| Maize is one of the most important crops planted all around the world. In China, maize has the highest yield and great significance in ensuring national food security. Chinese corn-growing areas are mostly located in arid and semi-arid regions. Among them, two-thirds of cultivated corns rely on natural precipitation. In the context of global climate change, the threat of drought is increasing. Soil moisture is an important component of the soil-vegetation-atmosphere continuum(SPAC), which is a key factor determining the water status of terrestrial ecosystems, and is also the main sources of water supply for crops. The precise monitoring of soil moisture is crucial to drought early warning as well as daily irrigation management.A widely used drought monitoring method in previous studies, which combines remote sensing data including surface temperature and vegetation index, is an eigenspace method. It could mainly be applied to regional drought monitoring with the strict requirements of the study area covering all kinds of scenarios from bare soil to closed canopy, from extreme drought to saturation, and the relative consistent weather conditions. The temporal and sp atial scales are commonly more than 10 days and the area of a province respectively. These methods provide relative water deficits comparing to adjacent areas, instead of the actual soil water content. Thus, it is urgent to develop a new method for estimat ing available soil moisture at a station level based on the data from single station with a long time span.In order to estimate soil moisture at different soil depths at a station level, based on the energy balance equation on the earth and the water def icit index(WDI), a soil moisture estimation model was established on the hypothesis that evapotr anspiration deficit ratio(i.e. WDI) linearly depends on soil relative humidity. Thus, the soil moisture could be estimated in terms of the remote sensing data(the normalized difference vegetation index(NDVI) and surface temperature(Ts)) and the ground observation(air temperature Ta).The soil water estimation model was validated based on the data from the drought process experiment on summer maize(Zea mays L.) responses to different irrigation treatments carried out during 2014 at Gucheng Eco-agrometeorological Experimental Station(39° 08′N, 115° 40′E) of China Meteorological Administration. The drought processes were established by setting up 7 different irrigation treatments(one-time irrigation amount of 10 mm, 30 mm, 60 mm, 90 mm, 120 mm, 150 mm and 225mm) at the trefoil stage of summer maize. The model was used to estimate soil relative humidity of different soil depths(0-50cm) in terms of the parameters from the genetic algorithm and nonlinear programming algorithm. The results indicated that the soil water estimation model developed in this paper was able to evaluate soil relative humidity of different soil depths in t he summer maize ecosystems, and the hypothesis is reasonable that evapotranspiration deficit ratio(i.e. WDI) linearly depends on soil relative humidity. Cross-validation of the soil water estimation model was conducted using leave-one-out method, and the estimation results were evaluated by the determination coefficients(R2), mean bias error(MBE), mean absolute error(MAE), relative maximum absolute error(RMAE), root-mean-square error(RMSE), percentage root-mean-square error(PRMSE) and constitutive relation error(CRE).The experiment established a soil moisture gradient within the 7 treatments, which resulting in different degrees of drought starting successively from trefoil stage to jointing stage. The occurrence and development of drought caused a delay in developing stages and prolonging of the tasseling-silking period, which shortened the grain filling stage, thus resulted in yield loss. Comparing to canopy water content, plant leaf area and transpiration rate, the remote sensing data(i.e. NDVI and surface temperature) showed more sensitive to soil water stress for the value divergence among different treatments. The fitting and cross validation results of the estimation model verified the rationality of the hypothesis. It showed that the estimation accuracy of 0-10 cm surface soil moisture was the highest(R 2= 0.90). The correlation of the estimated and measured soil relative humidity in deeper soil layers(up to 50cm) could pass the significance test at the level of 0.001 with the RMAEs less than 15% and the PRMSEs less than 20%.This research was a meaningful attempt to retrieve soil moisture from remote sensing data and the ground truth observation. The results would provide reference for drought detecting and monitoring as well as irrigation management. However, current research was based upon station scale experiment and single species(i.e. Zea mays L.). The regional applicability remained to be further investigated. |
PDF Full Text Request