Fig 1.
Map of study area indicating the distributional data for the leishmaniases (black circles) used to parameterise environmental models (a) cutaneous leishmaniasis (n = 803 squares) (b) visceral leishmaniasis (n = 201 squares).
Grey squares indicate the locations of pseudo-absence data points for one iteration of the model. Tick marks on the x and y axes indicate degrees latitude and longitude respectively and the study grain is 5 arc minute squares.
Fig 2.
The trajectory of changes in climate and land use under alternative future socio-economic storylines, climate change pathways and policies.
A matrix of six alternative scenarios is depicted, 3 SSPs in columns x 2 climate change pathways in rows. Each scenario was realised in terms of mapped values of cover of CLUE land use categories and Worldclim climate variables in 2005 (to reflect the recent past) and 2050.
Table 1.
Potential environmental predictors of leishmaniasis distribution.
Table 2.
Percentage contribution of top ten ranked predictors to models of cutaneous leishmaniasis where different sets of predictors were considered (averaged across 20 sub-models).
Table 3.
Percentage contribution of predictors to models of visceral leishmaniasis where different sets of predictors were considered (averaged across 20 sub-models).
Table 4.
Mean accuracy statistics for models of cutaneous and visceral leishmaniasis for models considering only abiotic predictors versus those considering mammal richness alongside abiotic predictors.
A. cutaneous leishmaniasis. B. visceral leishmaniasis.
Fig 3.
Predicted (a) mean and (b) standard deviation of the relative probability of presence of CL and (c) the sum of times CL is predicted to be present, across 20 runs of models built with abiotic variables only.
Fig 4.
Geographical variation in selected (a) climate and (b) landscape predictors of leishmaniasis distribution in the recent past period (2005).
Fig 5.
Predicted (a) mean and (b) standard deviation of the relative probability of presence of VL and (c) the sum of times VL is predicted to be present across 20 runs of models built with abiotic variables only.
Fig 6.
The predicted extent of (a) cutaneous and (b) visceral leishmaniasis under recent past conditions and under alternative climate change pathways and socio-economic storylines.
The heavy blacklines across the middle of the box indicate median predicted extent across the 20 model runs; the box indicates the interquartile range of the data whilst the whiskers indicates the extremes. Note that the threshold that best predicted disease presence in cross-validation across model runs was used to assign pixels to presence or absence classes (values in Table 3).
Fig 7.
Change in the predicted presence of cutaneous leishmaniasis between the recent past and future (2050) conditions under alternative climate change and socio-economic scenarios.
This was calculated by subtracting the number of times a pixel was predicted as present in the recent past (across 20 model runs) from the number of times a pixel was predicted as present in the future. Areas in pink show areas that are likely more favourable for disease and areas in blue show areas that are less favourable for the disease in the future.
Fig 8.
Change in the predicted relative probability of presence of visceral leishmaniasis between the recent past and future (2050) conditions under alternative climate change and socio-economic scenarios.
This was found by subtracting the number of times a pixel was predicted as present in the recent past (across 20 model runs) from the number of times a pixel was predicted as present in the future. Areas in pink show areas that are likely more favourable for disease and areas in blue show areas that are less favourable for the disease in the future.