**1. Introduction**

Food security is an important topic for every country in the world [1]. Accurate and timely estimation of the spring wheat yield on regional and national scales is becoming absolutely essential for developing countries like Mongolia. In particular, crop yield estimation and the monitoring of crop production can provide fundamental information for crop producers, decision-makers in planning harvest and for agricultural development overall [2]. The agriculture sector is the second contributor to the Mongolian economy after mining [3]. However, only 13% of agricultural production is sourced from crops, mostly spring wheat, the remaining 87% is from the livestock [4] since the Mongolian climate is more suitable for extensive grazing, which covers more than 80% of the total land area. The spring wheat is below 1% of the total land area and around 1.35 million hectares of the total land is suitable for crop cultivation [5]. The northern part of Mongolia has the most favorable natural conditions and a more suitable area for rain-fed crops [6]. Hence, most of the spring wheat is grown in the northern provinces due to above-average precipitation. However, precipitation can only support the basic water requirement of spring wheat, and little variation in precipitation would cause a big fluctuation in crop yield. The vegetation cover, crop yields, and their growth are highly dependent on the amount of precipitation and the related soil moisture [7,8]. Mongolia has an extreme continental climate, with a short growing season, high evaporation, and low precipitation, which pose serious limitations for the Mongolian agriculture development. Because of the high altitude, our country's climate is much colder than other countries in the same latitude. More than 80% of total spring wheat cultivation is rain-fed and only 5000 hectares is irrigated for spring wheat in Mongolia [4]. Therefore, agricultural production is particularly sensitive to climate variability and climatic conditions make agriculture very challenging. Due to the impacts of climate change, more extreme and continued droughts have occurred in many parts of Mongolia and have directly affected the vegetation and crop growth, biodiversity and socioeconomics in Mongolia [9]. Nanzad et al. [10] found that about 41–57% of Mongolia has been ravaged by mild to severe droughts for many of the last 17 years. A consecutive severe drought in 2002, 2005, 2007, 2010, 2013, and 2015 lowered spring wheat production severely as shown in Figure 1 and spring wheat had to be imported as local production declined due to weather conditions [4].

**Figure 1.** Whole Mongolian grain production with study area (National Statistics Office, 2019).

Weather information is normally used to forecast crop yield, but there is a lack of continuous measurement among others due to cost factors. Using Earth observation satellite imagery for monitoring temporal and spatial variation, combined with the point observation as a co-monitoring has advantages. Furthermore, satellite imagery is produced at a lower cost than the traditional way and is more easily accessible for use [11,12]. The use of remote sensing data helps to assess crop conditions in different fields at regional and whole country levels, even in remote areas, as it gives a timely and accurate measurement. Therefore, there have been many attempts in the applications of remote sensing in crop yield estimation, monitoring and mapping and most of these work streams indicated that remote sensing technology was prospective and promising [13–21]. A number of field studies have shown that models based on remote sensing data enable to estimate crop yield in many countries. Usually, remote sensing derived indices are connected to crop yield using empirical regression-based models [22,23]. During the past decades, remote sensing has been broadly used in forecasting crop yield. The Advanced Very-High-Resolution Radiometer (AVHHR) is the most popular sensor, the most widely used in terms of crop monitoring and yield forecasting since the early 1980s for a large scale [24,25]. In recent years, satellite-derived data such as Moderate Resolution Imaging Spectroradiometer (MODIS), Landsat, and Sentinel data were used for the yield prediction and monitoring and meaningful results have been obtained [11,12,26–28]. Lewis et al., [29] used AVHRR-NDVI data for maize production forecasts and correlated results showed that forecasts could be obtained one month before the harvest. In Spain, Vicente-Serrana et al. [30] combined AVHHR-NDVI data and drought indices and were able to predict wheat and barley yield four months before harvest. Moreover, Peterson [12] found the best timing to predict crop yield was from two to four months before the harvesting using NDVI, EVI, and NDWI of MODIS for different crops in Africa. Recently, some remote sensing indices such as the normalized multiband drought index (NMDI), vegetation supply water index (VSWI), and visible and shortwave infrared drought index (VSDI) were utilized in a number of studies for drought and crop monitoring and crop yield estimation according to previous studies [31–34]. The more promising method is using crop growth modeling that incorporates updated crop biophysical parameters such as leaf area index (LAI) and a fraction of absorbed photosynthetically active radiation (fPAR) retrieved from satellite imagery and by using survey information of crops throughout the growing season in local to regional areas. For example, Huang et al. [35] found that more accurate county-level winter wheat estimation was obtained using the WOFOST-PROSAIL model. Furthermore, many researchers have developed crop growth models to estimate crop yields [36–40]. However, the crop growth models require more specific information, such as daily weather data, soil properties, and crop growth determining factors, which would make analytical costs excessive. It is obvious that no general indicator can be used to predict crop yields in all regions. The applicability of the indicator will vary with the region, crop type, and crop growth stage.

Some recent studies in Mongolia were conducted to monitor the cropland cover changes, to assess land degradation for the agricultural region. Erdenee et al. [5] have used Landsat TM and ETM data in the detection of changes cropland over Tsagaannuur, Selenge provinces from 1989 and 2000. Otgonbayar et al. [41] investigated to prepare a cropland suitability map of Mongolia using Landsat and MODIS (MOD13, MOD15, and MOD17). Furthermore, Enkhjargal et al. [42] used MODIS and SPOT time-series remotely sensed data from 2000–2013 to estimate long-term soil moisture content in agricultural regions of Mongolia. Ariya [43] used Landsat images from 2000 and 2015 to assess land degradation for the agricultural area of Mongolia. Nevertheless, to date, no studies of crop yield estimation using remote sensing indicators have been done yet in Mongolia. Recently, remote sensing indicators are employed to monitor the drought across the pasture lands of Mongolia [44] and provide valuable information for drought management and reduction. Compared to drought, more attention should be paid to crop yield, while these indices were not getting enough attention in the field of crop yield in Mongolia. The small variation of precipitation would cause the big fluctuation of crop yield so that it is very important to forecast spring wheat yield early for food security in Mongolia. Although spring wheat accounts for a small proportion of Mongolia's land area, as the size of the spring wheat field is large enough that it provides the possibility of predicting spring wheat yields based on remote sensing technology in Mongolia.

Therefore, our analysis focuses on Selenge and Darkhan provinces of Mongolia due to the availability of high-quality spring wheat yield data for those regions. It is the first attempt to estimate the spring wheat yield using space observation technology in Mongolia. The main objectives of this study were as follows: (i) to evaluate the potential of using remote sensing nine indices to estimate spring wheat yield; (ii) to choose the more suitable remote sensing indices for predicting spring wheat yield; (iii) to identify the best timing and more accurate model to estimate spring wheat yield in Northern Mongolia.
