Hi Nicolas,

I believe that in this respect, one of the most critical challenges is understanding how the historical dynamic profile of social-ecological landscapes (SELs) evolved in response to internal processes (e.g., plant succession, management practices) and external drivers (e.g., rainfall, temperature, climate change, exchange rates). Humans trying to understand the current state or predict the future condition of SELs regularly resort to simple, easily interpreted surrogates as parts of the whole complexity that can be understood and used by non-scientists to make planning and management decisions. Yet, the overall information we can gain from a set of indicators will never match that of the whole system, since each individual indicator carries only partial information. Thus, the set of indicators needs to be constantly re-evaluated and re-interpreted in the light of the increasing understanding of the whole organization and functioning of systems. Fortunately, the complexity of living systems of people and nature emerges not from a random association of a large number of interacting factors but rather from a smaller number of key-controlling processes (Holling 2001; Gunderson and Holling 2002). Thus, much of the fundamental nature of systems can often be captured and described by single key-variables, as many features of the system’s state tend to shift in concert with a few important key-state variables. Remote sensing is a primary source of information to study the complexity of SESs at the landscape level in terms of dynamics at multiple spatial and temporal scales. It has become a proven tool for scientists to monitor synoptically, and to understand major natural and anthropogenic disturbance events and their historical regimes at regional and global scales It has provided valuable indices to describe and quantify natural and human-related land-cover transformations and processes, and one, in particular, has been widely used: the Normalized Difference Vegetation Index (NDVI). Briefly, NDVI can be used to quantify annual net primary productivity (Young and Harris 2005), which is a main supporting ecosystem service (Millennium Ecosystem Assessment 2005; Costanza et al. 2007). NDVI is broadlyrecognized as a spatially explicit robust indicator of vegetation photosynthesis related to social–ecologicalprocesses such as habitat-land use conversion (e.g.,urban sprawling) or crop rotation (Guerschman et al.2003; Potter et al. 2003; Young and Harris 2005). It is used to identify and assess the impact of disturbances such as drought, fire, flood, frost (Potter et al. 2003; Mildrexler et al. 2007), or other human-driven disturbances (Guerschman et al. 2003; Zurlini et al. 2006; Wylie et al. 2008; Zaccarelli et al. 2008.).Reducing the complexity of SELs to one-dimensional representation of state might seem oversimplified. Nonetheless, vegetation cover is the primary determinant of landscape mosaics and quite responsive to land cover transfor...

Dear Nicolas, although Giovanni is certainly right in addressing the complexity of the interacting human and environmental system at the landscape scale. Nonetheless, I would prefer an approach that first clarifies the information needs at hand. Do you wish to conduct an academic study concerned only with abstract ecological changes or is there a real-world decision-making issue to be solved? I would argue that you need at least a vision of potential decision-making problems in order to determine the best strategy and correspnding methods. Best regards, jan

I did not mean to suggest conducting an academic study concerned with abstract ecological changes. On the contrary, just because it is impossible to get a detailed estimate of all ecosystem services at work and their trade-offs in a certain area, in order to address the estimate of changes in the provision of ecosystem services, it is increasingly recognized that remote sensing indices and, in particular, NDVI-related indices are the most appropriate and easy-to-use methods to get the overall picture. As I said, NDVI is broadly recognized as a spatially explicit robust indicator of vegetation photosynthesis (i.e. a fundamental supporting service) related to social–ecological processes such as habitat-land use conversion (e.g.,urban sprawling) or crop rotation (Guerschman et al.2003; Potter et al. 2003; Young and Harris 2005). It is used to identify and assess the impact of disturbances such as drought, fire, flood, frost (Potter et al. 2003; Mildrexler et al. 2007), or other human-driven disturbances (Guerschman et al. 2003; Zurlini et al. 2006; Wylie et al. 2008; Zaccarelli et al. 2008.). Of course, remotely sensed NDVI does not directly provide estimates of specific ecosystem services like pollination, or water regulation, but because of its strong relationship with land cover, and its synoptic feature it can provide an overall landscape perspective of the ecosystem service flow (Costanza et al. 2010). Once you get the overall picture along with robust high frequency time series at the pixel level, you can then address real-world decision-making issues at different scales and play different games like: where can I put an element (e.g. roads, settlements etc) without interfering with the ecosystem service flow, or can I clean this out (e.g. a part of a forest, a meadow) and what does that imply in term of ecosystem service flow? In those cases, NDVI-related indices can provide basic support for a reasonable answer. In other terms, what can I gain and what am I going to loose in terms of ecosystems services as a result of certain decisions in planning or management? See e.g. http://link.springer.com/article/10.1007/s10980-012-9794-4. In any case, in order to make any decision you need to have a landscape perspective if not larger. So, for example, you could decide to reduce a certain ecosystem service at a location in favour of other services or to foster it in other locations (even in other parts of the world) to counterbalance such reduction or loss. If you want to refine the strategy further you can then make use of economic values for ecosystem services or PES (Payments for Ecosystem Services) methods to make more sound and appropriate decisions through e.g. cost-benefit analyses.

Dear Giovanni, I certainly agree that NVDI and remote sensing are important tools for ecosystem servie analysis. But "...but because of its strong relationship with land cover, and its synoptic feature it can provide an overall landscape perspective of the ecosystem service flow (Costanza et al. 2010)" is simply overstating its capacities. NVDI has a good relationship with land cover, however: many ecosystem services are laregely independent of landcover changes. E.g. it is next to impossible to deduce from NVDI measurements *alone* how the provisioning of irrigation water during the dry season will change downstream. Changes in soil nutrient status or non-timer forest resources cannot be seen. In Kambodia, we have the "empty forests" syndrome where the near extinction of regional fauna is invisible from space. The full set of cultural ecosystem services is extremely difficult to judge by NVDI changes. And so on. Yes, it provides a landscape perspective, but if the perspective provided fits your research needs has to be assessed always. In that sense, I am a bit more reluctant to advertise its "overall" use. All the best, jan

Dear Jan, thank you, I think we can agree on the fact that NDVI related indices and remote sensing in general are not a "panacea". I did not mean it at all after reading Elinor Ostrom's paper, But still I would like to stress the amount of information you can get from such a "bird's eye view" on what is occurring on Earth. Once you get such landscape perspective, you need of course to go in further details at finer spatial scales and see what is going on for specific ecosystem services. Thus you could discover that It can even detect "how the provisioning of irrigation water during the dry season will change downstream" e.g. from the outboost or shrinking of NDVI related to river bank vegetation given that is strongly dependend on solar radiation and water availability . Obviously you will never perfectly detect e.g. what is occurring in a lake or under forest canopy (but see , e.g., "Mapping litter decomposition by remote-detected indicators") neither you will be able to detect the full set of cultural services or soil nutrient conditions under vegetation cover unless they will affect vegetation composition and health (and in this case you could detect it) . I just wanted to stress the enormous potential associated with such methods to measure changes in the provision of ecosystem services that is increasingly confirmed by the outstanding number of applications published in the open literature.

I wish you all the best, Gianni

Dear colleagues. Thank you very much for your answers. I was thinking more about the identification of a subset of ecosystem functions and related services in a study region; then identify potential involved taxonomic groups and related species´functional traits. With this information I was thinking to have an approximation at different stages: (1) conduct field surveys to determine the status of these functions (types) groups in different natural-anthropogenic vegetation gradients (or chronosequences), (2) model species´ potential distribution and see spatial overlap at the identified functional types (eg Maxent, BioMod); and (3) modeling the provision of some ecosystem services on the landscape (e.g. Invest, Swat, Aries). Finally, for stages 2 and 3 we could evaluate the effects of habitat loss on the distribution (of functional types and services) a LULCC multitemporal study. Do you know about studies that have had these approximations?. Thanks