WaterSubject

Irrigation performance and water productivity can be benchmarked if estimates of spatially distributed yield and crop water use are available. A commonly used method to estimate crop evapotranspiration in irrigated areas is to multiply reference evapotranspiration values by appropriate crop coefficients. This study evaluated convenient ways to derive such coefficients using multispectral vegetation indices obtained by remote sensing. Detailed ground radiometric measurements were taken in small plots perpendicular to the crop rows to obtain canopy reflectance values. Ancillary measurements of green ground cover, plant height, leaf area index and biomass were taken in the cropped strip covered by the radiometer field-of-view. The results were up-scaled using 10 Landsat-5 and 1 Landsat-7 images. Crop measurements and ground radiometry were made at the time of Landsat overpass on two commercial fields, one grown with sugarbeet and the other with cotton. Crop height and ground cover were determined weekly in these two fields, three additional sugarbeet fields and one additional cotton field. The ground and satellite observations of canopy reflectance yielded similar results. Two vegetation indices, the normalized difference vegetation index (NDVI) and the soil adjusted vegetation index (SAVI) were evaluated. Both indices described the crop growth well, but SAVI was used in further evaluations because it could be conveniently related to both ground cover and the basal crop coefficient using a simple model. Based on these findings, crop water use variability was analyzed in a large sample of sugarbeet and cotton fields, within a homogeneous irrigation scheme in Southern Spain. The yield versus evapotranspiration data points were highly scattered for both cotton and sugarbeet. The yield values obtained from the sugarbeet fields and cotton fields were substantially lower than values predicted by a linear yield function, and close to a curvilinear yield function, respectively.

In the last 50 years, a substantial amount of research on irrigation and crop water management has focused on yield responses to water supply and/or evapotranspiration. The water–yield function represents the maximum yield for a given amount of water in the absence of biotic or abiotic stresses other than water. In practice, weeds, diseases, poor nutrition, and other problems adversely affect yield in addition to water stress. Thus, when yield data are plotted against evapotranspiration or water supply data obtained from commercial fields, most of the data points typically fall well below the yield function curve. Quantifying the among-farmers variability of water use and benchmarking water productivity in irrigated areas are the first steps for identifying the causes of this “yield gap” and deciding where improved crop management is needed.

In order to benchmark water use and water productivity successfully, accurate estimates of crop evapotranspiration and its spatial distribution are required. Numerous studies have assessed the utility of remote sensing techniques for estimating crop evapotranspiration at large scale. Some of these studies have employed signals in the thermal band obtained from remote sensors as inputs for energy balance equations that are solved to estimate evapotranspiration. A second approach is based on the FAO method for estimating crop evapotranspiration, in which reference evapotranspiration values are multiplied by crop coefficients. These coefficients may be derived from multispectral vegetation indices (VI) obtained by remote sensing.

The second approach for estimating crop evapotranspiration requires fewer inputs and theoretical background knowledge, thus it is simpler than the first approach, but it ignores reductions in evapotranspiration due to stomatal closure resulting from water deficits. Thus, an assumption underlying this approach is that variations in the size of the crop have much stronger effects on its evapotranspiration than variations in stomata conductance.

This assumption seems to be valid for irrigated crops, according to studies presented by several authors. The results of these studies have shown the strengths and potential of this approach for operational use by farmers. However, there is little agreement on the nature and generality of the relationships between crop coefficients and spectral vegetation indices.

Crop coefficients can be estimated from spectral measurements because the basal crop coefficient (K_cb, the component of the crop coefficient that represents transpiration) and the vegetation indices are both sensitive to leaf area index (LAI) and ground cover fraction. Some authors have suggested that relationships between K_cb and VI are linear, but others have found non-linear relationships. According to theoretical relations derived the linearity of these relationships depends on the crop architecture and the definition of VI applied.

Analyses of the relationships between vegetation indices, defined in various ways, and both LAI and f_c may facilitate the formulation of a comprehensive relationship between K_cb and VI. The cited work of supports this hypothesis, as do various recent studies on the relationships between transpiration and f_c or LAI, in orchards and vines. Moreover, FAO-56 guidelines recommend that K_cb values should be adjusted to account for variations in effective ground cover to obtain site-specific crop coefficients=.

In this paper, we present data on the relationships between spectral vegetation indices and crop growth variables obtained at different scales for two contrasting row crops, sugarbeet and cotton, and examine the validity and feasibility of using satellite-derived vegetation indices for determining evapotranspiration crop coefficients. Finally, we apply this methodology to assess water use variability and benchmark water productivity within a large irrigation scheme in Southern Spain.