Assessing water stress of desert vegetation using remote sensing : the case of the Tamarugo forest in the Atacama Desert (Northern Chile)

Water stress assessment of natural vegetation plays a key role in water management of desert ecosystems. It allows scientists and managers to relate water extraction rates to changes in vegetation water condition, and consequently to define safe water extraction rates for maintaining a healthy ecosystem. Previous research has shown that optical remote sensing constitutes a powerful tool for assessing vegetation water stress due to its capability of quantitatively estimating important parameters of vegetation such as leaf area index (LAI), green canopy fraction (GCF), and canopy water content (CWC). However, the estimation of these parameters using remote sensing can be challenging in the case of desert vegetation. Desert plants have to cope with high solar irradiation and limited water. In order to maintain an adequate water balance and to avoid photoinhibition, desert plants have evolved different adaptations. A common one is heliotropism or ‘solar tracking’, an ability of many desert species to move their leaves to avoid facing direct high solar irradiation levels during the day and season. This adaptation (paraheliotropism) can have an important effect on the canopy spectral reflectance measured by satellites as well as on vegetation indices such as the normalized difference vegetation index (NDVI). In this thesis, I propose a remote sensing based approach to assess water stress of desert vegetation, exemplified in the case of the Tamarugo (Prosopis tamarugo Phil) tree in the Atacama Desert (Northern Chile), a ‘solar tracker’ species, which is threatened by groundwater overexploitation.

In the first chapter of this thesis (general introduction), I explained the motivation of the PhD project and elaborated four research questions, which are later discussed in chapters 2, 3, 4, and 5. The thesis concluded with chapter 6, where I provide a synthesis of the main results, general conclusions and a final reflection and outlook.

In the second chapter, I studied the effects of water stress on Tamarugo plants under laboratory conditions and modelled the light-canopy interaction using the Soil-Leaf-Canopy radiative transfer model. I described for the first time pulvinar movement of Tamarugo and quantified its effects on canopy spectral reflectance with and without stress. I showed that different spectral indices have potential to assess water stress of Tamarugo by means of LAI and CWC. In the third chapter, I measured the effects of pulvinar movements on canopy reflectance for Tamarugos under field conditions and used high spatial resolution images to assess water stress at the tree level. I developed an automated process to first identify single trees and delineate their crowns, and secondly, to estimate LAI and GCF using spectral vegetation indices. These indices (NDVI and chlorophyll red-edge index) were negatively correlated to diurnal values of solar irradiation as a consequence of leaf pulvinar movements. For this reason, higher values of both vegetation indices are expected to occur in the morning and in winter (low solar radiation) than at midday or summer.

In the fourth chapter I studied the effects of diurnal pulvinar movements on NDVI time series from the MODIS-Terra satellite (acquired in the morning) and the MODIS-Aqua satellite (acquired at midday) for the period 2003-2012 and the seasonal effects of pulvinar movements on NDVI time series of Landsat images for the period 1998-2012 for Tamarugo areas with and without water stress. NDVI values measured by MODIS-Terra (morning) were higher than the NDVI values measured by MODIS-Aqua (afternoon) and the difference between the two, the ΔNDVImo-mi, showed good potential as water stress indicator. In a similar way, I observed a strong seasonal effect on the Landsat NDVI signal, attributed to pulvinar movements, and the difference between winter and summer, the ΔNDVIW-S, also showed good potential for detecting and quantifying water stress. The ΔNDVImo-mi, the ΔNDVIW-S and the NDVI itself measured systematically in winter time (NDVIW) were negatively correlated with in situ groundwater depth measurements.

In chapter five I used a dense NDVI time series of Landsat images for the period 1989-2013, combined with high spatial resolution satellite imagery and hydrogeological records, to provide a quantitative assessment of the water status of Tamarugo vegetation after 50 years of increasing groundwater extraction. The results showed that the NDVIW and ΔNDVIW-S of the Tamarugo vegetation declined 19% and 51%, respectively, as groundwater depleted (3 meters on average) for the period 1989-2013. Both variables were negatively correlated to groundwater depth both temporally and spatially. About 730.000 Tamarugo trees remained in the study area by 2011, from which 5.2% showed a GCF<0.25 which is associated to severe water stress. Based on this spatio-temporal analysis, I suggest that the survival of Tamarugo trees is limited to a maximum groundwater depth of 20 meters.

The main conclusions of this PhD thesis are summarized as follows:

Heliotropism or leaf ‘solar tracking’, a common adaptation among desert plants, has an important impact on canopy spectral reflectance. As shown in the case of the Tamarugo trees, widely used vegetation indices such as the NDVI were negatively correlated to solar irradiation (the stimulus for leaf solar tracking), showing a distinct diurnal and seasonal cycle.An early symptom of water stress in paraheliotropic plants (leaves facing away the sun) is the decline of the amplitude of the diurnal and seasonal NDVI cycles. Thus, remote sensing estimations of this amplitude (e.g. the NDVI difference between winter and summer or the difference between midday and morning) can be used to detect and map early water stress of paraheliotropic vegetation.At the tree level, very high spatial resolution images combined with object based image analysis and in-situ data provided accurate estimations of the water status of small desert vegetation features, such as isolated trees. For monitoring purposes, careful consideration of the time during the day and the season at which the images are taken needs to be taken to avoid misleading interpretations.Time series analysis of historical satellite images combined with very high spatial resolution images and hydrogeological records can provide a quantitative spatio-temporal assessment of the effects of long-term groundwater extraction on desert vegetation.