Green Revolution in the Peruvian desert

Localization
Peru
Date
2013-04-22
Author
Category

Can the Terra-i tool only detect areas of natural vegetation loss? The answer is: “No”! Indeed, increases in the normalized difference vegetation index (NDVI) associated with vegetation growth, either natural or introduced, can also be detected. This has being explored for the first time through a case study in a hotspot region of Piura, Peru, where pilot runs have demonstrated the use of NDVI increases in identifying vegetation growth. Analyses revealed a gradual change of habitat from semi-arid lands with little to no vegetation, to heavily vegetated areas covered with industrial sugarcane crops. Additionally, both scenarios of coverage - disturbed and undisturbed - were successfully isolated based on a tilted time series analysis technique called "Seasonal trend decomposition by LOESS".

Figure 1. Irrigation system technologies have improved water availability and therefore been associated with vegetation increases on desert or sparse lands, as the case study in the provinces of Paita and Sullana in the Piura region of Peru showed. Photo: Paolo Imbrauglio.

Pilot runs explored how the "Increase" product of the Terra-i tool can detect increases in vegetation greenness intensity or the NDVI value. These pilots were carried out across different areas of Latin America, and hot spot regions were identified based on the rapidity of the changes in their landscape. One such hot spot region included our case study area: the provinces of Sullana and Paita in the Piura region of Peru.

The Sullana and Paita provinces, located mainly within the Tumbes-Piura dry forest ecoregion, are characterized by a desert landscape with little or no vegetation and minimal rainfall [1]. In addition, the potential soil use in this region is very limited and sporadic, only favouring vegetation growth in periods of abrupt changes in rainfall patterns (i.e. during the “El Niño” phenomenon) [2].

Nevertheless, according to the "Increase" detections from the Terra-i system this region has shown a gradual increase of vegetation due to anthropogenic disturbances, particularly when the analysis includes sugarcane crops grown for ethanol production projects (Figure 2). A recent report prepared by FAO [1] sets the figure at around 5,757 hectares of sugarcane crop identified from 2009 to 2010, an area that is projected to increase to 10,000 ha or more in the coming years. The companies “Caña brava” and “Maple”, which have acquired the vast majority of surface area in the Piura region, are notable for their capacity to implement modernized irrigation technologies, supporting large-scale agricultural activities and greatly impacting the presence of natural coverage.

Figure 2. At left, a map of land-use change detections in the region of Piura (Peru) resulting from the Terra-i tool’s “Increase” product, showing changes from January 2004 to October 2012 (light to dark green spots). The points sampled to describe the behavior of NDVI time-series are represented by orange squares and yellow triangles, indicating points of natural coverage with and without anthropogenic disturbance, respectively. At right, pictures of the two areas with the greatest extent of sugarcane crops correspond to the “Caña Brava” (A) and “Maple” (B) companies, as well as the irrigation technology (C) used by these companies. Photos SPDA (A and C); Dallas Business Journal (B).

In addition to an exploration of vegetation increase, an in-depth analysis and interpretation of the NDVI time series used by Terra-i highlighted and compared the evolution of greenness over time between areas of natural coverage with and without anthropogenic intervention. The "Seasonal trend decomposition by LOESS" [5, 6] technique was used to discriminate both situations through the resulting trend (Xw) and seasonality (Xh) components.

To illustrate the meaning of the Xw and Xh components in terms of coverage behaviour (natural and disturbed), an annual crop (i.e. corn) can have a high value of seasonality (Xh), as it starts and finishes its cycle during the course of a single year, and a low value of trend (Xw), as an area dedicated to agriculture does not have large variations in weather except for extreme climatic events. On the other hand, a rainforest would present a lower value of Xh as the tree canopy remains green throughout the year, a factor that also accounts for almost 80% of the trend component (Xw).

Applying these concepts to our case study, it was possible to satisfactorily differentiate the behavior of the two components for natural areas with and without intervention (Figure 3). Strong seasonal behavior (Xh) was evident at the moment in which Terra-i detected increases in NDVI values, characteristic of the presence of annual crops (in this case, sugarcane). Also, the NDVI presented a positive trend component (Xw), as it passed from a bare ground situation to being densely covered by the crop. It is expected that this behavior will continue for a period of 4-5 years, until the crop’s productive conditions are stabilized. On the other hand, bare ground presented values close to 0 for the seasonality (Xh) and trend (Xw) components, associated with low values of vegetation greenness intensity (NDVI ≤ 0.2).

Figure 3. Evolution of the NDVI, Xh and Xw average from a sample of 10 pixels showing the transformation from natural coverage to an industrialized sugarcane crop (dashed lines) and an abrupt change in 2007 detected by the Terra-i system (yellow arrow). The behavior of the same component average for undisturbed natural cover (solid lines) is also displayed, using the same amount of sampled pixels.

In conclusion, the exploratory analysis of the "Increase" product for this case study successfully identified changes caused by vegetation increase. For this case land-use change was gradual, passing from areas with limited to no vegetation cover (natural habitat) to areas with fully industrialized sugarcane crops. Collaborative work with Pablo Vázquez, visiting researcher from INTA (Argentina), was a first step towards achieving classification of the land-use changes that Terra-i detects.

The Terra-i team kindly invites its users to explore—in addition to the "Increase" product— other outputs such as "Decrease" and "Floods", which are available for download on the Terra-i website. Also, the team expresses its interest in participating and providing support to research initiatives that are considering more detailed explorations and analyses of the tool’s results. Users can contact us at our email or in the contact us section on the website.

The article was authored by Paula Paz, Alejandro Coca, Louis Reymondin and Pablo Vázquez. Revision of English-language version by Caitlin Peterson (CIAT / CCAFS visiting researcher)

References

[1] SNV. Inclusión de pequeños agricultores en la cadena productiva de caña para etanol y productiva de caña para etanol y certificación ISCC. 2011. [Online] http://www.snvworld.org/sites/www.snvworld.org/files/publications/20111017_informe_final_fao_cana_brava.pdf

[2] BISA. EIA Proyecto agroindustrial de producción de etanol automotor. 2007. [Online] http://www.maple-energy.com/downloads/ENVIRONMENT/EIAProyetanol/INFORME%20%20FINAL-EIA%20ETANOL.pdf

[3] CIPCA. Actualización del mapa regional del sector agrario en Piura. 2011. [Online] http://www.cipca.org.pe/publicaciones/Estudios%20y%20Folletos/ESTUDIO%20MAPA%20REGIONAL%20AGRARIO.pdf

[4] Revista Agro negocios Peru. Articulo “Perú tiene los proyectos de caña de azúcar con riego tecnificado más grandes del mundo” . Publicacion 2012. [Online] http://www.agronegociosperu.org/noticias/121112_n1.htm

[5] Cleveland, R. B., Cleveland, W. S., McRae, J. E., & Terpenning, I. (1990). Stl: a seasonal-trend decomposition procedure based on loess (with discussion). Journal of Official Statistics. 6: 3-73. [Online] http://cs.wellesley.edu/~cs315/Papers/stl%20statistical%20model.pdf

[6] Hua Lu, Michael R. Raupach, Tim R. McVicar and Damian J. Barrett. 2003. Decomposition of vegetation cover into woody and herbaceous components using AVHRR NDVI time series. Remote Sensing of Environment. 86: 1-18. http://www.clw.csiro.au/publications/technical2001/tr35-01.pdf

 

Important articles Analysis category