A statistical study of NDVI sensitivity to seasonal and interannual rainfall variations in Southern Africa

Abstract

The relation between rainfall and the Normalized Di erence Vegetation Index (NDVI) in Africa south of 15ß S ( 1983± 1988) is studied. For 115 1ß by 1ß grid-points, the spatial distribution of annual NDVI and rainfall means is highly comparable. Both parameters have overall decreasing values from Mozambique to South-Western Africa. The strongest correlations occur when NDVI monthly values are compared with the bimonthly preceding rainfall amounts, attesting a time response of one to two months. At these time and space scales, NDVI does not appear to be sensitive to the seasonal and interannual rainfall variations in the Namib desert, South Namibia and western Cape Province. Along the Indian Ocean coast, it is weakly sensitive to the seasonal cycle only. It becomes largely sensitive to the seasonal cycle in Zimbabwe, and in South-Western Zambia. A high sensitivity to the interannual rainfall variability is only observed on the Southern African Plateau, around the Kalahari basin. Multivariate analyses show that geographical conditions of seasonal and interannual rainfall-NDVI associations strongly di er. While a sensitivity to seasonal rainfall is observed in areas where mean annual rainfall varies from 300 to 900 mm and where the contrast between wet and dry seasons is strong, the sensitivity to interannual rainfall anomalies is observed only for relative dry areas, where mean annual rainfall varies from 300 to 500 mm. In both cases, the relation is observed whatever the soil type or vegetation formation.

Effects of precipitation, temperature and topographic parameters on evergreen vegetation greenery in the Western Ghats, India

“…For example, several studies reported the relationship between NDVI to rainfall to be no longer sensitive to rainfall variations beyond a given rainfall threshold, particularly in wet tropical areas (Davenport and Nicholson, 1993;Wang et al, 2003). This rainfall amount has been shown to be 200 mm month −1 over equatorial Africa (Richard and Poccard, 1998), East Africa at 1200 mm/yr −1 (Nicholson and Farrar, 1994) or 600 mm/yr −1 (Fuller and Prince, 1996) above which NDVI curve saturates. In the case of the Western Ghats, the rainfall range during the wettest quarter is from 601 to 5490 mm/yr −1 , which is far higher than the values reported above, suggesting the saturation of NDVI.…”

“…With respect to NDVI response to rainfall, several studies reported that vegetation does not respond to immediate rainfall, rather it is affected by the history of soil moisture buildup (cumulative rainfall) (Davenport and Nicholson, 1993;Wang et al, 2003). For example, Richard and Poccard (1998) studied the sensitivity of NDVI to seasonal and inter-annual rainfall variations in southern Africa and reported the strongest correlations when NDVI monthly values are compared with the preceding bi-monthly rainfall amounts, attesting to a time response of one to two months. Further, their analysis using multivariate statistics, suggested differences in rainfall-NDVI associations based on geographical conditions, Farrar et al (1994) found that while the correlation between NDVI and precipitation is highest for a multimonth average, NDVI is controlled by soil moisture in the concurrent month.…”

Seasonally-managed wetland footprint delineation using Landsat ETM+ satellite imagery

“…Unlike the near-instantaneous change in temperature corresponding to the phase transformation of water to vapor, NDVI displays a delayed response to water availability. The time lag between increased moisture (via any combination of precipitation, irrigation, or exfiltration fluxes) and plant response has been shown to range from two months (Richard and Poccard, 1998) to fewer than five days (Wang et al, 2007). Eklundh (1998) found statistically significant time lag periods difficult to establish, concluding that consistency may be region-dependent.…”

Influence of Indian Ocean zonal circulation on variability and predictability of Ethiopia highlands vegetation

“…To better understand how summer rainfall affects crop production at state-level, past work has compared time series of farm yields, rain-gauge and satellite data (Tucker, 1979;Eklundh, 1998;Richard and Poccard, 1998;Al-Bakri and Suleiman, 2004;Budde et al, 2004). Remotely sensed color reflectance is used to quantify vegetation fraction since 1982 (Bannari et al, 1995), with bias correction for sensor degradation, orbital drift and atmospheric contamination (Chappell et al, 2001;Kawabata et al, 2001;Mennis, 2001;Tucker et al, 2005), and intercomparisons with crop yield (Lewis et al, 1998;Maselli et al, 2000).…”

“…Remotely sensed color reflectance is used to quantify vegetation fraction since 1982 (Bannari et al, 1995), with bias correction for sensor degradation, orbital drift and atmospheric contamination (Chappell et al, 2001;Kawabata et al, 2001;Mennis, 2001;Tucker et al, 2005), and intercomparisons with crop yield (Lewis et al, 1998;Maselli et al, 2000). Rainfall in semi-arid zones drives vegetation growth at seasonal and multi-annual time scales (Budde et al, 2004;Eklundh and Sjöström, 2005;Li et al, 2004;Vanacker et al, 2005) with a response lag of one to two months (Davenport et al, 1993;Eklundh, 1998;Richard and Poccard, 1998). Correlations between vegetation fraction and rainfall are higher for running sums and in zones with a single wet season of 600 to 1000 mm (Davenport et al, 1993 White, 1997;Herrmann et al, 2005;Vanacker et al, 2005;Zhang et al, 2005); whereas vegetation sensitivity to climate in desert and monsoon regions is limited.…”