Introduction

One third of Mongolia’s population is living in the livestock sector, which is based on a traditional nomadic life. This life style is highly dependent on, and influenced by, changes in weather and climate, and is therefore vulnerable to climate change impacts. The nomadic lifestyle of Mongolian herders means that, when affected by climate-induced events, they are at a serious risk of poverty (Climate Change Coordination Office, 2013).

A Dzud (massive mortality of livestock) causes great damage to livelihoods and to the economy of the country. Dzuds are generated by complex climate and social factors like 1) Severely cold weather, 2) Drought in summer, 3) Heavy snowfall, and 4) Livestock management (MARCC 2014). It appears that the severity of dzuds has increased in recent years, which poses a significant threat to Mongolia’s livestock sector (Fig. 2).

Fig. 2: Changes in livestock mortality rate in Mongolia during 1990–2013. Red bars indicate the occurrence of dzuds 

(https://www.theguardian.com/world/2017/jan/05/mongolian-herders-moving-to-city-climate-change).

 

Developing a way to predict the occurrence of dzuds would allow for appropriate preparation for these phenomena and would ultimately result in a significant reduction of animal losses in winters. The Mongolian government has therefore prioritized the development of a dzud prediction system to reduce national economic losses and combat poverty (MARCC 2014). Chuo University, Japan, has developed a dzud prediction system in response to this situation, supported by the Ministry of the Environment of Japan. This system is detailed below.

 

Methodology

The dzud prediction system predicts the occurrences of dzuds (dzud index) between December and April on December 1 every year. The dzud index is calculated from previous observed summer (May–July) temperature and precipitation data, which are obtained from national Mongolian statistical data and future prediction data of winter (December–February) temperature and precipitation by National Centers for Environmental Prediction Coupled Forecast System model version 2 (NCEP CFSv2) output, present pasture consumption, pasture stock, snow depth, and snow cover. The index comprises a scale between 0 and 1 (the higher value, the higher vulnerability), indicating vulnerability of the livestock sector to a dzud.

 

Validating the dzud index system

The dzud index was compared with the observed livestock mortality rates in each province of Mongolia. This index successfully tracked increases in livestock mortality rate, particularly from 2001 to 2002 and in 2010 (Fig. 3). Although there was an overestimation in 1997, the index almost accurately predicted periods of low livestock mortality rate (Fig. 3).

Fig. 3: Comparison between the dzud index and recorded livestock mortality rates in Bayanhongor Buutsagaan during 1991–2014 (Chuo University, 2017).

 

Fig. 4 shows the relationship between the dzud index and ratio of livestock loss range (1992–2015) for all the Mongolian provinces. Further examination of the relationship between the dzud index and livestock mortality rates across Mongolia from 1992 to 2015 showed that in years where the dzud index was 0.4–0.6, the most common livestock loss rate was 10–25% (Fig. 5). In this example, therefore, mid-range values on the dzud index appear to correspond to years with mid-range livestock loss rates (Fig. 4).

Fig. 4: Livestock loss rate distribution with increasing dzud index values from 1992 to 2015 across all Mongolian provinces (Chuo University, 2017).

 

Adjusting the dzud index

Adjustments to the dzud index were implemented to improve the predictive capability of the system based on the validation results detailed above (Figs. 3 and 4). These adjustments were as follows:

1. (1)The dzud index algorithm was fine-tuned to more accurately reproduce the ratio of livestock loss range (1992-2015) across all Mongolian provinces.
2. (2)Extreme cold day components (days with temperatures of < -35 %, < -30 %, and < -25 %) were added to the dzud index algorithm as parameters to reflect extreme cold events.
3. (3)The dzud index was altered to consider cattle-horses and sheep-goats as two separate classes.

 

Result of the dzud prediction

The results of the dzud prediction during December 2016 – April 2017 (calculated in 1 December 2016) are shown in Figure. 5 – Figure. 7.

Figure 5: The dzud index distribution on cows/horses (upper) and sheep and goats (under) during December 2016 – April 2017 (Chuo University, 2017).

Figure 6: Probability of more than 10% death of cows and horses (upper) and probability of more than 25% death of cows and horses (under) during December 2016 – April 2017 (Chuo University, 2017).

Figure 7: Probability of more than 10% death of sheep and goats (upper) and probability of more than 25% death of sheep and goats (under) during December 2016 – April 2017 (Chuo University, 2017).

 

Issues and future developments

The dzud index is calculated using observation data (national statistical data), which could include errors; this adversely affects the system’s predictive accuracy. Increasing the frequency of observation records for each parameter of the dzud index algorithm would help to combat this issue. The version of the system detailed here predicts on December 1 of each year, regarding the future dzud occurrences between December (same year) and April (next year) for four months. However, at present, moment-by-moment changes in meteorological conditions during winter (December–February) are not included in the prediction. This real-time information would be assimilated into the dzud index in an ideal system (Fig. 8).

Fig. 8: Schematic of an ideal dzud prediction system (Chuo University, 2017). CFSv2: Coupled Forecast System model version 2; IRIMHE: Information and Research Institute of Meteorology, Hydrology and Environment; NCEP: National Centers for Environmental Prediction; NEMA: National Emergency Management Agency; NOAA: National Oceanic and Atmosphere Administration.