Widening geographic range of Rift Valley fever disease clusters associated with climate change in East Africa

Abstract

Introduction

Rift Valley fever (RVF) virus, a mosquito-borne phlebovirus, is known for extreme weather-associated large epidemics characterised by abortions and perinatal mortalities in livestock; and fever, jaundice, encephalitis, retinitis and haemorrhagic syndrome in humans.^{1 2} First isolated in Kenya in 1931, the virus caused sporadic localised clusters of RVF disease during the rainy season until 1951 when a larger epidemic spread the virus to approximately 33% of the country.³ Since the 1960s Kenya has reported >10 major RVF epidemics associated with El Niño/Southern Oscillation (ENSO) heavy rainfall, including the two most extensive regional epidemics (Somalia, Kenya, Tanzania) in 1997–1998 and 2006–2007 that attracted the involvement of international partners including WHO as public health events of international concern.⁴ From the 1970s, periodic major epidemics have been reported in the eastern (Kenya, Somalia, Tanzania), southern (Zimbabwe, South Africa, Madagascar, Mayotte), northern (Egypt, Tunisia) and western (Senegal, Mali, Niger, Mauritania) regions of Africa, almost all associated with flooding conditions.⁴ In 2000, the virus crossed over to the Middle East, through livestock trade, causing a major epidemic in Saudi Arabia and Yemen that resulted in over 3500 confirmed human cases and >200 fatalities.^{5 6}

Recurrent RVF epidemics occur in certain ecosystems where the virus is endemic, following the convergence of several factors including the presence of mosquito vectors, adequate density of naïve livestock and extreme weather conditions that cause excessive flooding to trigger large-scale vector populations.⁷ In East Africa, such permissive ecologies, which were responsible for up to 90% of human cases during epidemics, were characterised by low-elevation plains (<1000 m above sea level) with slow-draining clay soil, and dense bushes or acacia vegetations.⁸ RVF epidemics can also occur in new areas, triggered by massive livestock movement as observed in Africa to Middle East livestock trade during annual Muslim Hajj, or political unrest resulting in large-scale displacement of communities.^{9 10}

Once the virus is introduced in a country or region, it becomes endemic through maintenance by Aedes mosquitoes via transovarial transmission.^{11 12} The virus can also be maintained through continuous low-level cycling in livestock, herbivore wildlife, humans and mosquitoes.^{13 14} During such cryptic cycles, a small number of infected mosquitoes routinely emerge to infect animals and humans; however, this cycling does not reach the epidemic threshold, either because the infected vector population is small and rapidly dies off, or a second vector responsible for the horizontal transmissions that occur in epidemics is not present in sufficient numbers.^{15 16}

There is a growing global concern that changing climate conditions and increasing extreme weather events will transform the landscape of infectious diseases by expanding permissive ecologies for pathogens, dispersing pathogens to new regions and bringing humans closer to pathogens through land use changes.¹⁷ Recent RVF studies pointed to an expanding geographic range and scope of the disease, likely associated with increasing rainfall intensity and or warming temperatures that perhaps broaden ecological areas supportive of virus maintenance.¹⁸ For example, a modelling study conducted in 2017–2018 in the greater horn of Africa (Ethiopia, Kenya, Somalia, Uganda and Tanzania) using data from epidemics, small RVF clusters (≤20 human cases) and seroprevalence findings identified a broader set of risk factors of RVF occurrence, including new soil types (nitisols, vertisols), both low and high elevation areas and highly variable temperatures during the driest months.¹³ In addition, Uganda, which was never involved in severe RVF epidemic occurring in East Africa, started reporting an increasing number of RVF outbreaks from 2016, and significant disease seroprevalence primarily in the highlands of western and central regions of the country.^{19 20} To gain an understanding of the changing RVF epidemiology, we investigated the geoecological and climatic factors associated with the spatiotemporal occurrence of small RVF clusters in Kenya, Uganda and Tanzania between 2008 and 2022.

Methods

Results

Characteristics and spatial distribution of small RVF clusters

We collected data on 100 RVF disease events in East Africa from 2008 through 2022. Of the 100 events, 44 were reported in Uganda, 41 in Kenya and 15 in Tanzania. Among these, 91 were classified as small clusters, characterised by a median of 3 human cases (IQR, 1–3) and 3 livestock cases (IQR, 1–7), with minimal human mortality (IQR, 0–1) as detailed in table 1. 35% of the clusters occurred in regions that had never reported RVF cases before, spanning a broader range of altitudes as low as 77 m to as high as 2867 m above sea level, with a median altitude of 1224 m. In Uganda, 93% of the clusters were situated in altitudes ≥1000 m, with 48% (17/35) in Kenya, and 25% (3/12) in Tanzania (table 1, figure 1). Geographically, 43% (15/35) of the small RVF clusters in Kenya were concentrated in the southwestern highlands, while 75% (33/44) in Uganda occurred in the southcentral highlands (figure 1). The southcentral Kenya and southwestern Uganda regions are agriculturally productive highlands, characterised by intensive and semi-intensive livestock production systems. Overall, small clusters occurred across the entire spectrum of livestock production systems, including intensive, semi-intensive and extensive, whereas RVF epidemics were primarily reported in dry areas with extensive livestock production systems. The majority (65%) of the soils where the small RVF clusters occurred were clayey, specifically nitisols, planosols and ferralsols soil types, followed by loamy (18%), and sandy soil types (15%).

Open in a separate window

Figure 1

Location where small Rift Valley fever clusters reported in East Africa between 2008 and 2022.

Table 1

Characteristics of small RVF clusters in East Africa, 2008–2022

Characteristics	Kenya (n=35)	Uganda (n=44)	Tanzania (n=12)	Total (n=91)
Median no. of human cases (IQR)	3 (2–3)	1 (1–2)	1 (1–6)	1 (1–3)
Median no. of livestock cases (IQR)	3 (2–7)	1 (1–1)	9 (3–14)	3 (1–7)
Median human case mortality	0 (0)	1 (0–1)	0 (0)	1 (0–1)
No. of clusters in new areas	6 (17%)	23 (52%)	3 (25%)	32 (35%)
Types of livestock production systems
Intensive	14 (40%)	3 (7%)	3 (25%)	20 (22%)
Semi-intensive	1 (3%)	38 (86%)	2 (17%)	41 (45%)
Extensive	20 (57%)	3 (7%)	7 (58%)	30 (33%)
Soil types
Clay	23 (68%)	24 (67%)	7 (58%)	54 (65%)
Loam	8 (23%)	3 (8%)	4 (33%)	15 (18%)
Sand	3 (9%)	9 (25%)	1 (8%)	13 (15%)
No. of RVF events with positive NDVI change 1 month prior to cases
Yes	20 (59%)	20 (51%)	4 (36%)	43 (51%)
No	14 (41%)	19 (49%)	7 (64%)	41 (49%)
Median altitude (m) (IQR)	743.8 (1429)	1276 (304)	491 (769)	1224 (994)

Open in a separate window

N=RVF disease events, n=sporadic RVF clusters.

NDVI, normalised difference vegetation index; RVF, Rift Valley fever.

A positive NDVI change 1 month prior to RVF occurrence was observed in 51% of the small clusters. The small RVF clusters were situated in areas with notable land use changes, particularly those leading to significant water accumulation such as rice farming, mining and irrigation. In areas with water pools, either as a result of natural or man-made conditions, 72% of the small RVF clusters occurred. On average 28% of the small RVF clusters were in areas with man-made pools resulting from construction of ponds or dams, swamps, irrigation schemes and excavated mines.

Seasonality and short-term geoecological determinants

When plotted by month of occurrence, the small RVF clusters demonstrated a higher frequency (75/91, 82%) during and immediately after the wet seasons as shown in figure 2. In univariable analysis, a history of previous RVF occurrence, the presence of significant water pools, various types of land use changes, both long-term and short-term temperature variations, and increases in NDVI changes were associated with an elevated IRR with p values of ≤0.05 (online supplemental table 2). Results from the multivariate analysis are presented in table 2. The multivariate model in online supplemental table 3 exhibited a higher AIC indicating a poorer fit compared with the model presented in table 2. Temperature and NDVI demonstrated a significant positive association with the occurrence of small RVF clusters while soil types and elevation were not found to be associated with the changes in IRR. A 1-unit positive change in NDVI a month prior to the disease occurrence, or 1⁰ C temperature rise in the third month prior to occurrence were associated with disease IRR of 1.93 (95% CI 1.01, 3.71) and 1.20 (95% CI 1.1, 1.2), respectively as shown in table 2.

Table 2

Multivariate analysis of potential drivers of RVF clusters in East Africa, 2008–2022

Factor	Coefficient	IRR	95% CI	SE	P value
Intercept	−10.53	–	–	–	–
Tanzania	0.90	2.45	1.07, 5.92	0.42	0.03*
Uganda	−1.22	0.30	0.15, 0.57	0.32	<0.001*
Temperature (third month prior)	0.18	1.20	1.11, 1.29	0.04	<0.001*
Humidity (first month prior)	−0.01	0.99	0.95, 1.02	0.02	0.48
First month prior NDVI	2.29	9.88	0.85, 119.52	1.05	0.03
Positive NDVI change	0.66	1.93	1.01, 3.71	0.30	0.03*
Loam soils	−0.20	0.82	0.39, 1.76	0.36	0.58
Sandy soils	−0.50	0.60	0.21, 1.74	0.48	0.30
Deviance	–	–	DF	–	–
Null deviance	149.52		75	–	–
Residual deviance	76.28		67	–	–
Dispersion parameter	1.00		–	–	–
AIC	423.69		–	–	–
2× log-likelihood	−403.69		–	–	–

Open in a separate window

Theta: 1.002; SE: 0.175

Temperature (third month prior) refers to the temperature of the third month prior to RVF disease event.

Humidity (first month prior) refers to humidity of the first month prior to RVF disease event occurrence, first month prior NDVI refers to NDVI value of the first month before RVF disease event.

*shows statistically significant variables in the model

AIC, Akaike Information Criterion; IRR, incidence rate ratio; NDVI, normalised difference vegetation index; RVF, Rift Valley fever.

Open in a separate window

Figure 2

Seasonal distribution of Rift Valley fever (RVF clusters reported in East Africa between 2008 and 2022.

Supplementary data

bmjgh-2023-014737supp002.pdf

Supplementary data

bmjgh-2023-014737supp003.pdf

Long-term climatic trends in East Africa, 1981–2022

There is substantial spatial heterogeneity in climatological temperature (figure 3A–D) and rainfall patterns (figure 3E–H) across East Africa, partly associated with the topographic complexity of the region (climatology is calculated as the 30-year average over the current normal period 1991–2020 as defined by the World Meteorological Organisation). Annual mean temperatures show substantial warming across the entire region, with larger trends (calculated over 1981–2022) across most of Kenya than in Uganda and Tanzania (figure 4A). We analysed the spatial patterns of trends during the long (March to May) and short (October toDecember) rains seasons; 82.4% of the RVF cluster occurred during and after these seasons. Temperature trends during the long rains varied between 0.12 (^oC/decade) in southwestern Uganda and 0.3 (^oC/decade) in northern Kenya (figure 3B). This warming trend is more widespread and amplified across all three countries during the short rainy seasons (figure 3D). Trends in rainfall are mixed, showing increasing wetness in southwestern Uganda, western Kenya and most of Tanzania but increasing dryness in northwestern Uganda and the expansive low-elevation eastern Kenya region stretching from the coastal south to the arid and semi-arid regions north during the long rainy seasons (figure 3E,F). In contrast, the short rains seasons experienced stronger and more widespread wetting trends across Kenya, Uganda and northwestern Tanzania but strong decreasing trends across southern Tanzania (figure 3G,H).

Open in a separate window

Figure 3

Seasonal temperature and rainfall patterns in Kenya, Uganda and Tanzania.

Open in a separate window

Figure 4

Effect of annual and seasonal temperature and rainfall on occurrence of Rift Valley fever incidences.

Figure 3A,C depicts the long-term climatology (average over 1991–2020) and figure 3B,D illustrates decadal trends (calculated from 1981 to 2022) of mean temperature during the two main rainy seasons—long rains (March to May) and short rains (October to December) seasons over East Africa. Figure 3E–H shows the same as above but for total precipitation. Black dots in figure 3B,D,F and figure 3H represent areas with trends that are significant at the 10% level based on the Wald test.

Annual precipitation trends reflected the patterns of trends in seasonal rainfall, showing increased wetness across southern and eastern Uganda, western Kenya and Tanzania (figure 4B). Another important climate variability parameter is trends in annual number of dry days (<1 mm of rain/day). There has been a decreasing trend in the number of dry days (ie, trend towards more rainy days) across most of Uganda, while the number of dry days is increasing across much of Kenya and parts of Tanzania (figure 4C). Parts of Kenya with a high number of clusters experienced dry day increases of ~3–6 days/decade while parts of Uganda with a high number of clusters experienced dry day decreases of 3–4.5 days/decade, despite increasing trends in annual rainfall (figure 4B,C).

Long-term climate variability in areas with a high density of RVF clusters

Because of limited number of RVF clusters from Tanzania, we use the data from Kenya and Uganda to assess the impact of climatic variability on incidence of RVF clusters. The assessment compared annual and seasonal mean temperature and rainfall trends in two regions; area A in southwestern Uganda where 76% of the RVF clusters of the country occurred (figure 4A—turquoise box), and area B in southwestern Kenya where 30% of in-country clusters occurred (figure 4A—blue box). Time-series of temperature and total rainfall during the long-rains (March to May), intervening dry seasons (June to August) and short-rains (October to December) in the RVF cluster area A and area B were compared (figure 4D–K).

The two areas with a high concentration of RVF cluster areas (figure 4A, area A and area B) in Uganda and Kenya experienced significant warming trends (p<0.05) across all seasons and annually over the past four decades. Area B in Kenya experienced slightly stronger warming than area A (figure 4D–G). For example, during the short-rain seasons, temperature in area A increased at 0.16°C/decade while in area B temperature increased at 0.21°C/decade (figure 4G). Relative to other areas in Uganda, parts of area A experienced weaker annual warming while annual trends in area B were similar in magnitude to trends across other parts of Kenya (figure 4A). The consistent 2°C–3°C higher temperature in Uganda across all seasons (figure 4D–G) is not unique to high concentration RVF cluster sites but rather reflects the normal climatological difference across the two regions.

Annual total rainfall over the high concentrations RVF cluster areas showed increasing trends in area B in Kenya being approximately 1.7 times higher than in area A (figure 4H), although trends are not statistically significant. Similarly, although not significant, rainfall over the area B in Kenya showed a faster increasing trend relative to area A during the dry seasons and short rains seasons but not in the long-rains seasons (figure 4I–K). Trends in the short-rain seasons in area B were >2.5 times higher than in area A.

Top panels decadal trends (calculated from 1981 to 2022) in (figure 4A) annual temperature, (figure 4B) annual rainfall and (figure 4C) annual number of dry days (days with <1 mm/day) overlayed with the RVF clusters (yellow dots) in Kenya, Uganda and Tanzania. Black dots in figure 4A–C represent trends that are significant at the 10% level based on the Wald test. The turquoise and blue boxes in figure 4A indicate RVF clusters in Uganda and Kenya, respectively. Time-series of area-average, mean temperature (figure 4D–G) and rainfall (figure 4H–K) in the RVF cluster area A (Uganda) and area B (Kenya) calculated annual and for the long-rains, dry and short-rains seasons. Numbers in each panel represent the decadal trend magnitude (in °C/decade in figure 4D–G and mm/decade in figure 4H–K and p values of the linear trend in brackets for the corresponding time-series.

A time-series plots of number of RVF clusters in the two hotspots; area A in Uganda (figure 5—top panel) and area B in Kenya (figure 5—bottom panel) during the long-rain seasons showed increasing trend as temperature and rainfall increased in these highland regions.

Open in a separate window

Figure 5

Increasing frequency of Rift Valley fever (RVF) clusters in two highland area of Kenya and Uganda associated with increasing precipitation and temperature.

Discussion

We performed spatiotemporal modelling involving various geoecological and short-term climatic factors associated with the increasing occurrence of small RVF clusters reported across widening geographic regions in East Africa in recent years. In addition, we analysed long-term temperature and precipitation metrics trends in the region over four decades (1981–2022) and evaluated the association of these trends with hotspots of RVF clusters in Kenya and Uganda. The described RVF clusters are characterised by <5 human cases and <10 livestock cases. Notably, 35% of these clusters occurred in geographic areas that had not previously reported RVF disease. The small RVF clusters occurred during or immediately after rainy seasons, highlighting the critical role that mosquito vectors continue to play in the dissemination of the RVF virus. Comparable trends of small RVF clusters have been documented in various African countries in recent years.^{53 54} In contrast to the large epidemics associated with extreme weather and primarily occurring in areas of low elevation, specific soil types and extensive livestock production, the small RVF clusters occurred in diverse ecosystems. These included both highlands and lowlands with varied soil types and encompassing the entire spectrum of livestock production systems (intensive, semi-intensive and extensive). Notably, a substantial proportion of clusters in Uganda (85%) and Kenya (55%) occurred in the highlands where mixed farming with intensive and semi-intensive livestock production was prevalent.

When we evaluated >40 independent variables for association with RVF occurrence, comparable short-term and long-term climatic factors emerged as primary drivers shaping the evolving landscape of RVF disease. Warmer temperatures and rainfall 1–3 months prior correlated with a higher RVF IRR, with 80% of the clusters occurring during this phase. Factors crucial for extreme weather-associated RVF epidemics like excessive flooding, low elevation, clayey soil types and extensive livestock production systems were not significant for small RVF cluster occurrence. Our regional analysis of long-term climatic trends revealed an annual mean temperature increase of up to 0.12–0.3°C/decade in parts of East Africa, indicating a rise in annual average temperatures of up to 1.2°C over the past 40 years. These climatic variations have directly contributed to escalating frequency and severity of droughts, particularly in the arid and semi-arid lowlands such as northeastern Kenya.^{55 56} Our findings on climatic trends analysis reveal increasing rainfall trends in scattered highlands of the regions. Importantly, our data indicates that areas in the highlands of southwestern Uganda and Kenya experiencing rising temperature and rainfall trends alongside diverging trends in dry days coincide with significantly increasing frequency of RVF clusters. The disease landscape in Uganda more vividly reflects the impact of climatic variations. Despite not being part of regional RVF epidemics and having no documented significant human RVF clusters for decades, the country has reported over 60 small RVF clusters across the country since 2016.

Our findings suggest that the combination of increasing temperature, increasing rainfall trends; and a decrease in the annual number of dry days is a driving factor behind the escalating public health risk of RVF disease in Uganda. With ongoing global warming, East Africa is projected to undergo changes in temperature and precipitation patterns with continued increases in seasonal and annual mean temperatures, particularly in highland areas. Seasonal and annual mean temperatures are expected to keep rising with the most significant increases occurring in the highland areas.⁵⁷ Even with a 2°C of global warming aligning with the Paris Climate Agreement goals, annual temperatures in the region may increase by 1.5°C–2°C accompanied by over 40% increase in the number of days with temperature above 35°C.⁵⁸ In terms of precipitation, the observed wetting during the short rain seasons is projected to persist, accompanied by an increase in heavy rain event whereas the changes in rainfall patterns during the long rain seasons remain uncertain.^{59 60} The likely impact of these future climatic trends is a heightened frequency of small RVF clusters across the geographic areas described in our study, with a reduced frequency of occurrence for the extreme weather-associated large epidemics in the region’s lowlands.

The global impact of the spreading RVF clusters includes a wider dispersal of the virus leading to the creation of new regions with virus endemicity. This escalation increases the risk of more widespread epidemics or pandemics in the future. Changing climate conditions and drivers of natural climate variability such as ENSO that causes extensive heavy rains and flooding across Africa and other regions can trigger widespread virus amplification and human infections.^{60 61} This is particularly concerning in rural Africa and Middle East where there are significant populations of livestock to support virus amplification and enhanced human infections through consumption of animal products.^{60 61} Studies have shown widespread distribution of Aedes mosquito vectors globally, including Europe, Australia and North America.⁶² It is crucial for global public health institutions to consistently approve human RVF vaccine therapeutics and implement swift distribution plans through regional vaccine banks.^{63 64} Fortunately, the highly conserved RVF virus genome and minimal antigenic variation observed during epidemics guarantee the high efficacy of vaccines and immunotherapeutics.^{65 66}

The study has limitations. First, reliance on available data introduces potential bias because of the incompleteness of some of the data points and under-reporting in regions with limited health infrastructure. In addition, the presence of potential multicollinearity of variables and the use of presence-only data should be considered when interpreting the data. Conversely, an increase in disease reporting rate over time and in new geographic areas is difficult to disentangle from increases in actual disease prevalence.

While we acknowledge these potential limitations, our analyses suggest distinct ecological and climatic risk factors underlying the changing epidemiology of RVF disease in East Africa, which should be investigated for direct causal relationship. Specifically, our study highlights the increasing frequency of small RVF clusters in previously unaffected areas, associated with a combination of higher temperature and rainfall.

Acknowledgments

Footnotes

References