We focus our analysis on maize production in both countries, as maize is the main staple food in Lesotho, constituting 77% of agricultural production and 80% of the rural diet14. During normal years, around 30% of the domestic demand for maize is satisfied by national production, with the remaining being imported duty-free from South Africa through the Southern African Customs Union free trading zone.
The 2007 drought was characterised by anomalously low precipitation (compared to 1979–2018) over the eastern part of South Africa and Lesotho (Fig. 1a). during the agricultural growing season (January–February–March, JFM)14,15. Although temperature was also anomalously high (1 in 5-year event), we focus on JFM total precipitation anomalies as the dominant climatic driver of maize yield during this event given the severity of the precipitation anomaly and the predominant rainfed agriculture in this region14,15,16. We averaged precipitation data over Lesotho and the main maize growing region of South Africa (boxes in Fig. 1a). From reanalysis data (ERA517) over 1979–2018 (see “Methods”), we estimate the 2007 event to be the most extreme JFM total precipitation deficit on record in both countries (Fig. 1b,c), resulting in the most severe co-occurring drought on record (Fig. S1a). Figure 1d and e show the maize production time series (FAOSTAT18) in both countries, and Fig. 1g shows the maize deficit (minimum maize demand for sufficient calorie intake minus domestic production) for Lesotho. Lesotho has seen a large, nonlinear, trend in maize deficit over the last decades (Fig. 1g, black line), driven by increasing population, soil erosion, limited agricultural expansion (only 13% of the country is arable land), poor land-use practices, decreasing soil fertility, and the high number of HIV/AIDS infections that reduces the labour supply14,19,20,21. This has increased the reliance on South Africa for maize imports to meet its national cereal requirement. Precipitation is highly correlated between South Africa and Lesotho (ρ = 0.94, 1979–2018), making production anomalies in both countries also correlated (ρ = 0.38, 1981–2013), although this has changed over the years (S1 and Fig. S1b). The fact that Lesotho and South Africa are in close proximity to each other, and hence subject to the same meteorological conditions and possible future changes in the climate system, makes synchronous maize failures likely to happen in this region now and continuously in the future.
Figure 1
Overview of 2007 drought and impacts. (a) Spatial distribution of the JFM precipitation anomaly over South Africa (relative to 1979–2018 average) derived from reanalysis data17, with the boxes indicating the areas that the climate data is averaged over. (b) JFM precipitation time series for South Africa with the 2007 event highlighted in red. (c) Same as (b) but for Lesotho. (d) Maize production data from the FAOSTAT database18 with the 2007 event in red. (e) Same as (d) but for Lesotho. (f) Price per tonnes for maize in South Africa from FAOSTAT. Note only data from 1991 onwards is available. (g) Maize deficit (minimum maize demand for sufficient calorie intake minus domestic production) over time in Lesotho, with the black line showing the non-linear trend line over the years, derived using a lowess function. Figure (a) was generated using the ‘Basemap’ package (https://matplotlib.org/basemap/index.html) and Python Programming Language (version 3.7).
In 2007, the maize production was reduced by 40% in Lesotho compared to 2006, whereas in South Africa, 2007 was the second consecutive production failure, with both 2006 and 2007 witnessing production of 31% below average (over the 1981–2013 period). This reduced the maize available for export to Lesotho (~ 2% of production in South Africa), resulting in a 35,500 tonnes of maize shortage in Lesotho14. The production of the two substitute crops, sorghum (20% of cropped area) and wheat (10% of cropped area), also decreased by 42% and 4%, respectively, compared to 200614,15, limiting the dietary substitution to alternative cereal crops. Although sorghum is a more drought tolerant crop14,15,22, farmers still decide to plant maize during dry years because of dietary preferences for maize and because markets for sorghum are less well developed19. On top of the production loss, the maize price in South Africa, which drives maize prices and hence purchasing power in Lesotho, increased by 41% compared to 2006 and 100% compared to 2005 (Fig. 1f). With the vast majority of small scale farmers in Lesotho (~ 60% population) not being self-sufficient, many households are vulnerable to price spikes as they rely on the grain markets to buy maize. In particular, the poorest households, that spend up to 65% of their total expenditure on staple foods21, experienced disproportionate impacts14. Overall, the combined shortage and price spike caused imminent food insecurity for 400,000 people in Lesotho, approximately 20% of the total population.
Influence of climate change
We perform a multi-method and multi-model EEA analysis13,23 to develop a synthesis assessment on the role of CC in the occurrence of the drought event in both countries and the co-occurrence of it. EEA compares the likelihood of a given extreme weather event occurring in the actual world (ACT) as observed with its likelihood of occurrence in a counterfactual world (NAT) without human influence on the climate system, the ratio between the two likelihoods being the change in probability, or risk ratio (RR), of an event due to CC. We employ a multi-method and multi-model approach to account for both model-related uncertainties and uncertainties related to methodological assumptions23. The RRs for each country’s drought are derived using long-term observational data (CRU-TS24) as well as a range of climate models (weather@home25,26, HadGEM3-A27, as well as ETH-CAM4 and MIROC5 from the HAPPI experiment28) (see “Methods”). Moreover, for the joint probability of occurrence, we use the four climate models with a sufficient number of simulations to approximate the tail-end of the distribution (weather@home, HadGEM3-A, ETH-CAM4, MIROC5). The joint probability of the 2007 event in Lesotho and South Africa is derived by analysing the bivariate distribution of the drought in these two countries.
We estimate that CC made the 2007 drought event 5.36 (10–90%: 1.51–32.50, Fig. S5a) times more likely in Lesotho and 4.70 (10–90%: 1.53–26.30, Fig. S5b) times more likely in South Africa. The lower tails of the joint exceedance distribution (dry–dry) for the four models are empirically derived and shown in Fig. 2a–d, which are based on the joint distribution functions for the ACT and NAT (Fig. 2e–h). A clear shift can be observed for three out of four models, and overall the likelihood of occurrence of the synchronous drought event increased by a factor 2.14 (10–90%: 1.42–3.16, Fig. S5c) due to CC.
Figure 2
Joint probability plot of the 2007 drought. (a) The lower tail (dry–dry) joint probability plot in the weather@home model with red showing the ACT world and blue showing the NAT world without CC. The horizontal grey bars indicate the 10–90% uncertainty estimates of the event using bootstrapping, with the black marker showing the mean estimate. The tails of the distribution are derived empirically. (b–d) Same as (a) but for HadGEM3-A (b), ETH-CAM4 (c) and MIROC5 (d). (e) Joint probability plot, with contour lines derived from kernel density estimates, in the weather@home model. The black marker indicates the 2007 event. (f–h) Same as (e) but for HadGEM3-A (f), ETH-CAM4 (g) and MIROC5 (h). The uncertainty around the 2007 event is only visible for weather@home given the larger uncertainty in this model, whereas for the other models the uncertainty is not clearly visible relative to the size of the marker.
The role of climate change in food insecurity
We then build a probabilistic model using a set of regression models (see “Methods” for validation) that predicts the production and price anomalies (deviation from detrended time series) based on the precipitation anomalies. We combine this with an approximation of export fraction (percentage of export over production in South Africa) from South Africa to Lesotho (see “Methods”), which varies from ~ 0.9% during normal years to up to 2% during dry years (e.g. 2007). Food availability ( 0 surplus) is quantified as the difference between Lesotho’s domestic maize deficit and what is available to import from South Africa. The total imported food value is expressed as the maize deficit in Lesotho and the price of maize in South Africa, which is used as a cumulative indicator of the impacts of prices on household purchasing power. Using the derived RRs, we can construct many plausible counterfactual scenarios of the food security situation without CC which we use to stress-test the system. Stress-testing, referring to exploration of the vulnerabilities of a system based on many plausible scenarios29, is used to evaluate the sensitivities of the synchronous maize failures to climate shocks, and its implications for food security in Lesotho.
We sample 50,000 possible scenarios using the probabilistic model. The spread in the model predictions includes the uncertainties in estimating maize production based on precipitation and, for the NAT simulations, the uncertainties in the fraction of exports from South Africa to Lesotho, which are sampled between observed bounds (0.5–2.5%). Figure 3a shows the results of the probabilistic model, together with the range of the distribution. Our probabilistic model approximates the observed food shortage with a mean of -54,086 tonnes (10–90%: − 119,477 to 11,288 tonnes, observed shortage is within the 68th percentile of the ACT distribution). The NAT distribution plot (blue) has a mean of 8,583 tonnes (10–90%: − 85,802 to 122.760), indicating that a large-scale shortage would have been less likely in a world without anthropogenic CC. For illustrative purposes, we contrast the climate-induced effects to the existing chronic trend of increasing food insecurity in Lesotho and create a hypothetical maize production time series that does not exhibit a declining production trend (NAT + no decline). As can be observed from Fig. 3a, this has a considerably larger effect on the shortage than CC, with a mean of 118,846 tonnes (10–90%: 30,153 to 238,991 tonnes). Figure 3b tracks the probability of maize surplus for four different ranges of fractions of export (F.E.) from South Africa to Lesotho. The probability of maize surplus varies from 9.6 to 87.9% for the NAT simulations and 96.8–100% for the NAT + no decline simulations across the difference F.E.. The distributions of the ACT and NAT in the upper left panel of Fig. 3b (under the F.E. scenario of 0.5–1.0%) are almost identical, indicating that the anthropogenic CC effect is approximately equal to a 1.0–1.5% change in export fraction (compared to the original 2.0% in the ACT). Figure 3b also illustrates the large sensitivity of Lesotho’s food shortage to the amount of production available for export to Lesotho from South Africa. This sensitivity is nonlinear as a less severe drought leads to less crop losses in both countries, and a smaller F.E. need to cover this deficit. Moreover, a less severe drought (less maize loss in Lesotho) also means a higher probability that maize loss can be substituted with available sorghum and/or wheat crops (provided they can fully substitute the sufficient calorie intake). Figure 3c depicts the value of imported maize, which is estimated to be 34.7 million USD (10–90%: 26.2 to 43.9 million USD) in the ACT, 21.9 million USD (10–90%: 11.5 to 32.9 million USD) in the NAT, and -1.7 million USD (10–90%: -12.3 to 8.6 million USD) in the NAT + no decline world (negative value meaning that households would be able to consume or sell the surplus maize). Given that an average rural household spends 175 USD on food and beverages annually30, the average per household expenditure on imported maize (total value of imported maize divided by the number of households in 2007) to cover basic food needs constitutes 50.1% (10–90%: 40.4 to 60.2%) in the ACT world and 31.7% (10–90%: 19.6 to 43.8%) in the NAT world. In other words, purchasing power per household is ~ 37% lower in the ACT compared to the NAT. In particular, small-scale farmers would be hit hardest, as their self-sufficiency level is decreased due to lower yield, making them increasingly reliant on food markets (that see higher prices). Using data from the 2009/2010 Agricultural Census of rural households in Lesotho31, we approximate the exposure of rural farmers to the combined effect of lower production and price volatility (see “Methods”). 2009/2010 is a relatively neutral year, with no large precipitation, production or price anomaly, and the area of maize planted was almost equal to 2007. Based on the size of the farmland that households operate on (which are aggregated into 13 groups of field sizes), the household size, the yield, and maize price estimates, we evaluate the average household expenditure per group of farmers to cover their minimum consumption requirement. Furthermore, we estimate the aggregated number of households that are self-sufficient, i.e. household production is larger than minimum consumption requirement. As shown in Fig. 3d, the ACT simulations increase the average household expenditure for farmers relative to the reference period. The slope of the ACT relative to the reference period is steeper, indicating that a large share of farmers experienced a non-linear increase in expenditure. This showcases that the majority of rural farmers are affected by a combined effect of production loss and price increase, changing a large number of households from net sellers to net buyers. The number of rural households that are self-sufficient (Fig. 3e) shifts from 48.2% (10–90%: 35.3 to 54.7%) in the reference period to 14.9% (10–90%: 7.0 to 22.8%) in the ACT and 33.0% (10–90%: 15.1 to 54.7%) in the NAT world. Although CC made the drought more severe, which negatively impacted rural households’ self-sufficiency, the NAT world simulations still show a larger number of households that are not self-sufficient relative to the reference period of a relatively neutral year. Therefore, climate variability, whether or not driven by CC, induces strong year-to-year variability in the level of self-sufficiency of rural farmers. The ability to cope with such variations in expenditure to purchase staple foods varies per household and depends on availability of alternative labour income, remittances (main source of income for up to 10–15% of households in some districts30), social safety nets and the ability to reduce or change consumption patterns14,21.
Figure 3
Illustrative plot showing the sensitivity of the food security situation in Lesotho to the 2007-like drought event. (a) Probability density plots of the availability of maize to cover minimum consumption needs using the probabilistic model for the ACT (red) and NAT (blue). The horizontal bars at the top illustrate the mean (black marker) and 10–90% uncertainty range of the distribution. The grey distribution plot is the NAT result added to a scenario without no trend in maize deficit in Lesotho. (b) A breakdown of plot (a) under four different ranges of the fraction of exports (F.E., exports from South Africa to Lesotho over total production South Africa) with the numbers in blue and grey representing the probability of having a positive maize availability value. (c) The monetary value of maize imported from South Africa to cover minimum consumption needs in Lesotho. A negative value indicates that the minimum consumption is satisfied, and the surplus maize can be either consumed or sold. (d) The cumulative distribution plot of the household expenditure for 13 groups of farming households with varying sizes of farmland operation. The black line is based on the 2009–2010 Agricultural Census of rural households, and the red and blue lines show the model results for the statistical model using the ACT and NAT conditions. The thick line is the mean value and the filled shading the 10–90% uncertainty range based on the 50,000 realizations of the model. (e) The fraction of rural households that are self-sufficient in meeting their minimum maize requirement for the reference period (2009–2010), compared to those under ACT and NAT conditions.
Source link : https://www.nature.com/articles/s41598-021-83375-x
Author :
Publish date : 2021-02-16 08:00:00
Copyright for syndicated content belongs to the linked Source.
