*2.2. MODIS Data and Processing*

This study applied the use of 1 km spatial resolution, daily intervals MODIS/MOD1B time-series data to evaluate spring wheat yield. The MODIS data provided a high temporal resolution and wide coverage areas but a low spatial resolution. MODIS time-series data during the growing season (June to August) were obtained for the study area from the Atmosphere Archive and Distribution System (LAADS) of (NASA) for 2000–2017. Available online https://ladsweb.modaps.eosdis.nasa.gov (accessed on 20-10-2019). The study area covered over three tiles of granules of MODIS data such as an H24V03, H24V04, and H25V04 (H is horizontal and V is vertical coordinate). All downloaded MODIS 18 years' images were re-projected from sinusoidal to the Albers equal area projection. The reflectance bands (NIR, red, blue, and SWIR) were calculated using MOD021KM level 1b calibrated data. In the processing section, all the collected images were re-projected, mosaicked and calibrated for atmospheric and geometric correction using ENVI IDL software. In this study, the nine vegetation and drought indices (NDVI, NMDI, NDWI, VCI, TCI, VHI, NDDI, VSDI, and VSWI) are calculated from cloud-free and corrected reflectance bands.

## *2.3. Crop Data*

In this study, we have utilized annual spring wheat yield data for eight agrometeorological stations from 2000 to 2017 in Northern Mongolia (Table 1). The sowing stage of spring wheat is generally in the first decade of May. All the crops present an important vegetative development in the June–August period and spring wheat harvest occurs generally in September. The statistics of spring wheat yield were obtained from the agrometeorological division of Information and Research Institute of Meteorology, Hydrology, and Environment (IRIMHE) in Mongolia. Agrometeorological stations measure the crop phenology stage, growth condition and damage, crop density and height for every 10 days from May to September. Finally, the spring wheat yield sown from sampling surveys at the end of the growing season was also measured. Spring wheat yield was collected in 50 × 50 cm plots, in four repeated samplings at each agrometeorological station. From the homologies sample plots, four samples of spring wheat were taken through crop cutting from a different area at equal distance and their average was taken to minimize random errors. Before thrashing and weighting the spring wheat grain yield, the sample plot was placed in the oven for 5–10 min and at 20◦C–25◦C to easily split the grain yield and the straw. The final spring wheat grain weight of each station was converted to 100kg ha−<sup>1</sup> unit. As shown in (Figure 3), the phenological stages of the normal growth cycle of spring wheat and mean climate variables in the study area from May to September (2000–2017). In order to obtain the crop remote sensing indices values correctly, we had to solve image masking for regional spring wheat yield estimation. Applying a cropland mask to select remote sensing indices values as input to a crop yield model significantly improves the accuracy of the crop yield estimation [18,46]. A copy of the crop cover mask was obtained from the land cover map of Mongolia, provided by Elbegjargal et al. [47], and used to reduce the influence of non-agricultural areas on the remote sensing indices signal [48]. Finally, all areas with non-agricultural land were masked out and the regional annual yield was estimated and mapped for only cropland areas yield estimated maps were produced by model (M4) in the study area from 2000 to 2017.


**Table 1.** Information of meteorological stations with location (spring wheat yield available stations).

**Figure 3.** Growing stage of spring wheat and climate variables distribution in the study area.

#### **3. Methodologies**

The purpose of this study is to create a predictive measure of spring wheat yield computed from satellite data. The methodology has three main parts. Firstly, we calculate the nine remote sensing indices using MODIS data. Secondly, nine remote sensing indices are used for the relationship between actual spring wheat yield as input for the testing of the estimation model of spring wheat yield. Thirdly, we develop spring wheat yield regression models and map estimated spring wheat yield for the regional level to the 18 years. The general flowchart of this research, the processing method, and the individual steps are illustrated in Figure 4.

**Figure 4.** Flowchart of processing method.
