Remote sensing for landscape epidemiology : spatial analysis of plague hosts in Kazakhstan

Abstract

The spatial distribution of hosts is a crucial aspect for the understanding of infectious disease dynamics. In Kazakhstan, the great gerbil (Rhombomys opimus) is the main host for plague (Yersinia pestis infection) and poses a public health threat, yet their spatial distribution is unknown. Great gerbils are social animals that live in family groups in burrows in desert environments. Plague is transmitted from gerbil to gerbil via fleas, and can also be transmitted to humans. In Kazakhstan, an abundance threshold exists above which plague can spread over larger areas. This means that the great gerbils’ distribution is crucial to the spread and persistence of this disease.

To gain information on the spatial distribution of the great gerbils, the spatial distribution of their burrows needs to be known. High-resolution satellite images were used in combination with a classification technique called Random Forests to identify the burrows of the great gerbils, which resulted in overall accuracies ranging from 87 to 97%.

Once the burrow distribution was known, the relation between the abundance of great gerbil burrows and the landscape was studied. Multiple regression showed that burrow density was negatively related to Greenness, especially in the floodplain areas.

The spatial distribution of the host species is thought to influence the plague dynamics, such as the direction of plague spread, however no detailed analysis existed on the possible corridors and barriers that were present. Corridors and barriers were mapped using a burrow density threshold, and were mostly aligned on the NWSE axis. To investigate whether plague spread was radially symmetric in the past, a plague presence and absence data set was used. Results showed that plague spread had occurred mostly along the NWSE axis. This associates great-gerbil-burrow density with the direction of plague spread.

In plague-prediction models, great-gerbil burrows are assumed to be randomly or regularly distributed. However, this assumption has never been validated. Spatial point-pattern statistics were used on field data to see whether the occupied burrows are clustered within the total population of burrows. The results showed that burrows irrespective of occupancy were regularly distributed. It also showed that occupied burrows are clustered, but that this can only be detected in squares of 500 m and larger.

To be able to predict plague outbreaks in Kazakhstan effectively, the real-time distribution of occupied burrows needs to be known, preferably for large areas. An NDVI time-series was used to identify occupied burrows, and classify them from space. Classification accuracies are promising with an overall accuracy of 65%.

It is concluded that great gerbil burrows irrespective of occupancy are spatially structured and related to the landscape at larger scales, and that they are regularly distributed at local scales. Occupied burrows are spatially clustered. This should be incorporated into plague models. Furthermore, with the newly developed method, occupied burrows can now be detected from space. This will aid the prediction of plague outbreaks. Together, these findings allow for more realistic approaches to disease ecology models for both this system and for other structured host populations.