The Resilience of Annual Vegetation Primary Production Subjected to Different Climate
The resilience of annual vegetation primary production subjected to different climate change scenarios
Rakefet Shafran-Nathan *a, Tal Svoray a, Avi Perevolotsky b
a Dept. of Geography and Environmental Development, Ben-Gurion University of the Negev, Beer-Sheva, 84105, Israel.
b Dept. of Agronomy and Natural Resources, The Volcani Center,
P.O. Box 6, Bet Dagan 50250, Israel
Submitted to Climatic Change
* Corresponding author at:
Tel: ++972 8 6479054;
Fax: ++972 8 6472821;
e-mail:
2 Materials and methods
2.1 The productivity model
We used a GIS-based productivity model that operated daily in each 25-m2 grid cell. The model is based on fuzzy algebra and simulates the effects of solar radiation, hydraulic conductivity, rock coverage, rainfall, evaporation and temperature, on primary production processes of herbaceous vegetation. The model was fully described by Svoray et al. (2008), therefore we present only its basic formulation (Eq. 1):
[1]
That is to define how much each cell X = {x} is related to set (A) (Svoray et al. 2004). The membership function defines the degree of membership of each cell or attribute in the fuzzy set (A) (Burrough et al. 1992; Robinson 2003). To integrate environmental and climatic membership values a joint membership function is calculated (Eq. 2). The weights (λs) determine the hierarchy of the contributions of the various ecosystem controllers to the new fuzzy set that is created in the new prediction layer. Eq. 2 shows the simulation of water-availability conditions (mijt), as calculated on a daily basis for all grid cells.
in which RD is rainfall depth [mm]; TMP is air temperature [0C], EVP is the daily evaporation rate [mm]; RAD is the radiation flux [W/m2]; DEF is the soil moisture deficit [mm]; RC is the rock cover [%]; and RDE is the soil moisture storage [mm].
The model predicts the ANPP of herbaceous vegetation during the course of a growing season; it covers the period starting immediately after the first rainfall event of the season, which usually occurs in October–November. Germination- conditions are accumulated () on a daily basis until conditions for germination in a grid cell, relative to a threshold value, are fulfilled. If a cell “has germinated”, the model starts to accumulate daily conditions for production (), until the end of the growing phase.
2.2 From model validation to long-term data
Model predictions () were validated against recorded long-term (tens of years) biomass harvests in both sites. Dry standing biomass yields, at the end of the growing season, were used as indicators (Thrash 1998) for validating the model predictions of production processes. Because of year-to-year variations in seasonal rainfall distribution, the exact end of the growing season differs among years. Total dry biomass was measured in 30-48 (depending on year) randomly selected plots at the DME site and in 48 permanent plots at the SAE site. Each studied plot extended over at least (75 × 75) m2 – a (3 × 3)-cells kernel window in the model – and represented a specific habitat in terms of spatial uniformity of soil moisture resources. The harvest-sampling method (Ungar et al. 1999) was used and grazing activities were excluded from the plots, which ensured that weather effects formed the dominant determining factor. The plots were digitized to enable geometric correspondence between average harvested biomass and average model prediction in each plot.
Validation was based on linear regression analysis between harvested biomass samples and model predictions. This procedure assumed that if there was a significant association between harvested biomass and accumulated model predictions, regression coefficients could be used to transform model predictions to biomass values (Paruelo et al. 1997). This transformation was considered to be valid only for years in which vegetation samples were harvested. However, occurrence of a sequence of seasons with reliable validation data enables the relationship between predictions and biomass to be applied to seasons that did not have their own validation data. Model predictions of vegetation production were validated against harvested samples taken from the two sites, at the end of 15 growing seasons (1994-2008) at the SAE site (0.63 < R2 < 0.85; p < 0.0001; Table 1), and after six growing seasons (2003-2008) at the DME site (0.53 < R2 < 0.66; p < 0.001; Table 1). A new linear distribution was calculated for each group by means of the GLM (General Linear Model; STATISTICA 8). The new regression coefficients enabled conversion of model predictions to biomass values. The model simulated a 30-year period (1979-2008) at the SAE site and 21 years (1986-1990; 1993-2008) at the DME site (Fig. 1). The model also was subjected to an exhaustive uncertainty and error propagation analysis (Livne and Svoray 2010).
References
Burrough PA, Macmillan RA, Vandeursen W (1992) Fuzzy classification methods for determining land suitability from soil-profile observations and topography. J Soil Sci 43:193-210
Livne E, Svoray T (2010) Components of uncertainty in primary production model: The study of DEM, classification and location error. Int J Geog Info Sci 25:473–488
Paruelo JM, Epstein HE, Lauenroth WK, Burke IC (1997) ANPP estimates from NDVI for the central grassland region of the United States. Ecology 78:953-958
Robinson VB (2003) A Perspective on the fundamentals of fuzzy sets and their use in geographic information systems. Trans GIS 7:3-30
Svoray T, Shafran-Nathan R, Henkin Z, Perevolotsky A (2008) Spatially and temporally explicit modeling of conditions for primary production of annuals in dry environments. Ecol Model 218:339-353
Thrash I, (1998) Impact of water provision on herbaceous vegetation in Kruger National Park, South Africa. J Arid Environ 38:437-450
Ungar ED, Perevolotsky A, Yonatan R, Barkai D, Hefets Y, Barham H (1999) Primary production of natural pastures of the hilly northern Negev contributory factors and management implications. Ecol Environ 5:130–140 (in Hebrew)
Table 1. Linear regression R2 and significance (*P < 0.05; ** P < 0.005; *** P < 0.00005) for validation analysis between cumulative model daily scores and harvested vegetation samples in the SAE and the DME.
Fig. 1. Long-term model predictions of ANPP (grid cells, black curve) and annual precipitation: (A) during 30 years in the SAE, and (B) during 21 years in the DME.
Fig 1
6