Rice yield and nutrient balance in southwest China
Data files
Sep 22, 2023 version files 35.47 GB
Abstract
The optimal application of nutrients, such as nitrogen and phosphorus, to the soil is crucial for achieving high crop yields with minimal environmental impact. However, the effect of spatio-temporal changes in soil nutrient supply on crop yield is poorly understood in China. Here, we present a framework that combines environmental data, fertilizer field experiments, and machine learning to estimate the rice yield responses to different nutrient conditions and overall farmland nutrient sustainability in southwest China from 2009 to 2019. This dataset contains data on the spatial distribution of rice yield forecasts and farmland nutrient balances in Southwest China in 2009 and 2019.
README: Rice yield and nutrient balance in southwest China
https://doi.org/10.5061/dryad.b8gtht7jv
The optimal application of nutrients, such as nitrogen and phosphorus, to the soil is crucial for achieving high crop yields with minimal environmental impact. However, the effect of spatio-temporal changes in soil nutrient supply on crop yield is poorly understood in China. Here, we present a framework that combines environmental data, fertilizer field experiments, and machine learning to estimate the rice yield responses to different nutrient conditions and overall farmland nutrient sustainability in southwest China from 2009 to 2019. This dataset contains data on the spatial distribution of rice yield forecasts and farmland nutrient balances in Southwest China in 2009 and 2019.
Description of data and file structure
The spatial resolution of the raster data is 30m. The data format is gdb/tiff and the file names mean the following:
- FNSZ: farmland nutrient sustainability zones (1: Degradation Risk Zone; 2: Sustainable Zone; 3: Excess Risk Zone).
- Spatial distribution of Soil organic matter (g/kg), Soil pH, Soil available phosphorus (mg/kg), Soil available potassium (mg/kg), Nitrogen fertilizer rate (kg/ha), Phosphorus fertilizer rate (kg/ha), Potassium fertilizer rate (kg/ha) in 2009 and 2019: MO2009, MO2019, pH2009, pH2019, AP2009, AP2019, AK2009, AK2019, FRN2009, FRN2019, FRP2009, FRP2019, FRK2009, FRK2019.
- Spatial distribution of regression coefficients of soil-based yield GWRK model: YS_GWRK_C_Intercept, YS_GWRK_C_Soil OM, YS_GWRK_C_Soil AP, YS_GWRK_C_Soil AK, YS_GWRK_C_Soil pH, YS_GWRK_C_Altitude, YS_GWRK_C_Slope, YS_GWRK_C_PRE, YS_GWRK_C_TEM, YS_GWRK_C_NDVI, YS_GWRK_C_Residual.
- Spatial distribution of regression coefficients of relative yield GWRK model: YF_GWRK_C_Intercept, YF_GWRK_C_Soil OM, YF_GWRK_C_Soil AP, YF_GWRK_C_Soil AK, YF_GWRK_C_Soil pH, YF_GWRK_C_Altitude, YF_GWRK_C_Slope, YF_GWRK_C_PRE, YF_GWRK_C_TEM, YF_GWRK_C_NDVI, YF_GWRK_C_Residual.
- Spatial distributions of the soil-based yield and the fertilized yield (kg/ha): YS2009, YS2019, YF2009, YF2019.
- Spatial distributions of rice Nitrogen, Phosphorus, and Potassium fertilizer utilization efficiency: REN, REP, REK
- Spatial distributions of rice Nitrogen, Phosphorus, and Potassium theoretical fertilizer rate (kg/ha): TFRN2009, TFRN2019, TFRP2009, TFRP2019, TFRK2009, TFRK2019.
Version updates will be made as needed with such updates noted in the metadata.
Methods
Crop soil-based yield (YS) and fertilized yield (YF) can effectively reflect the contribution of farmland inherent soil productivity and fertilization for rice yield, respectively. Here, using 1027 field experiments (plain: 213, hill: 611, mountain: 203) as training samples, we established GWRK models to simulate the effects of soil and fertilizer nutrient supply on YS and YF of rice. YS refers to rice yield without fertilization in the current year, expressed as the yield of treatment 1 (N0P0K0) of the "3414" experiment. The YS is important for two reasons: firstly, it shows the spatial distribution of rice yield reduction under the hypothetical condition of no fertilizer use; and secondly, the fertilizer increased yield (YF-YS) provides the possibility to calculate the theoretical fertilizer requirement of rice in order to assess the nutrient balance. The independent variables (explanatory variables) corresponding to the YS model include 9 environmental factors: soil OM (X1), pH (X2), soil AP (X3), soil AK (X4), PRE (X5), TEM (X6), NDVI (X7), altitude (X8), and slope (X9).
YF refers to rice yield under normal fertilization conditions (representative fertilization rate and cultivation management of local farmers). Since YF significantly depends on YS, we adopted the relative yield coefficient (RYC) as the dependent variable fitting model to improve the prediction accuracy, which is expressed as the ratio of the yield of treatment 1 and treatment 6 (N0P0K0/N2P2K2) in the "3414" experiment. Based on the YS model, the RYC model adds 3 fertilization quantity factors: FRN (X10), FRP (X11), and FRK (X12), increasing the number of independent variables to 12. Finally, YF is calculated on the basis of RYC and YS. Fertilizer increased yield (FIY) is used to express the increase in rice yield due to fertilizer application by subtracting the soil base yield from the fertilizer applied yield; fertilizer contribution ratio (FCR) was used to measure the contribution of fertilizer to rice yield. Finally, the nutrient balance rates were input into the SOFM model to assess the farmland nutrient sustainability in the Sichuan Basin.