem to be present in this study as either an input or for calibration purposes. This leads to the question, how accurate is this model.

In a future post I aim to explore a series of different crop modelling studies to investigate whether or not a lack of historical observations is common throughout this field of research.

As always, feel free to respond to what I have covered in today’s post by leaving a comment below.  

Climate change: part 1

 Climate change is a huge topic for any geographer (unless you are a human geographer then the use of social spaces may float your metaphorical boat)! The academic scene is blooming with studies analysing the implications of climate change. Papers focusing on the implications on crop yields and food security are of particular use for this blog. My previous post (in which I referred to Gaun’s paper, which is well worth a read, even if I do say so myself!) gave a brief overview of the influences of future rainfall variations on crop yields. In this post I thought I would take a step back and assess the general implications of climate change for rainfall and temperature. This will help me take critical stance when reading such academic papers.

Temperature

Under mean emissions scenarios, the CMIP 5 projections estimate, by the end of the 21^st century (2070-2099) African surface temperatures will have risen by 2-4 degrees Celsius; values which are robust across the majority of GCM models used in Aloysius et al (2015). Increased levels of evapotranspiration can decrease reliability of surface stores of water and increase the risk of water stress.

Rainfall

Predicating future rainfall patterns is much harder than predicting their partner in crime, temperature.  In simple terms the wet is going to get wetter and the dry is going to get drier (The Guardian, 2011) Now, I am a geography student and that explanation doesn’t quite make the cut.

Increased temperatures allow a greater volume of water vapour to be held in the atmosphere (up to 7% per degree of warming) which could result in an approximate rise of mean annual precipitation by 1-2% (per degree of warming) (Feng et al., 2013). In relation to the global distribution of rainfall, it is predicted that higher latitudes will receive more rainfall at the expense of regions in the tropics (IPCC, 2007). Details surrounding changes in frequency and intensity of rainfall patterns, in Africa, can be found in my previous post. Climate change is predicted to alter, not only the magnitude of rainfall, but also its seasonal distribution and variability. Arid and Semi-arid regions, such as Africa, will be hit the hardest these changes as they rely on seasonal regimes of rainfall for agricultural production.

The typical African diet is reliant on a diet of rain fed crops. This fact indicates these impacts will be particularly felt across Africa, more so than other regions (Desanker, 2002). Changes to both actual evapotranspiration (reducing the water retained in soil and surface stores) and increased variability in rainfall events will reduce the reliability and consistency of crop yields. The Tanzanian report on climate change indicates areas which receive 1 annual rain event (ie. The lower latitudes) will experience a reduction in mean annual precipitation. The report continues by stating this will cause a 33% decline in maize yields (a staple crop in Tanzania). Such a statement suggest changes to future rainfall patterns will dominate crop yields in Africa. This hypothesis lies outside this blog post but I shall return to the topic of temperature vs rainfall for crop yields.

I aim to use this information to help accurately analyse the large body of literature using crop models to simulate agricultural yields under climate change scenarios. Over the next few blog posts I intend to continue reading through the mass of academic literature, examine the consistency of study results and assess the application of such studies in aiding policy decisions.

Crop Models: A basic introduction

Crop models have 'cropped' up a few times in my previous posts. Mathews et al., (2007) highlighted the importance of incorporating crop models when considering agricultural policy. This factor has resulted in a boom in crop modelling studies throughout academia and, therefore, provides a great focus for my blog over the next few months.

So if you know me I love a good model. They make sense. We get along. Models are used in a number of different contexts: Hydrological, Climate simulation and Agriculture. Crop models are a recent development, mirroring the trend in increased popularity of modelling in academic geography.

So what is a crop model? (Stay with me, I know this could be a bit intense if you are not a model lover like myself). The aim of a crop simulation model is to estimate crop yields as a function of certain parameters including weather, soil conditions and potential agricultural management policies. This is done by firstly simulating natural conditions. Certain parameters (within the model) are altered in order to create probabilistic future scenarios (such as changes in soil moisture), thus allowing future crop yields (under these scenarios) to be calculated. A classic example of this would be predicting the effects of changes in precipitation, as result of climate change, on agricultural yields (Kang et al., 2009). 

The use of these models is particularly relevant to agricultural in the African regions. Exponential population growth and climate change are both factors which will put pressures on the African agriculture sector. Based on this premises crop models (in combination with development models) should be incorporated into future policy decisions (Dourado-Neto et al., 1998). 

Crop Modelling is becoming more common in academic research (Argawall and Mall, 2002; Tubiello et al., 2000). However, as with all modelling studies, a certain level of uncertainty is associated. This uncertainty is something I wish to address in a later post so will not bore you with the details now.

This was just a brief introduction to the overall aims of crop modelling. Over the coming few weeks I hope to explore the different applications (and coinciding crop models) of agriculture studies, the uncertainty associated with crop modelling and how these models can help achieve sustainable agriculture. 

Keep it Growing

Climate change and exponential population growth have both been identified as pending threats to future agricultural production (Sivakumar et al., 2005). Such pressures have stressed the need for achieving sustainable agriculture. The idea of achieving sustainability in the agricultural sector has been discussed extensively in both academia and politics (Garnett et al., 2013; Netting, 1993). Yet this is still yet to be achieved!

Chapter 14 of Agenda 21 states by the year 2025, 83 percent of the expected global populations will be living in a developing country.  Agriculture has yet to meet the demand of the present day without incorporating future needs. There is an urgency to meet these demands by increasing production of already cultivated land, whilst avoiding further encroachment on land unsuitable for agricultural use.

Figure 1: Timeline of the term Sustainable agriculture. Demonstrating the extent to which it has been discussed on a global scale. 

This topic has been discussed extensively (demonstrated by figure 1). Most recently the ideology of achieving sustainable agriculture has been incorporated within the sustainable development goals (United Nations, 2015). Sustainable Development Goal Two: focuses on ending hunger. It aims to address the issues surrounding hunger, food security and malnutrition through achieving sustainable agricultural. Target 2.3 established issue of increased demand for food production with an aim to double agricultural yields by 2030.

Thoughts and reflections:

The sustainable development goals will have positive implications. Bringing the attention and consequently funding to the issues of agriculture will never be negative. Despite not meeting the targets, the MDGs focused donor funding and encouraged some progress on the topic of hunger (the level of hunger dropping by 27% since 2000) (Sachs, 2012). Hunger is still currently an issue in Sub Saharan Africa as 23.8% of the total population are classed as chronically undernourished (Global Hunger Index, 2015). Furthermore future predictions of population growth (8.5 million by 2030) need to be encompassed into the target of the SDG goal 2.3. The predicted “doubling” in agricultural production will need to incorporate not only the current gap in hunger but also the expected population growth, currently it is unclear if this has been accounted for. However uncertainty, associated with the extent and rate of population growth, may prevent this target from being ever effective in achieving sustainable agriculture and preventing hunger.